diff --git a/src/include/86box/vid_8514a.h b/src/include/86box/vid_8514a.h
index b2932c725..1f601e149 100644
--- a/src/include/86box/vid_8514a.h
+++ b/src/include/86box/vid_8514a.h
@@ -19,7 +19,16 @@
 #ifndef VIDEO_8514A_H
 #define VIDEO_8514A_H
 
+typedef struct {
+    int ena,
+        x, y, xoff, yoff, cur_xsize, cur_ysize,
+        v_acc, h_acc;
+    uint32_t addr, pitch;
+} hwcursor8514_t;
+
 typedef struct ibm8514_t {
+	hwcursor8514_t hwcursor;
+	hwcursor8514_t hwcursor_latch;
     uint8_t pos_regs[8];
 
     int force_old_addr;
@@ -29,8 +38,10 @@ typedef struct ibm8514_t {
 
     uint32_t vram_size;
     uint32_t vram_mask;
+    uint32_t pallook[512];
 
     PALETTE   vgapal;
+    uint8_t   hwcursor_oddeven;
     uint8_t   dac_mask;
     uint8_t   dac_status;
     uint32_t *map8;
@@ -38,7 +49,9 @@ typedef struct ibm8514_t {
     int       dac_pos;
     int       dac_r;
     int       dac_g;
+    int		  dac_b;
     int       internal_pitch;
+    int 	  hwcursor_on;
 
     struct {
         uint16_t subsys_cntl;
diff --git a/src/include/86box/vid_svga.h b/src/include/86box/vid_svga.h
index ab8144a77..2456f40d6 100644
--- a/src/include/86box/vid_svga.h
+++ b/src/include/86box/vid_svga.h
@@ -89,6 +89,7 @@ typedef struct svga_t {
     int dac_pos;
     int dac_r;
     int dac_g;
+    int dac_b;
     int vtotal;
     int dispend;
     int vsyncstart;
@@ -341,6 +342,9 @@ extern void    ati68860_ramdac_set_render(void *priv, svga_t *svga);
 extern void    ati68860_ramdac_set_pallook(void *priv, int i, uint32_t col);
 extern void    ati68860_hwcursor_draw(svga_t *svga, int displine);
 
+extern void    ati68875_ramdac_out(uint16_t addr, int rs2, int rs3, uint8_t val, void *priv, svga_t *svga);
+extern uint8_t ati68875_ramdac_in(uint16_t addr, int rs2, int rs3, void *priv, svga_t *svga);
+
 extern void    att49x_ramdac_out(uint16_t addr, int rs2, uint8_t val, void *priv, svga_t *svga);
 extern uint8_t att49x_ramdac_in(uint16_t addr, int rs2, void *priv, svga_t *svga);
 
@@ -396,6 +400,7 @@ extern float   tvp3026_getclock(int clock, void *priv);
 
 #    ifdef EMU_DEVICE_H
 extern const device_t ati68860_ramdac_device;
+extern const device_t ati68875_ramdac_device;
 extern const device_t att490_ramdac_device;
 extern const device_t att491_ramdac_device;
 extern const device_t att492_ramdac_device;
@@ -409,6 +414,9 @@ extern const device_t bt485a_ramdac_device;
 extern const device_t gendac_ramdac_device;
 extern const device_t ibm_rgb528_ramdac_device;
 extern const device_t ics2494an_305_device;
+extern const device_t ati18810_device;
+extern const device_t ati18811_0_device;
+extern const device_t ati18811_1_device;
 extern const device_t ics2595_device;
 extern const device_t icd2061_device;
 extern const device_t ics9161_device;
diff --git a/src/include/86box/vid_svga_render.h b/src/include/86box/vid_svga_render.h
index 1f587d4f5..0c48303c9 100644
--- a/src/include/86box/vid_svga_render.h
+++ b/src/include/86box/vid_svga_render.h
@@ -73,6 +73,15 @@ void svga_render_ABGR8888_highres(svga_t *svga);
 void svga_render_RGBA8888_lowres(svga_t *svga);
 void svga_render_RGBA8888_highres(svga_t *svga);
 
+void ibm8514_render_8bpp(svga_t *svga);
+void ibm8514_render_15bpp(svga_t *svga);
+void ibm8514_render_16bpp(svga_t *svga);
+void ibm8514_render_24bpp(svga_t *svga);
+void ibm8514_render_BGR(svga_t *svga);
+void ibm8514_render_32bpp(svga_t *svga);
+void ibm8514_render_ABGR8888(svga_t *svga);
+void ibm8514_render_RGBA8888(svga_t *svga);
+
 extern void (*svga_render)(svga_t *svga);
 
 #endif /*VID_SVGA_RENDER_H*/
diff --git a/src/include/86box/video.h b/src/include/86box/video.h
index 78395a2f6..92362b8a9 100644
--- a/src/include/86box/video.h
+++ b/src/include/86box/video.h
@@ -302,7 +302,9 @@ extern void ibm8514_device_add(void);
 extern const device_t mach8_isa_device;
 extern const device_t mach32_isa_device;
 extern const device_t mach32_vlb_device;
+extern const device_t mach32_mca_device;
 extern const device_t mach32_pci_device;
+extern const device_t mach32_onboard_pci_device;
 
 /* ATi Mach64 */
 extern const device_t mach64gx_isa_device;
diff --git a/src/machine/m_at_socket4.c b/src/machine/m_at_socket4.c
index b8e080d62..6d8611684 100644
--- a/src/machine/m_at_socket4.c
+++ b/src/machine/m_at_socket4.c
@@ -269,6 +269,7 @@ machine_at_valuepointp60_init(const machine_t *model)
     pci_init(PCI_CONFIG_TYPE_2 | PCI_NO_IRQ_STEERING);
     pci_register_slot(0x00, PCI_CARD_NORTHBRIDGE, 0, 0, 0, 0);
     pci_register_slot(0x01, PCI_CARD_IDE,         0, 0, 0, 0);
+    pci_register_slot(0x03, PCI_CARD_VIDEO,       3, 3, 3, 3);
     pci_register_slot(0x06, PCI_CARD_NORMAL,      3, 2, 1, 4);
     pci_register_slot(0x0E, PCI_CARD_NORMAL,      2, 1, 3, 4);
     pci_register_slot(0x0C, PCI_CARD_NORMAL,      1, 3, 2, 4);
@@ -280,6 +281,9 @@ machine_at_valuepointp60_init(const machine_t *model)
 
     device_add(&i430lx_device);
 
+    if (gfxcard[0] == VID_INTERNAL)
+        device_add(&mach32_onboard_pci_device);
+
     return ret;
 }
 
diff --git a/src/machine/machine_table.c b/src/machine/machine_table.c
index d16713e6c..19a715a0e 100644
--- a/src/machine/machine_table.c
+++ b/src/machine/machine_table.c
@@ -7754,7 +7754,7 @@ const machine_t machines[] = {
             .max_multi = MACHINE_MULTIPLIER_FIXED
         },
         .bus_flags = MACHINE_PS2_PCI,
-        .flags = MACHINE_IDE_DUAL | MACHINE_APM | MACHINE_ACPI,
+        .flags = MACHINE_IDE_DUAL | MACHINE_VIDEO | MACHINE_APM | MACHINE_ACPI,
         .ram = {
             .min = 2048,
             .max = 131072,
@@ -7764,7 +7764,7 @@ const machine_t machines[] = {
         .kbc_device = NULL,
         .kbc_p1 = 0,
         .gpio = 0,
-        .device = NULL,
+        .device = &mach32_onboard_pci_device,
         .fdc_device = NULL,
         .sio_device = NULL,
         .vid_device = NULL,
diff --git a/src/qt/qt_settingsdisplay.cpp b/src/qt/qt_settingsdisplay.cpp
index 5f2f81c58..a22ba29da 100644
--- a/src/qt/qt_settingsdisplay.cpp
+++ b/src/qt/qt_settingsdisplay.cpp
@@ -102,11 +102,17 @@ SettingsDisplay::onCurrentMachineChanged(int machineId)
         ui->comboBoxVideoSecondary->setEnabled(true);
         ui->pushButtonConfigureSecondary->setEnabled(true);
     }
-	if (video_card_get_flags(gfxcard[0]) != VIDEO_FLAG_TYPE_8514)
-		ibm8514_has_vga = 0;
-	if (video_card_get_flags(gfxcard[0]) != VIDEO_FLAG_TYPE_XGA)
-		xga_has_vga = 0;
-
+    if (gfxcard[0] == VID_INTERNAL) {
+        if (video_get_type_monitor(0) != VIDEO_FLAG_TYPE_8514)
+            ibm8514_has_vga = 0;
+        if (video_get_type_monitor(0) != VIDEO_FLAG_TYPE_XGA)
+            xga_has_vga = 0;
+    } else {
+        if (video_card_get_flags(gfxcard[0]) != VIDEO_FLAG_TYPE_8514)
+            ibm8514_has_vga = 0;
+        if (video_card_get_flags(gfxcard[0]) != VIDEO_FLAG_TYPE_XGA)
+            xga_has_vga = 0;
+    }
     ui->comboBoxVideo->setCurrentIndex(selectedRow);
     if (gfxcard[1] == 0)
         ui->pushButtonConfigureSecondary->setEnabled(false);
diff --git a/src/video/CMakeLists.txt b/src/video/CMakeLists.txt
index 3d3d2d466..59205f235 100644
--- a/src/video/CMakeLists.txt
+++ b/src/video/CMakeLists.txt
@@ -18,7 +18,8 @@ add_library(vid OBJECT agpgart.c video.c vid_table.c vid_cga.c vid_cga_comp.c
     vid_incolor.c vid_colorplus.c vid_genius.c vid_pgc.c vid_im1024.c
     vid_sigma.c vid_wy700.c vid_ega.c vid_ega_render.c vid_svga.c vid_8514a.c
     vid_svga_render.c vid_ddc.c vid_vga.c vid_ati_eeprom.c vid_ati18800.c
-    vid_ati28800.c vid_ati_mach8.c vid_ati_mach64.c vid_ati68860_ramdac.c vid_bt48x_ramdac.c
+    vid_ati28800.c vid_ati_mach8.c vid_ati_mach64.c vid_ati68875_ramdac.c
+    vid_ati68860_ramdac.c vid_bt48x_ramdac.c
     vid_av9194.c vid_icd2061.c vid_ics2494.c vid_ics2595.c vid_cl54xx.c
     vid_et3000.c vid_et4000.c vid_sc1148x_ramdac.c vid_sc1502x_ramdac.c
     vid_et4000w32.c vid_stg_ramdac.c vid_ht216.c vid_oak_oti.c vid_paradise.c
diff --git a/src/video/vid_8514a.c b/src/video/vid_8514a.c
index 8e55bcfb5..f46d8a7e7 100644
--- a/src/video/vid_8514a.c
+++ b/src/video/vid_8514a.c
@@ -63,30 +63,16 @@ ibm8514_log(const char *fmt, ...)
 
 #define READ_PIXTRANS_WORD(cx, n)                                                                    \
     if ((cmd <= 1) || (cmd == 5)) {                                                                    \
-        if (dev->local) { \
-            temp = svga->vram[((dev->accel.cy * dev->pitch) + (cx) + (n)) & svga->vram_mask];             \
-            temp |= (svga->vram[((dev->accel.cy * dev->pitch) + (cx) + (n + 1)) & svga->vram_mask] << 8); \
-        } else { \
-            temp = dev->vram[((dev->accel.cy * dev->pitch) + (cx) + (n)) & dev->vram_mask];             \
-            temp |= (dev->vram[((dev->accel.cy * dev->pitch) + (cx) + (n + 1)) & dev->vram_mask] << 8); \
-        } \
+        temp = dev->vram[((dev->accel.cy * dev->pitch) + (cx) + (n)) & dev->vram_mask];             \
+        temp |= (dev->vram[((dev->accel.cy * dev->pitch) + (cx) + (n + 1)) & dev->vram_mask] << 8); \
     } else {                                                                                         \
-        if (dev->local) { \
-            temp = svga->vram[(dev->accel.dest + (cx) + (n)) & svga->vram_mask];                           \
-            temp |= (svga->vram[(dev->accel.dest + (cx) + (n + 1)) & svga->vram_mask] << 8);               \
-        } else { \
-            temp = dev->vram[(dev->accel.dest + (cx) + (n)) & dev->vram_mask];                           \
-            temp |= (dev->vram[(dev->accel.dest + (cx) + (n + 1)) & dev->vram_mask] << 8);               \
-        } \
+        temp = dev->vram[(dev->accel.dest + (cx) + (n)) & dev->vram_mask];                           \
+        temp |= (dev->vram[(dev->accel.dest + (cx) + (n + 1)) & dev->vram_mask] << 8);               \
     }
 
 #define READ(addr, dat) \
-    if (dev->local) { \
-        if ((svga->bpp == 15) || (svga->bpp == 16)) { \
-            dat = vram_w[(addr) & (svga->vram_mask >> 1)]; \
-        } else { \
-            dat = (svga->vram[(addr) & (svga->vram_mask)]); \
-        } \
+    if (dev->bpp) { \
+        dat = vram_w[(addr) & (dev->vram_mask >> 1)]; \
     } else { \
         dat = (dev->vram[(addr) & (dev->vram_mask)]); \
     }
@@ -194,14 +180,9 @@ ibm8514_log(const char *fmt, ...)
     }
 
 #define WRITE(addr, dat)                                         \
-    if (dev->local) { \
-        if ((svga->bpp == 15) || (svga->bpp == 16)) { \
-            vram_w[((addr)) & (svga->vram_mask >> 1)]                    = dat; \
-            svga->changedvram[(((addr)) & (svga->vram_mask >> 1)) >> 11] = changeframecount; \
-        } else { \
-            svga->vram[((addr)) & (svga->vram_mask)]                = dat; \
-            svga->changedvram[(((addr)) & (svga->vram_mask)) >> 12] = changeframecount; \
-        } \
+    if (dev->bpp) { \
+        vram_w[((addr)) & (dev->vram_mask >> 1)]                    = dat; \
+        dev->changedvram[(((addr)) & (dev->vram_mask >> 1)) >> 11] = changeframecount; \
     } else { \
         dev->vram[((addr)) & (dev->vram_mask)]                = dat; \
         dev->changedvram[(((addr)) & (dev->vram_mask)) >> 12] = changeframecount; \
@@ -212,7 +193,7 @@ int ibm8514_has_vga = 0;
 int
 ibm8514_cpu_src(svga_t *svga)
 {
-    const ibm8514_t *dev = &svga->dev8514;
+    ibm8514_t *dev = &svga->dev8514;
 
     if (!(dev->accel.cmd & 0x100))
         return 0;
@@ -226,7 +207,7 @@ ibm8514_cpu_src(svga_t *svga)
 int
 ibm8514_cpu_dest(svga_t *svga)
 {
-    const ibm8514_t *dev = &svga->dev8514;
+    ibm8514_t *dev = &svga->dev8514;
 
     if (!(dev->accel.cmd & 0x100))
         return 0;
@@ -579,8 +560,10 @@ ibm8514_accel_out_fifo(svga_t *svga, uint16_t port, uint32_t val, int len)
                 dev->data_available  = 0;
                 dev->data_available2 = 0;
                 dev->accel.cmd       = val;
-                if (dev->accel.cmd & 0x100)
-                    dev->accel.cmd_back = 0;
+                if (port == 0xdae8) {
+                    if (dev->accel.cmd & 0x100)
+                        dev->accel.cmd_back = 0;
+                }
                 ibm8514_accel_start(-1, 0, -1, 0, svga, len);
             }
             break;
@@ -978,9 +961,9 @@ ibm8514_accel_out(uint16_t port, uint32_t val, svga_t *svga, int len)
             case 0x4ae8:
                 if (!val)
                     break;
-                dev->accel.advfunc_cntl = val & 7;
-                ibm8514_on              = (dev->accel.advfunc_cntl & 1);
-                vga_on                  = !ibm8514_on;
+                dev->accel.advfunc_cntl = val & 0x0f;
+                ibm8514_on = val & 0x01;
+                vga_on = !ibm8514_on;
                 ibm8514_log("IBM 8514/A: VGA ON = %i, val = %02x\n", vga_on, val);
                 svga_recalctimings(svga);
                 break;
@@ -1011,14 +994,14 @@ ibm8514_accel_in(uint16_t port, svga_t *svga, int len)
     ibm8514_t *dev  = &svga->dev8514;
     uint32_t   temp = 0;
     int        cmd;
-    int        vpos      = dev->displine + svga->y_add;
+    int        vpos      = 0;
     int        vblankend = svga->vblankstart + svga->crtc[0x16];
 
     switch (port) {
         case 0x2e8:
-            vpos = dev->displine + svga->y_add;
-            if (vblankend > dev->vtotal) {
-                vblankend -= dev->vtotal;
+            vpos = dev->vc & 0x7ff;
+            if (vblankend > dev->v_total) {
+                vblankend -= dev->v_total;
                 if (vpos >= svga->vblankstart || vpos <= vblankend)
                     temp |= 2;
             } else {
@@ -1051,13 +1034,13 @@ ibm8514_accel_in(uint16_t port, svga_t *svga, int len)
             break;
 
         case 0x42e8:
-            vpos = dev->displine + svga->y_add;
-            if (vblankend > dev->vtotal) {
-                vblankend -= dev->vtotal;
+            vpos = dev->vc & 0x7ff;
+            if (vblankend > dev->v_total) {
+                vblankend -= dev->v_total;
                 if (vpos >= svga->vblankstart || vpos <= vblankend)
                     dev->subsys_stat |= 1;
             } else {
-                if (vpos >= svga->vblankstart && vpos <= vblankend)
+                 if (vpos >= svga->vblankstart && vpos <= vblankend)
                     dev->subsys_stat |= 1;
             }
             if (len != 1) {
@@ -1172,7 +1155,7 @@ void
 ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat, svga_t *svga, UNUSED(int len))
 {
     ibm8514_t *dev     = &svga->dev8514;
-    uint16_t  *vram_w  = (uint16_t *) svga->vram;
+    uint16_t  *vram_w  = (uint16_t *) dev->vram;
     uint16_t   src_dat = 0;
     uint16_t   dest_dat;
     uint16_t   old_dest_dat;
@@ -1181,7 +1164,7 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
     uint16_t   clip_b          = dev->accel.multifunc[3];
     uint16_t   clip_r          = dev->accel.multifunc[4];
     int        pixcntl         = (dev->accel.multifunc[0x0a] >> 6) & 3;
-    uint16_t   mix_mask        = ((svga->bpp == 8) || (svga->bpp == 24)) ? 0x80 : 0x8000;
+    uint16_t   mix_mask        = dev->bpp ? 0x8000 : 0x80;
     uint16_t   compare         = dev->accel.color_cmp;
     int        compare_mode    = dev->accel.multifunc[0x0a] & 0x38;
     int        cmd             = dev->accel.cmd >> 13;
@@ -1194,7 +1177,7 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
     int        and3     = dev->accel.cur_x & 3;
     uint16_t   poly_src = 0;
 
-    if ((svga->bpp == 8) || (svga->bpp == 24)) {
+    if (!dev->bpp) {
         compare &= 0xff;
         frgd_color &= 0xff;
         bkgd_color &= 0xff;
@@ -1225,7 +1208,7 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
             count >>= 3;
         }
 
-        if ((svga->bpp == 15) || (svga->bpp == 16)) {
+        if (dev->bpp) {
             if ((dev->accel.cmd & 0x200) && (count == 2))
                 count >>= 1;
         }
@@ -1312,6 +1295,8 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
     /*Bit 4 of the Command register is the draw yes bit, which enables writing to memory/reading from memory when enabled.
       When this bit is disabled, no writing to memory/reading from memory is allowed. (This bit is almost meaningless on
       the NOP command)*/
+    ibm8514_log("CMD8514: CMD=%d, full=%04x, ssvdraw=%x.\n", cmd, dev->accel.cmd, dev->accel.ssv_draw);
+
     switch (cmd) {
         case 0: /*NOP (Short Stroke Vectors)*/
             if (dev->accel.ssv_state == 0)
@@ -1356,7 +1341,7 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
 
                     mix_dat <<= 1;
                     mix_dat |= 1;
-                    if ((svga->bpp == 15) || (svga->bpp == 16))
+                    if (dev->bpp)
                         cpu_dat >>= 16;
                     else
                         cpu_dat >>= 8;
@@ -1441,7 +1426,7 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
 
                     mix_dat <<= 1;
                     mix_dat |= 1;
-                    if ((svga->bpp == 15) || (svga->bpp == 16))
+                    if (dev->bpp)
                         cpu_dat >>= 16;
                     else
                         cpu_dat >>= 8;
@@ -1693,7 +1678,7 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
 
                     mix_dat <<= 1;
                     mix_dat |= 1;
-                    if ((svga->bpp == 15) || (svga->bpp == 16))
+                    if (dev->bpp)
                         cpu_dat >>= 16;
                     else
                         cpu_dat >>= 8;
@@ -1786,7 +1771,7 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
 
                         dev->accel.temp_cnt--;
                         mix_dat >>= 1;
-                        if ((svga->bpp == 15) || (svga->bpp == 16))
+                        if (dev->bpp)
                             cpu_dat >>= 16;
                         else
                             cpu_dat >>= 8;
@@ -1914,7 +1899,7 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
 
                         mix_dat <<= 1;
                         mix_dat |= 1;
-                        if ((svga->bpp == 15) || (svga->bpp == 16))
+                        if (dev->bpp)
                             cpu_dat >>= 16;
                         else
                             cpu_dat >>= 8;
@@ -2020,7 +2005,7 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
                 if (dev->accel.cur_y >= 0x600)
                     dev->accel.cy |= ~0x5ff;
 
-                if (dev->local && dev->accel.ge_offset && (svga->bpp == 24))
+                if ((dev->local >= 2) && dev->accel.ge_offset && (svga->bpp == 24))
                     dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
                 else
                     dev->accel.dest = dev->accel.cy * dev->pitch;
@@ -2072,7 +2057,7 @@ ibm8514_accel_start(int count, int cpu_input, uint32_t mix_dat, uint32_t cpu_dat
                         if (!(dev->accel.cmd & 0x40) && (frgd_mix == 2) && (bkgd_mix == 2) && (pixcntl == 0) && (cmd == 2)) {
                             if (!(dev->accel.sx & 1)) {
                                 dev->accel.output = 1;
-                                if (dev->local && dev->accel.ge_offset && (svga->bpp == 24))
+                                if ((dev->local >= 2) && dev->accel.ge_offset && (svga->bpp == 24))
                                     dev->accel.newdest_out = (dev->accel.ge_offset << 2) + ((dev->accel.cy + 1) * dev->pitch);
                                 else
                                     dev->accel.newdest_out = (dev->accel.cy + 1) * dev->pitch;
@@ -2195,7 +2180,7 @@ rect_fill_pix:
 
                             mix_dat <<= 1;
                             mix_dat |= 1;
-                            if ((svga->bpp == 15) || (svga->bpp == 16))
+                            if (dev->bpp)
                                 cpu_dat >>= 16;
                             else
                                 cpu_dat >>= 8;
@@ -2320,7 +2305,7 @@ rect_fill_pix:
 
                             mix_dat <<= 1;
                             mix_dat |= 1;
-                            if ((svga->bpp == 15) || (svga->bpp == 16))
+                            if (dev->bpp)
                                 cpu_dat >>= 16;
                             else
                                 cpu_dat >>= 8;
@@ -2348,7 +2333,7 @@ rect_fill_pix:
                                 else
                                     dev->accel.cy--;
 
-                                if (dev->local && dev->accel.ge_offset && (svga->bpp == 24))
+                                if ((dev->local >= 2) && dev->accel.ge_offset && (svga->bpp == 24))
                                     dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
                                 else
                                     dev->accel.dest = dev->accel.cy * dev->pitch;
@@ -2403,7 +2388,7 @@ rect_fill_pix:
                                 }
                             }
                             mix_dat >>= 1;
-                            if ((svga->bpp == 15) || (svga->bpp == 16))
+                            if (dev->bpp)
                                 cpu_dat >>= 16;
                             else
                                 cpu_dat >>= 8;
@@ -2440,7 +2425,7 @@ rect_fill_pix:
                                 else
                                     dev->accel.cy--;
 
-                                if (dev->local && dev->accel.ge_offset && (svga->bpp == 24))
+                                if ((dev->local >= 2) && dev->accel.ge_offset && (svga->bpp == 24))
                                     dev->accel.dest = (dev->accel.ge_offset << 2) + (dev->accel.cy * dev->pitch);
                                 else
                                     dev->accel.dest = dev->accel.cy * dev->pitch;
@@ -2547,7 +2532,7 @@ rect_fill_pix:
                                 }
                                 mix_dat <<= 1;
                                 mix_dat |= 1;
-                                if ((svga->bpp == 15) || (svga->bpp == 16))
+                                if (dev->bpp)
                                     cpu_dat >>= 16;
                                 else
                                     cpu_dat >>= 8;
@@ -2649,7 +2634,7 @@ rect_fill_pix:
 
                                 mix_dat <<= 1;
                                 mix_dat |= 1;
-                                if ((svga->bpp == 15) || (svga->bpp == 16))
+                                if (dev->bpp)
                                     cpu_dat >>= 16;
                                 else
                                     cpu_dat >>= 8;
@@ -2854,7 +2839,7 @@ rect_fill:
                                     }
 
                                     if (poly_src) {
-                                        dev->accel.fill_state = !dev->accel.fill_state;
+                                        dev->accel.fill_state ^= 1;
                                     }
 
                                     if (dev->accel.fill_state) {
@@ -3036,7 +3021,7 @@ rect_fill:
 
                 mix_dat <<= 1;
                 mix_dat |= 1;
-                if ((svga->bpp == 15) || (svga->bpp == 16))
+                if (dev->bpp)
                     cpu_dat >>= 16;
                 else
                     cpu_dat >>= 8;
@@ -3222,7 +3207,7 @@ bitblt_pix:
 
                             mix_dat <<= 1;
                             mix_dat |= 1;
-                            if ((svga->bpp == 15) || (svga->bpp == 16))
+                            if (dev->bpp)
                                 cpu_dat >>= 16;
                             else
                                 cpu_dat >>= 8;
@@ -3312,7 +3297,7 @@ bitblt_pix:
                                 }
                             }
                             mix_dat >>= 1;
-                            if ((svga->bpp == 15) || (svga->bpp == 16))
+                            if (dev->bpp)
                                 cpu_dat >>= 16;
                             else
                                 cpu_dat >>= 8;
@@ -3423,7 +3408,7 @@ bitblt_pix:
                             }
                             mix_dat <<= 1;
                             mix_dat |= 1;
-                            if ((svga->bpp == 15) || (svga->bpp == 16))
+                            if (dev->bpp)
                                 cpu_dat >>= 16;
                             else
                                 cpu_dat >>= 8;
@@ -3622,7 +3607,7 @@ bitblt:
                             }
                         }
                     } else {
-                        if ((svga->bpp == 24) && dev->local && (dev->accel.cmd == 0xc2b5)) {
+                        if ((svga->bpp == 24) && (dev->local >= 2) && (dev->accel.cmd == 0xc2b5)) {
                             int64_t cx;
                             int64_t dx;
 
@@ -3753,10 +3738,11 @@ bitblt:
     }
 }
 
-static void
+void
 ibm8514_render_8bpp(svga_t *svga)
 {
     ibm8514_t *dev = &svga->dev8514;
+    int       x;
     uint32_t  *p;
     uint32_t   dat;
 
@@ -3771,18 +3757,18 @@ ibm8514_render_8bpp(svga_t *svga)
             dev->firstline_draw = dev->displine;
         dev->lastline_draw = dev->displine;
 
-        for (int x = 0; x <= dev->h_disp; x += 8) {
+        for (x = 0; x <= dev->h_disp; x += 8) {
             dat  = *(uint32_t *) (&dev->vram[dev->ma & dev->vram_mask]);
-            p[0] = dev->map8[dat & 0xff];
-            p[1] = dev->map8[(dat >> 8) & 0xff];
-            p[2] = dev->map8[(dat >> 16) & 0xff];
-            p[3] = dev->map8[(dat >> 24) & 0xff];
+            p[0] = dev->pallook[dat & 0xff];
+            p[1] = dev->pallook[(dat >> 8) & 0xff];
+            p[2] = dev->pallook[(dat >> 16) & 0xff];
+            p[3] = dev->pallook[(dat >> 24) & 0xff];
 
             dat  = *(uint32_t *) (&dev->vram[(dev->ma + 4) & dev->vram_mask]);
-            p[4] = dev->map8[dat & 0xff];
-            p[5] = dev->map8[(dat >> 8) & 0xff];
-            p[6] = dev->map8[(dat >> 16) & 0xff];
-            p[7] = dev->map8[(dat >> 24) & 0xff];
+            p[4] = dev->pallook[dat & 0xff];
+            p[5] = dev->pallook[(dat >> 8) & 0xff];
+            p[6] = dev->pallook[(dat >> 16) & 0xff];
+            p[7] = dev->pallook[(dat >> 24) & 0xff];
 
             dev->ma += 8;
             p += 8;
@@ -3791,6 +3777,216 @@ ibm8514_render_8bpp(svga_t *svga)
     }
 }
 
+void
+ibm8514_render_15bpp(svga_t *svga)
+{
+    ibm8514_t *dev = &svga->dev8514;
+    int x;
+    uint32_t  *p;
+    uint32_t   dat;
+
+    if ((dev->displine + svga->y_add) < 0) {
+        return;
+    }
+
+    if (dev->changedvram[dev->ma >> 12] || dev->changedvram[(dev->ma >> 12) + 1] || svga->fullchange) {
+        p = &buffer32->line[dev->displine + svga->y_add][svga->x_add];
+
+        if (dev->firstline_draw == 2000)
+            dev->firstline_draw = dev->displine;
+        dev->lastline_draw = dev->displine;
+
+        for (x = 0; x <= dev->h_disp; x += 8) {
+            dat      = *(uint32_t *) (&dev->vram[(dev->ma + (x << 1)) & dev->vram_mask]);
+            p[x]     = video_15to32[dat & 0xffff];
+            p[x + 1] = video_15to32[dat >> 16];
+
+            dat      = *(uint32_t *) (&dev->vram[(dev->ma + (x << 1) + 4) & dev->vram_mask]);
+            p[x + 2] = video_15to32[dat & 0xffff];
+            p[x + 3] = video_15to32[dat >> 16];
+
+            dat      = *(uint32_t *) (&dev->vram[(dev->ma + (x << 1) + 8) & dev->vram_mask]);
+            p[x + 4] = video_15to32[dat & 0xffff];
+            p[x + 5] = video_15to32[dat >> 16];
+
+            dat      = *(uint32_t *) (&dev->vram[(dev->ma + (x << 1) + 12) & dev->vram_mask]);
+            p[x + 6] = video_15to32[dat & 0xffff];
+            p[x + 7] = video_15to32[dat >> 16];
+        }
+        dev->ma += (x << 1);
+        dev->ma &= dev->vram_mask;
+    }
+}
+
+void
+ibm8514_render_16bpp(svga_t *svga)
+{
+    ibm8514_t *dev = &svga->dev8514;
+    int x;
+    uint32_t  *p;
+    uint32_t   dat;
+
+    if ((dev->displine + svga->y_add) < 0) {
+        return;
+    }
+
+    if (dev->changedvram[dev->ma >> 12] || dev->changedvram[(dev->ma >> 12) + 1] || svga->fullchange) {
+        p = &buffer32->line[dev->displine + svga->y_add][svga->x_add];
+
+        if (dev->firstline_draw == 2000)
+            dev->firstline_draw = dev->displine;
+        dev->lastline_draw = dev->displine;
+
+        for (x = 0; x <= dev->h_disp; x += 8) {
+            dat      = *(uint32_t *) (&dev->vram[(dev->ma + (x << 1)) & dev->vram_mask]);
+            p[x]     = video_16to32[dat & 0xffff];
+            p[x + 1] = video_16to32[dat >> 16];
+
+            dat      = *(uint32_t *) (&dev->vram[(dev->ma + (x << 1) + 4) & dev->vram_mask]);
+            p[x + 2] = video_16to32[dat & 0xffff];
+            p[x + 3] = video_16to32[dat >> 16];
+
+            dat      = *(uint32_t *) (&dev->vram[(dev->ma + (x << 1) + 8) & dev->vram_mask]);
+            p[x + 4] = video_16to32[dat & 0xffff];
+            p[x + 5] = video_16to32[dat >> 16];
+
+            dat      = *(uint32_t *) (&dev->vram[(dev->ma + (x << 1) + 12) & dev->vram_mask]);
+            p[x + 6] = video_16to32[dat & 0xffff];
+            p[x + 7] = video_16to32[dat >> 16];
+        }
+        dev->ma += (x << 1);
+        dev->ma &= dev->vram_mask;
+    }
+}
+
+void
+ibm8514_render_24bpp(svga_t *svga)
+{
+    ibm8514_t *dev = &svga->dev8514;
+    int       x;
+    uint32_t *p;
+    uint32_t  dat;
+
+    if ((dev->displine + svga->y_add) < 0)
+        return;
+
+    if (dev->changedvram[dev->ma >> 12] || dev->changedvram[(dev->ma >> 12) + 1] || svga->fullchange) {
+        p = &buffer32->line[dev->displine + svga->y_add][svga->x_add];
+
+        if (dev->firstline_draw == 2000)
+            dev->firstline_draw = dev->displine;
+        dev->lastline_draw = dev->displine;
+
+        for (x = 0; x <= dev->h_disp; x += 4) {
+            dat  = *(uint32_t *) (&dev->vram[dev->ma & dev->vram_mask]);
+            p[x] = dat & 0xffffff;
+
+            dat      = *(uint32_t *) (&dev->vram[(dev->ma + 3) & dev->vram_mask]);
+            p[x + 1] = dat & 0xffffff;
+
+            dat      = *(uint32_t *) (&dev->vram[(dev->ma + 6) & dev->vram_mask]);
+            p[x + 2] = dat & 0xffffff;
+
+            dat      = *(uint32_t *) (&dev->vram[(dev->ma + 9) & dev->vram_mask]);
+            p[x + 3] = dat & 0xffffff;
+
+            dev->ma += 12;
+        }
+        dev->ma &= dev->vram_mask;
+    }
+}
+
+void
+ibm8514_render_BGR(svga_t *svga)
+{
+    ibm8514_t *dev = &svga->dev8514;
+    int       x;
+    uint32_t *p;
+    uint32_t  dat;
+
+    if ((dev->displine + svga->y_add) < 0)
+        return;
+
+    if (dev->changedvram[dev->ma >> 12] || dev->changedvram[(dev->ma >> 12) + 1] || svga->fullchange) {
+        p = &buffer32->line[dev->displine + svga->y_add][svga->x_add];
+
+        if (dev->firstline_draw == 2000)
+            dev->firstline_draw = dev->displine;
+        dev->lastline_draw = dev->displine;
+
+        for (x = 0; x <= dev->h_disp; x += 4) {
+            dat  = *(uint32_t *) (&dev->vram[dev->ma & dev->vram_mask]);
+            p[x] = ((dat & 0xff0000) >> 16) | (dat & 0x00ff00) | ((dat & 0x0000ff) << 16);
+
+            dat      = *(uint32_t *) (&dev->vram[(dev->ma + 3) & dev->vram_mask]);
+            p[x + 1] = ((dat & 0xff0000) >> 16) | (dat & 0x00ff00) | ((dat & 0x0000ff) << 16);
+
+            dat      = *(uint32_t *) (&dev->vram[(dev->ma + 6) & dev->vram_mask]);
+            p[x + 2] = ((dat & 0xff0000) >> 16) | (dat & 0x00ff00) | ((dat & 0x0000ff) << 16);
+
+            dat      = *(uint32_t *) (&dev->vram[(dev->ma + 9) & dev->vram_mask]);
+            p[x + 3] = ((dat & 0xff0000) >> 16) | (dat & 0x00ff00) | ((dat & 0x0000ff) << 16);
+
+            dev->ma += 12;
+        }
+        dev->ma &= dev->vram_mask;
+    }
+}
+
+void
+ibm8514_render_ABGR8888(svga_t *svga)
+{
+    ibm8514_t *dev = &svga->dev8514;
+    int       x;
+    uint32_t *p;
+    uint32_t  dat;
+
+    if ((dev->displine + svga->y_add) < 0)
+        return;
+
+    if (dev->changedvram[dev->ma >> 12] || dev->changedvram[(dev->ma >> 12) + 1] || svga->fullchange) {
+        p = &buffer32->line[dev->displine + svga->y_add][svga->x_add];
+
+        if (dev->firstline_draw == 2000)
+            dev->firstline_draw = dev->displine;
+        dev->lastline_draw = dev->displine;
+
+        for (x = 0; x <= dev->h_disp; x++) {
+            dat  = *(uint32_t *) (&dev->vram[(dev->ma + (x << 2)) & dev->vram_mask]);
+            *p++ = ((dat & 0xff0000) >> 16) | (dat & 0x00ff00) | ((dat & 0x0000ff) << 16);
+        }
+        dev->ma += (x * 4);
+        dev->ma &= dev->vram_mask;
+    }
+}
+
+void
+ibm8514_render_RGBA8888(svga_t *svga)
+{
+    ibm8514_t *dev = &svga->dev8514;
+    int       x;
+    uint32_t *p;
+    uint32_t  dat;
+
+    if ((dev->displine + svga->y_add) < 0)
+        return;
+
+    if (dev->changedvram[dev->ma >> 12] || dev->changedvram[(dev->ma >> 12) + 1] || svga->fullchange) {
+        p = &buffer32->line[dev->displine + svga->y_add][svga->x_add];
+
+        if (dev->firstline_draw == 2000)
+            dev->firstline_draw = dev->displine;
+        dev->lastline_draw = dev->displine;
+
+        for (x = 0; x <= dev->h_disp; x++) {
+            dat  = *(uint32_t *) (&dev->vram[(dev->ma + (x << 2)) & dev->vram_mask]);
+            *p++ = dat >> 8;
+        }
+        dev->ma += (x * 4);
+        dev->ma &= dev->vram_mask;
+    }
+}
+
 static void
 ibm8514_render_overscan_left(ibm8514_t *dev, svga_t *svga)
 {
@@ -3828,7 +4024,18 @@ ibm8514_poll(ibm8514_t *dev, svga_t *svga)
     int      wy;
 
     if (!dev->linepos) {
+        if ((dev->displine == dev->hwcursor_latch.y) && dev->hwcursor_latch.ena) {
+            dev->hwcursor_on      = dev->hwcursor_latch.cur_ysize - dev->hwcursor_latch.yoff;
+            dev->hwcursor_oddeven = 0;
+        }
+
+        if ((dev->displine == (svga->hwcursor_latch.y + 1)) && dev->hwcursor_latch.ena && dev->interlace) {
+            dev->hwcursor_on      = dev->hwcursor_latch.cur_ysize - (dev->hwcursor_latch.yoff + 1);
+            dev->hwcursor_oddeven = 1;
+        }
+
         timer_advance_u64(&svga->timer, svga->dispofftime);
+        svga->cgastat |= 1;
         dev->linepos = 1;
 
         if (dev->dispon) {
@@ -3841,6 +4048,9 @@ ibm8514_poll(ibm8514_t *dev, svga_t *svga)
                 video_wait_for_buffer();
             }
 
+            if (dev->hwcursor_on)
+                dev->changedvram[dev->ma >> 12] = dev->changedvram[(dev->ma >> 12) + 1] = dev->interlace ? 3 : 2;
+
             svga->render(svga);
 
             svga->x_add = (overscan_x >> 1);
@@ -3848,6 +4058,14 @@ ibm8514_poll(ibm8514_t *dev, svga_t *svga)
             ibm8514_render_overscan_right(dev, svga);
             svga->x_add = (overscan_x >> 1);
 
+            if (dev->hwcursor_on) {
+                if (svga->hwcursor_draw)
+                    svga->hwcursor_draw(svga, dev->displine + svga->y_add);
+                dev->hwcursor_on--;
+                if (dev->hwcursor_on && dev->interlace)
+                    dev->hwcursor_on--;
+            }
+
             if (dev->lastline < dev->displine)
                 dev->lastline = dev->displine;
         }
@@ -3855,10 +4073,15 @@ ibm8514_poll(ibm8514_t *dev, svga_t *svga)
         dev->displine++;
         if (dev->interlace)
             dev->displine++;
+        if ((svga->cgastat & 8) && ((dev->displine & 0x0f) == (svga->crtc[0x11] & 0x0f)) && svga->vslines)
+            svga->cgastat &= ~8;
+        svga->vslines++;
         if (dev->displine > 1500)
             dev->displine = 0;
     } else {
         timer_advance_u64(&svga->timer, svga->dispontime);
+        if (dev->dispon)
+            svga->cgastat &= ~1;
         dev->hdisp_on = 0;
 
         dev->linepos = 0;
@@ -3872,13 +4095,13 @@ ibm8514_poll(ibm8514_t *dev, svga_t *svga)
                 dev->ma = dev->maback;
             } else {
                 dev->sc++;
-                dev->sc &= 31;
+                dev->sc &= 0x1f;
                 dev->ma = dev->maback;
             }
         }
 
         dev->vc++;
-        dev->vc &= 2047;
+        dev->vc &= 0x7ff;
 
         if (dev->vc == dev->dispend) {
             dev->dispon = 0;
@@ -3893,6 +4116,7 @@ ibm8514_poll(ibm8514_t *dev, svga_t *svga)
         }
         if (dev->vc == dev->v_syncstart) {
             dev->dispon = 0;
+            svga->cgastat |= 8;
             x           = dev->h_disp;
 
             if (dev->interlace && !dev->oddeven)
@@ -3914,6 +4138,7 @@ ibm8514_poll(ibm8514_t *dev, svga_t *svga)
             dev->oddeven ^= 1;
 
             changeframecount = dev->interlace ? 3 : 2;
+            svga->vslines    = 0;
 
             if (dev->interlace && dev->oddeven)
                 dev->ma = dev->maback = (dev->rowoffset << 1);
@@ -3930,6 +4155,9 @@ ibm8514_poll(ibm8514_t *dev, svga_t *svga)
             dev->displine = (dev->interlace && dev->oddeven) ? 1 : 0;
 
             svga->x_add = (overscan_x >> 1);
+
+            dev->hwcursor_on    = 0;
+            dev->hwcursor_latch = dev->hwcursor;
         }
     }
 }
@@ -3948,19 +4176,17 @@ ibm8514_recalctimings(svga_t *svga)
         dev->rowcount    = !!(dev->disp_cntl & 0x08);
         dev->dispend     = ((dev->vdisp >> 1) + 1);
         if (dev->dispend == 766)
-            dev->dispend = 768;
+            dev->dispend += 2;
 
         if (dev->dispend == 598)
-            dev->dispend = 600;
+            dev->dispend += 2;
 
         if (dev->accel.advfunc_cntl & 4) {
-            if (!vga_on && dev->ibm_mode) {
-                if (dev->h_disp == 8) {
-                    dev->h_disp      = 1024;
-                    dev->dispend     = 768;
-                    dev->v_total     = 1536;
-                    dev->v_syncstart = 1536;
-                }
+            if (dev->h_disp == 8) {
+                dev->h_disp = 1024;
+                dev->dispend = 768;
+                dev->v_total = 1536;
+                dev->v_syncstart = 1536;
             }
 
             if (dev->dispend == 598)
@@ -3975,20 +4201,14 @@ ibm8514_recalctimings(svga_t *svga)
                 dev->v_total >>= 1;
             }
 
-            if (ibm8514_has_vga) {
-                dev->pitch     = dev->ext_pitch;
-                dev->rowoffset = dev->ext_crt_pitch;
-            } else
-                dev->rowoffset = 128;
+            dev->rowoffset = 0x80;
 
             ibm8514_log("1024x768 clock mode, hdisp = %d, htotal = %d, vtotal = %d, vsyncstart = %d, interlace = %02x\n", dev->h_disp, dev->h_total, dev->v_total, dev->v_syncstart, dev->interlace);
             svga->clock = (cpuclock * (double) (1ULL << 32)) / 44900000.0;
         } else {
-            if (!vga_on && dev->ibm_mode) {
-                if (dev->h_disp == 1024) {
-                    dev->h_disp  = 640;
-                    dev->dispend = 480;
-                }
+            if (dev->h_disp == 1024) {
+                dev->h_disp = 640;
+                dev->dispend = 480;
             }
 
             if (dev->interlace) {
@@ -4000,11 +4220,7 @@ ibm8514_recalctimings(svga_t *svga)
                 dev->v_total >>= 1;
             }
 
-            if (ibm8514_has_vga) {
-                dev->pitch     = dev->ext_pitch;
-                dev->rowoffset = dev->ext_crt_pitch;
-            } else
-                dev->rowoffset = 128;
+            dev->rowoffset = 0x80;
 
             svga->clock = (cpuclock * (double) (1ULL << 32)) / 25175000.0;
         }
@@ -4059,15 +4275,14 @@ ibm8514_init(const device_t *info)
     dev->vram        = calloc(dev->vram_size, 1);
     dev->changedvram = calloc(dev->vram_size >> 12, 1);
     dev->vram_mask   = dev->vram_size - 1;
-    dev->map8        = svga->pallook;
+    dev->map8        = dev->pallook;
 
     dev->type     = info->flags;
-    dev->ibm_mode = 1;
-    dev->bpp      = 8;
+    dev->bpp      = 0;
 
     ibm8514_io_set(svga);
 
-    if (info->flags & DEVICE_MCA) {
+    if (dev->type & DEVICE_MCA) {
         dev->pos_regs[0] = 0x7f;
         dev->pos_regs[1] = 0xef;
         mca_add(ibm8514_mca_read, ibm8514_mca_write, ibm8514_mca_feedb, NULL, svga);
diff --git a/src/video/vid_ati68860_ramdac.c b/src/video/vid_ati68860_ramdac.c
index f28bd0dcf..ffc96fb01 100644
--- a/src/video/vid_ati68860_ramdac.c
+++ b/src/video/vid_ati68860_ramdac.c
@@ -65,22 +65,22 @@ typedef struct ati68860_ramdac_t {
 } ati68860_ramdac_t;
 
 void
-ati68860_ramdac_out(uint16_t addr, uint8_t val, void *priv, svga_t *svga)
+ati68860_ramdac_out(uint16_t addr, uint8_t val, void *p, svga_t *svga)
 {
-    ati68860_ramdac_t *ramdac = (ati68860_ramdac_t *) priv;
+    ati68860_ramdac_t *ramdac = (ati68860_ramdac_t *) p;
 
     switch (addr) {
         case 0:
-            svga_out(0x3c8, val, svga);
+            svga_out(ibm8514_on ? 0x2ec : 0x3c8, val, svga);
             break;
         case 1:
-            svga_out(0x3c9, val, svga);
+            svga_out(ibm8514_on ? 0x2ed : 0x3c9, val, svga);
             break;
         case 2:
-            svga_out(0x3c6, val, svga);
+            svga_out(ibm8514_on ? 0x2ea : 0x3c6, val, svga);
             break;
         case 3:
-            svga_out(0x3c7, val, svga);
+            svga_out(ibm8514_on ? 0x2eb : 0x3c7, val, svga);
             break;
         default:
             ramdac->regs[addr & 0xf] = val;
@@ -168,23 +168,23 @@ ati68860_ramdac_out(uint16_t addr, uint8_t val, void *priv, svga_t *svga)
 }
 
 uint8_t
-ati68860_ramdac_in(uint16_t addr, void *priv, svga_t *svga)
+ati68860_ramdac_in(uint16_t addr, void *p, svga_t *svga)
 {
-    const ati68860_ramdac_t *ramdac = (ati68860_ramdac_t *) priv;
-    uint8_t                  temp   = 0;
+    ati68860_ramdac_t *ramdac = (ati68860_ramdac_t *) p;
+    uint8_t            temp   = 0;
 
     switch (addr) {
         case 0:
-            temp = svga_in(0x3c8, svga);
+            temp = svga_in(ibm8514_on ? 0x2ec : 0x3c8, svga);
             break;
         case 1:
-            temp = svga_in(0x3c9, svga);
+            temp = svga_in(ibm8514_on ? 0x2ed : 0x3c9, svga);
             break;
         case 2:
-            temp = svga_in(0x3c6, svga);
+            temp = svga_in(ibm8514_on ? 0x2ea : 0x3c6, svga);
             break;
         case 3:
-            temp = svga_in(0x3c7, svga);
+            temp = svga_in(ibm8514_on ? 0x2eb : 0x3c7, svga);
             break;
         case 4:
         case 8:
@@ -207,9 +207,9 @@ ati68860_ramdac_in(uint16_t addr, void *priv, svga_t *svga)
 }
 
 void
-ati68860_set_ramdac_type(void *priv, int type)
+ati68860_set_ramdac_type(void *p, int type)
 {
-    ati68860_ramdac_t *ramdac = (ati68860_ramdac_t *) priv;
+    ati68860_ramdac_t *ramdac = (ati68860_ramdac_t *) p;
 
     if (ramdac->ramdac_type != type) {
         ramdac->ramdac_type = type;
@@ -237,17 +237,17 @@ ati68860_ramdac_init(UNUSED(const device_t *info))
 }
 
 void
-ati68860_ramdac_set_render(void *priv, svga_t *svga)
+ati68860_ramdac_set_render(void *p, svga_t *svga)
 {
-    ati68860_ramdac_t *ramdac = (ati68860_ramdac_t *) priv;
+    ati68860_ramdac_t *ramdac = (ati68860_ramdac_t *) p;
 
     svga->render = ramdac->render;
 }
 
 void
-ati68860_ramdac_set_pallook(void *priv, int i, uint32_t col)
+ati68860_ramdac_set_pallook(void *p, int i, uint32_t col)
 {
-    ati68860_ramdac_t *ramdac = (ati68860_ramdac_t *) priv;
+    ati68860_ramdac_t *ramdac = (ati68860_ramdac_t *) p;
 
     ramdac->pallook[i] = col;
 }
@@ -255,11 +255,11 @@ ati68860_ramdac_set_pallook(void *priv, int i, uint32_t col)
 void
 ati68860_hwcursor_draw(svga_t *svga, int displine)
 {
-    const ati68860_ramdac_t *ramdac = (ati68860_ramdac_t *) svga->ramdac;
-    int                      offset;
-    uint8_t                  dat;
-    uint32_t                 col0 = ramdac->pallook[0];
-    uint32_t                 col1 = ramdac->pallook[1];
+    ati68860_ramdac_t *ramdac = (ati68860_ramdac_t *) svga->ramdac;
+    int                offset;
+    uint8_t            dat;
+    uint32_t           col0 = ramdac->pallook[0];
+    uint32_t           col1 = ramdac->pallook[1];
 
     offset = svga->dac_hwcursor_latch.xoff;
     for (uint32_t x = 0; x < 64 - svga->dac_hwcursor_latch.xoff; x += 4) {
diff --git a/src/video/vid_ati68875_ramdac.c b/src/video/vid_ati68875_ramdac.c
new file mode 100644
index 000000000..981b0c40b
--- /dev/null
+++ b/src/video/vid_ati68875_ramdac.c
@@ -0,0 +1,219 @@
+/*
+ * 86Box    A hypervisor and IBM PC system emulator that specializes in
+ *          running old operating systems and software designed for IBM
+ *          PC systems and compatibles from 1981 through fairly recent
+ *          system designs based on the PCI bus.
+ *
+ *          This file is part of the 86Box distribution.
+ *
+ *          Emulation of the Mach32-compatible ATI 68875 RAMDAC and clones.
+ *
+ *
+ *
+ * Authors: TheCollector1995.
+ *
+ *          Copyright 2022-2023 TheCollector1995.
+ */
+#include <stdio.h>
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+#include <wchar.h>
+#include <86box/86box.h>
+#include <86box/device.h>
+#include <86box/mem.h>
+#include <86box/timer.h>
+#include <86box/video.h>
+#include <86box/vid_svga.h>
+#include <86box/vid_svga_render.h>
+
+typedef struct ati68875_ramdac_t {
+    uint8_t gen_cntl;
+    uint8_t in_clk_sel;
+    uint8_t out_clk_sel;
+    uint8_t mux_cntl;
+    uint8_t palette_page_sel;
+    uint8_t test_reg;
+} ati68875_ramdac_t;
+
+static void
+ati68875_set_bpp(ati68875_ramdac_t *ramdac, svga_t *svga)
+{
+    if (ramdac->mux_cntl == 0xff)
+        return;
+
+    if (ramdac->mux_cntl & 0x20)
+        svga->bpp = 8;
+    else {
+        svga->bpp = 24;
+    }
+    svga_recalctimings(svga);
+}
+
+void
+ati68875_ramdac_out(uint16_t addr, int rs2, int rs3, uint8_t val, void *p, svga_t *svga)
+{
+    ati68875_ramdac_t *ramdac = (ati68875_ramdac_t *) p;
+    uint8_t rs = (addr & 0x03);
+
+    rs |= (!!rs2 << 2);
+    rs |= (!!rs3 << 3);
+
+    switch (rs) {
+        case 0x00: /* Palette Write Index Register (RS value = 0000) */
+        case 0x01: /* Palette Data Register (RS value = 0001) */
+        case 0x02: /* Pixel Read Mask Register (RS value = 0010) */
+        case 0x03:
+            svga_out(addr, val, svga);
+            break;
+        case 0x08: /* General Control Register (RS value = 1000) */
+            ramdac->gen_cntl = val;
+            break;
+        case 0x09: /* Input Clock Selection Register (RS value = 1001) */
+            ramdac->in_clk_sel = val;
+            break;
+        case 0x0a: /* Output Clock Selection Register (RS value = 1010) */
+            ramdac->out_clk_sel = val;
+            break;
+        case 0x0b: /* MUX Control Register (RS value = 1011) */
+            ramdac->mux_cntl = val;
+            ati68875_set_bpp(ramdac, svga);
+            break;
+        case 0x0c: /* Palette Page Register (RS value = 1100) */
+            ramdac->palette_page_sel = val;
+            break;
+        case 0x0e: /* Test Register (RS value = 1110) */
+            ramdac->test_reg = val;
+            break;
+        case 0x0f: /* Reset State (RS value = 1111) */
+            ramdac->mux_cntl = 0x2d;
+            svga->bpp = 8;
+            svga_recalctimings(svga);
+            break;
+    }
+
+    return;
+}
+
+uint8_t
+ati68875_ramdac_in(uint16_t addr, int rs2, int rs3, void *p, svga_t *svga)
+{
+    ati68875_ramdac_t *ramdac = (ati68875_ramdac_t *) p;
+    ibm8514_t *dev = &svga->dev8514;
+    uint8_t rs = (addr & 0x03);
+    uint8_t temp = 0;
+
+    rs |= (!!rs2 << 2);
+    rs |= (!!rs3 << 3);
+
+    switch (rs) {
+        case 0x00: /* Palette Write Index Register (RS value = 0000) */
+        case 0x01: /* Palette Data Register (RS value = 0001) */
+        case 0x02: /* Pixel Read Mask Register (RS value = 0010) */
+        case 0x03:
+            temp = svga_in(addr, svga);
+            break;
+        case 0x08: /* General Control Register (RS value = 1000) */
+            temp = ramdac->gen_cntl;
+            break;
+        case 0x09: /* Input Clock Selection Register (RS value = 1001) */
+            temp = ramdac->in_clk_sel;
+            break;
+        case 0x0a: /* Output Clock Selection Register (RS value = 1010) */
+            temp = ramdac->out_clk_sel;
+            break;
+        case 0x0b: /* MUX Control Register (RS value = 1011) */
+            temp = ramdac->mux_cntl;
+            break;
+        case 0x0c: /* Palette Page Register (RS value = 1100) */
+            temp = ramdac->palette_page_sel;
+            break;
+        case 0x0e: /* Test Register (RS value = 1110) */
+            switch (ramdac->test_reg & 0x07) {
+                case 0x00:
+                    temp = ibm8514_on ? dev->dac_r : svga->dac_r;
+                    break;
+                case 0x01:
+                    temp = ibm8514_on ? dev->dac_g : svga->dac_g;
+                    break;
+                case 0x02:
+                    temp = ibm8514_on ? dev->dac_b : svga->dac_b;
+                    break;
+                case 0x03:
+                    temp = 0x75;
+                    break;
+                case 0x04:
+                    if (ibm8514_on) {
+                        dev->dac_r++;
+                        temp = dev->dac_r;
+                    } else {
+                        svga->dac_r++;
+                        temp = svga->dac_r;
+                    }
+                    break;
+                case 0x05:
+                    if (ibm8514_on) {
+                        dev->dac_g++;
+                        temp = dev->dac_g;
+                    } else {
+                        svga->dac_g++;
+                        temp = svga->dac_g;
+                    }
+                    break;
+                case 0x06:
+                    if (ibm8514_on) {
+                        dev->dac_b++;
+                        temp = dev->dac_b;
+                    } else {
+                        svga->dac_b++;
+                        temp = svga->dac_b;
+                    }
+                    break;
+                case 0x07:
+                    if (ramdac->test_reg & 0x80)
+                        temp = ibm8514_on ? dev->dac_r : svga->dac_r;
+                    else if (ramdac->test_reg & 0x40)
+                        temp = ibm8514_on ? dev->dac_g : svga->dac_g;
+                    else if (ramdac->test_reg & 0x20)
+                        temp = ibm8514_on ? dev->dac_b : svga->dac_b;
+                    break;
+            }
+            break;
+    }
+
+    return temp;
+}
+
+static void *
+ati68875_ramdac_init(const device_t *info)
+{
+    ati68875_ramdac_t *ramdac = (ati68875_ramdac_t *) malloc(sizeof(ati68875_ramdac_t));
+    memset(ramdac, 0, sizeof(ati68875_ramdac_t));
+
+    ramdac->mux_cntl = 0x2d;
+
+    return ramdac;
+}
+
+static void
+ati68875_ramdac_close(void *priv)
+{
+    ati68875_ramdac_t *ramdac = (ati68875_ramdac_t *) priv;
+
+    if (ramdac)
+        free(ramdac);
+}
+
+const device_t ati68875_ramdac_device = {
+    .name          = "ATI 68875 RAMDAC",
+    .internal_name = "ati68875_ramdac",
+    .flags         = 0,
+    .local         = 0,
+    .init          = ati68875_ramdac_init,
+    .close         = ati68875_ramdac_close,
+    .reset         = NULL,
+    { .available = NULL },
+    .speed_changed = NULL,
+    .force_redraw  = NULL,
+    .config        = NULL
+};
diff --git a/src/video/vid_ati_mach64.c b/src/video/vid_ati_mach64.c
index 98dcfd14d..2e05a6e03 100644
--- a/src/video/vid_ati_mach64.c
+++ b/src/video/vid_ati_mach64.c
@@ -439,12 +439,12 @@ mach64_out(uint16_t addr, uint8_t val, void *priv)
             svga->crtcreg = val & 0x3f;
             return;
         case 0x3D5:
+            if (svga->crtcreg > 0x20)
+                return;
             if ((svga->crtcreg < 7) && (svga->crtc[0x11] & 0x80))
                 return;
             if ((svga->crtcreg == 7) && (svga->crtc[0x11] & 0x80))
                 val = (svga->crtc[7] & ~0x10) | (val & 0x10);
-            if (svga->crtcreg > 0x18)
-                return;
             old                       = svga->crtc[svga->crtcreg];
             svga->crtc[svga->crtcreg] = val;
 
@@ -454,7 +454,7 @@ mach64_out(uint16_t addr, uint8_t val, void *priv)
                         svga->fullchange = 3;
                         svga->ma_latch   = ((svga->crtc[0xc] << 8) | svga->crtc[0xd]) + ((svga->crtc[8] & 0x60) >> 5);
                     } else {
-                        svga->fullchange = changeframecount;
+                        svga->fullchange = svga->monitor->mon_changeframecount;
                         svga_recalctimings(svga);
                     }
                 }
@@ -493,7 +493,7 @@ mach64_in(uint16_t addr, void *priv)
         case 0x3D4:
             return svga->crtcreg;
         case 0x3D5:
-            if (svga->crtcreg > 0x18)
+            if (svga->crtcreg > 0x20)
                 return 0xff;
             return svga->crtc[svga->crtcreg];
 
@@ -1521,13 +1521,13 @@ mach64_start_line(mach64_t *mach64)
 #define WRITE(addr, width)                                                                  \
     if (width == 0) {                                                                       \
         svga->vram[(addr) &mach64->vram_mask]                = dest_dat;                    \
-        svga->changedvram[((addr) &mach64->vram_mask) >> 12] = changeframecount;            \
+        svga->changedvram[((addr) &mach64->vram_mask) >> 12] = svga->monitor->mon_changeframecount;            \
     } else if (width == 1) {                                                                \
         *(uint16_t *) &svga->vram[((addr) << 1) & mach64->vram_mask] = dest_dat;            \
-        svga->changedvram[(((addr) << 1) & mach64->vram_mask) >> 12] = changeframecount;    \
+        svga->changedvram[(((addr) << 1) & mach64->vram_mask) >> 12] = svga->monitor->mon_changeframecount;    \
     } else if (width == 2) {                                                                \
         *(uint32_t *) &svga->vram[((addr) << 2) & mach64->vram_mask] = dest_dat;            \
-        svga->changedvram[(((addr) << 2) & mach64->vram_mask) >> 12] = changeframecount;    \
+        svga->changedvram[(((addr) << 2) & mach64->vram_mask) >> 12] = svga->monitor->mon_changeframecount;    \
     } else {                                                                                \
         if (dest_dat & 1) {                                                                 \
             if (mach64->dp_pix_width & DP_BYTE_PIX_ORDER)                                   \
@@ -1540,7 +1540,7 @@ mach64_start_line(mach64_t *mach64)
             else                                                                            \
                 svga->vram[((addr) >> 3) & mach64->vram_mask] &= ~(1 << (7 - ((addr) &7))); \
         }                                                                                   \
-        svga->changedvram[(((addr) >> 3) & mach64->vram_mask) >> 12] = changeframecount;    \
+        svga->changedvram[(((addr) >> 3) & mach64->vram_mask) >> 12] = svga->monitor->mon_changeframecount;    \
     }
 
 void
@@ -3069,7 +3069,7 @@ mach64_ext_writeb(uint32_t addr, uint8_t val, void *priv)
             case 0x17:
                 WRITE8(addr, mach64->crtc_off_pitch, val);
                 svga_recalctimings(&mach64->svga);
-                svga->fullchange = changeframecount;
+                svga->fullchange = svga->monitor->mon_changeframecount;
                 break;
 
             case 0x18:
@@ -4541,7 +4541,7 @@ mach64_force_redraw(void *priv)
 {
     mach64_t *mach64 = (mach64_t *) priv;
 
-    mach64->svga.fullchange = changeframecount;
+    mach64->svga.fullchange = mach64->svga.monitor->mon_changeframecount;
 }
 
 // clang-format off
diff --git a/src/video/vid_ati_mach8.c b/src/video/vid_ati_mach8.c
index c31c0d237..3d019251e 100644
--- a/src/video/vid_ati_mach8.c
+++ b/src/video/vid_ati_mach8.c
@@ -40,27 +40,34 @@
 #include <86box/vid_svga_render.h>
 #include <86box/vid_ati_eeprom.h>
 
-#define BIOS_MACH8_ROM_PATH      "roms/video/mach8/BIOS.BIN"
-#define BIOS_MACH32_ISA_ROM_PATH "roms/video/mach32/MACH32ISA.VBI"
-#define BIOS_MACH32_VLB_ROM_PATH "roms/video/mach32/MACH32VLB.VBI"
-#define BIOS_MACH32_PCI_ROM_PATH "roms/video/mach32/MACH32PCI.BIN"
+#define BIOS_MACH8_ROM_PATH             "roms/video/mach8/BIOS.BIN"
+#define BIOS_MACH32_ISA_ROM_PATH        "roms/video/mach32/ATi Mach32 Graphics Pro ISA.BIN"
+#define BIOS_MACH32_VLB_ROM_PATH        "roms/video/mach32/MACH32VLB.VBI"
+#define BIOS_MACH32_MCA_ROM_PATH        "roms/video/mach32/MACH32MCA_Olivetti.BIN"
+#define BIOS_MACH32_PCI_ROM_PATH        "roms/video/mach32/intelopt_00000.rom"
 
 typedef struct mach_t {
     ati_eeprom_t eeprom;
     svga_t       svga;
 
     rom_t         bios_rom;
+    rom_t		  bios_rom2;
     mem_mapping_t mmio_linear_mapping;
 
-    int     mca_bus;
-    int     pci_bus;
-    int     vlb_bus;
+    int mca_bus;
+    int pci_bus;
+    int vlb_bus;
+    int has_bios;
+
     uint8_t regs[256];
     uint8_t pci_regs[256];
     uint8_t int_line;
     uint8_t pci_slot;
     uint8_t irq_state;
+
     int     index;
+    int     ramdac_type;
+    int     old_mode;
 
     uint32_t memory;
 
@@ -68,14 +75,20 @@ typedef struct mach_t {
     uint16_t config2;
 
     uint8_t  pos_regs[8];
+    uint8_t  pci_cntl_reg;
     uint8_t  cursor_col_0;
     uint8_t  cursor_col_1;
     uint8_t  ext_cur_col_0_r;
     uint8_t  ext_cur_col_1_r;
     uint8_t  ext_cur_col_0_g;
     uint8_t  ext_cur_col_1_g;
+    uint16_t cursor_col_0_rg;
+    uint16_t cursor_col_1_rg;
+    uint16_t cursor_col_b;
     uint16_t cursor_offset_lo;
     uint16_t cursor_offset_hi;
+    uint16_t cursor_offset_hi_reg;
+    uint16_t cursor_vh_offset;
     uint16_t cursor_x;
     uint16_t cursor_y;
     uint16_t misc;
@@ -85,6 +98,7 @@ typedef struct mach_t {
     uint8_t  ap_size;
     uint8_t  bank_w;
     uint8_t  bank_r;
+    uint16_t shadow_set;
 
     struct {
         uint8_t  line_idx;
@@ -166,17 +180,18 @@ typedef struct mach_t {
     } accel;
 
     atomic_int force_busy;
-    atomic_int force_busy2;
 } mach_t;
 
 static video_timings_t timing_gfxultra_isa = { .type = VIDEO_ISA, .write_b = 3, .write_w = 3, .write_l = 6, .read_b = 5, .read_w = 5, .read_l = 10 };
 static video_timings_t timing_mach32_vlb   = { .type = VIDEO_BUS, .write_b = 2, .write_w = 2, .write_l = 1, .read_b = 20, .read_w = 20, .read_l = 21 };
+static video_timings_t timing_mach32_mca   = { .type = VIDEO_MCA, .write_b = 4, .write_w = 5, .write_l = 10, .read_b = 5, .read_w = 5, .read_l = 10 };
 static video_timings_t timing_mach32_pci   = { .type = VIDEO_PCI, .write_b = 2, .write_w = 2, .write_l = 1, .read_b = 20, .read_w = 20, .read_l = 21 };
 
 static void     mach_accel_outb(uint16_t port, uint8_t val, void *priv);
 static void     mach_accel_outw(uint16_t port, uint16_t val, void *priv);
 static uint8_t  mach_accel_inb(uint16_t port, void *priv);
 static uint16_t mach_accel_inw(uint16_t port, void *priv);
+static uint8_t  mach_in(uint16_t addr, void *priv);
 
 static void mach32_updatemapping(mach_t *mach);
 
@@ -198,64 +213,67 @@ mach_log(const char *fmt, ...)
 #    define mach_log(fmt, ...)
 #endif
 
-#define READ_PIXTRANS_BYTE_IO(cx, n, vgacore) \
-    if ((mach->accel.cmd_type == 2) || (mach->accel.cmd_type == 5)) { \
-        if (vgacore) { \
-            if ((svga->bpp == 15) || (svga->bpp == 16)) \
-                if (n == 0) \
-                    mach->accel.pix_trans[(n)] = vram_w[(dev->accel.dest + (cx) + (n)) & (svga->vram_mask >> 1)] & 0xff; \
-                else \
-                    mach->accel.pix_trans[(n)] = vram_w[(dev->accel.dest + (cx) + (n)) & (svga->vram_mask >> 1)] >> 8; \
-            else \
-                mach->accel.pix_trans[(n)] = svga->vram[(dev->accel.dest + (cx) + (n)) & svga->vram_mask]; \
-        } else \
-            mach->accel.pix_trans[(n)] = dev->vram[(dev->accel.dest + (cx) + (n)) & dev->vram_mask]; \
+#define WRITE8(addr, var, val)                        \
+    switch ((addr) & 1) {                             \
+        case 0:                                       \
+            var = (var & 0xff00) | (val);             \
+            break;                                    \
+        case 1:                                       \
+            var = (var & 0x00ff) | ((val) << 8);      \
+            break;                                    \
     }
 
-#define READ_PIXTRANS_WORD(cx, n, vgacore)                                                                    \
+#define READ8(addr, var)                \
+    switch ((addr) & 1) {               \
+        case 0:                         \
+            temp = (var) & 0xff;        \
+            break;                      \
+        case 1:                         \
+            temp = ((var) >> 8) & 0xff; \
+            break;                      \
+    }
+
+#define READ_PIXTRANS_BYTE_IO(cx, n) \
+    if ((mach->accel.cmd_type == 2) || (mach->accel.cmd_type == 5)) { \
+        if (dev->bpp) { \
+            if (n == 0) \
+                mach->accel.pix_trans[(n)] = vram_w[(dev->accel.dest + (cx) + (n)) & (dev->vram_mask >> 1)] & 0xff; \
+            else \
+                mach->accel.pix_trans[(n)] = vram_w[(dev->accel.dest + (cx) + (n)) & (dev->vram_mask >> 1)] >> 8; \
+        } else { \
+            mach->accel.pix_trans[(n)] = dev->vram[(dev->accel.dest + (cx) + (n)) & dev->vram_mask]; \
+        } \
+    }
+
+#define READ_PIXTRANS_WORD(cx, n)                                                                    \
     if ((cmd == 0) || (cmd == 1) || (cmd == 5) || (mach->accel.cmd_type == -1)) { \
-        if (vgacore) { \
-            if ((svga->bpp == 15) || (svga->bpp == 16)) { \
-                temp = vram_w[((dev->accel.cy * dev->pitch) + (cx) + (n)) & (svga->vram_mask >> 1)];             \
-            } else { \
-                temp = svga->vram[((dev->accel.cy * dev->pitch) + (cx) + (n)) & svga->vram_mask];             \
-                temp |= (svga->vram[((dev->accel.cy * dev->pitch) + (cx) + (n + 1)) & svga->vram_mask] << 8); \
-            } \
+        if (dev->bpp) { \
+            temp = vram_w[((dev->accel.cy * dev->pitch) + (cx) + (n)) & (dev->vram_mask >> 1)];             \
         } else { \
             temp = dev->vram[((dev->accel.cy * dev->pitch) + (cx) + (n)) & dev->vram_mask];             \
             temp |= (dev->vram[((dev->accel.cy * dev->pitch) + (cx) + (n + 1)) & dev->vram_mask] << 8); \
         } \
     } else if ((mach->accel.cmd_type == 2) || (mach->accel.cmd_type == 5)) { \
-        if ((svga->bpp == 8) || (svga->bpp == 24)) { \
-            if (vgacore) { \
-                temp = svga->vram[((dev->accel.dest) + (cx) + (n)) & svga->vram_mask];                           \
-                temp |= (svga->vram[((dev->accel.dest) + (cx) + (n + 1)) & svga->vram_mask] << 8);               \
-            } else { \
-                temp = dev->vram[((dev->accel.dest) + (cx) + (n)) & dev->vram_mask];                           \
-                temp |= (dev->vram[((dev->accel.dest) + (cx) + (n + 1)) & dev->vram_mask] << 8);               \
-            } \
-        } else if ((svga->bpp == 15) || (svga->bpp == 16)) { \
-            temp = vram_w[((dev->accel.dest) + (cx) + (n)) & (svga->vram_mask >> 1)];                           \
+        if (!dev->bpp) { \
+            temp = dev->vram[((dev->accel.dest) + (cx) + (n)) & dev->vram_mask];                           \
+            temp |= (dev->vram[((dev->accel.dest) + (cx) + (n + 1)) & dev->vram_mask] << 8);               \
+        } else { \
+            temp = vram_w[((dev->accel.dest) + (cx) + (n)) & (dev->vram_mask >> 1)];                           \
         } \
     } else if ((mach->accel.cmd_type == 3) || (mach->accel.cmd_type == 4)) { \
-        if ((svga->bpp == 8) || (svga->bpp == 24)) { \
-            if (vgacore) { \
-                temp = svga->vram[((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (cx) + (n)) & svga->vram_mask]; \
-                temp |= (svga->vram[((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (cx) + (n + 1)) & svga->vram_mask] << 8); \
-            } else { \
-                temp = dev->vram[((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (cx) + (n)) & dev->vram_mask]; \
-                temp |= (dev->vram[((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (cx) + (n + 1)) & dev->vram_mask] << 8); \
-            } \
-        } else if ((svga->bpp == 15) || (svga->bpp == 16)) { \
-            temp = vram_w[((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (cx) + (n)) & (svga->vram_mask >> 1)]; \
+        if (!dev->bpp) { \
+            temp = dev->vram[((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (cx) + (n)) & dev->vram_mask]; \
+            temp |= (dev->vram[((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (cx) + (n + 1)) & dev->vram_mask] << 8); \
+        } else { \
+            temp = vram_w[((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (cx) + (n)) & (dev->vram_mask >> 1)]; \
         } \
     }
 
-#define READ(addr, dat, vgacore) \
-        if ((svga->bpp == 8) || (svga->bpp == 24)) \
-            dat = vgacore ? (svga->vram[(addr) & (svga->vram_mask)]) : (dev->vram[(addr) & (dev->vram_mask)]); \
-        else if ((svga->bpp == 15) || (svga->bpp == 16)) \
-            dat = vram_w[(addr) & (svga->vram_mask >> 1)];
+#define READ(addr, dat) \
+        if (!dev->bpp) \
+            dat = dev->vram[(addr) & (dev->vram_mask)]; \
+        else \
+            dat = vram_w[(addr) & (dev->vram_mask >> 1)];
 
 #define MIX(mixmode, dest_dat, src_dat)                                                       \
     {                                                                                         \
@@ -360,18 +378,13 @@ mach_log(const char *fmt, ...)
     }
 
 
-#define WRITE(addr, dat, vgacore)                                         \
-    if ((svga->bpp == 8) || (svga->bpp == 24)) { \
-        if (vgacore) { \
-            svga->vram[((addr)) & (svga->vram_mask)]                = dat; \
-            svga->changedvram[(((addr)) & (svga->vram_mask)) >> 12] = changeframecount; \
-        } else { \
-            dev->vram[((addr)) & (dev->vram_mask)]                = dat; \
-            dev->changedvram[(((addr)) & (dev->vram_mask)) >> 12] = changeframecount; \
-        } \
-    } else if ((svga->bpp == 15) || (svga->bpp == 16)) { \
-        vram_w[((addr)) & (svga->vram_mask >> 1)]                = dat; \
-        svga->changedvram[(((addr)) & (svga->vram_mask >> 1)) >> 11] = changeframecount; \
+#define WRITE(addr, dat)                                         \
+    if (!dev->bpp) { \
+        dev->vram[((addr)) & (dev->vram_mask)]                = dat; \
+        dev->changedvram[(((addr)) & (dev->vram_mask)) >> 12] = changeframecount; \
+    } else { \
+        vram_w[((addr)) & (dev->vram_mask >> 1)]                = dat; \
+        dev->changedvram[(((addr)) & (dev->vram_mask >> 1)) >> 11] = changeframecount; \
     }
 
 static int
@@ -393,7 +406,7 @@ mach_pixel_read(mach_t *mach)
 }
 
 static void
-mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint32_t cpu_dat, mach_t *mach, ibm8514_t *dev, UNUSED(int len))
+mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint32_t cpu_dat, mach_t *mach, ibm8514_t *dev)
 {
     svga_t   *svga = &mach->svga;
     int       compare_mode;
@@ -408,7 +421,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
     uint16_t  src_dat  = 0;
     uint16_t  dest_dat = 0;
     uint16_t  old_dest_dat;
-    uint16_t *vram_w    = (uint16_t *) svga->vram;
+    uint16_t *vram_w    = (uint16_t *) dev->vram;
     uint16_t  mix       = 0;
     int16_t   clip_l    = dev->accel.clip_left & 0x7ff;
     int16_t   clip_t    = dev->accel.clip_top & 0x7ff;
@@ -417,7 +430,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
     uint32_t  mono_dat0 = 0;
     uint32_t  mono_dat1 = 0;
 
-    if ((svga->bpp == 8) || (svga->bpp == 24)) {
+    if (!dev->bpp) {
         rd_mask &= 0xff;
         dest_cmp_clr &= 0xff;
     }
@@ -431,13 +444,12 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
 
     if ((mono_src == 2) || (bkgd_sel == 2) || (frgd_sel == 2) || mach_pixel_read(mach)) {
         mach->force_busy  = 1;
-        mach->force_busy2 = 1;
         dev->force_busy   = 1;
         dev->force_busy2  = 1;
     }
 
     if (cpu_input) {
-        if ((svga->bpp == 15) || (svga->bpp == 16)) {
+        if (dev->bpp) {
             if ((mach->accel.dp_config & 0x200) && (count == 2)) {
                 count >>= 1;
             }
@@ -515,7 +527,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                             mix = 1;
                             break;
                         case 1:
-                            if (dev->accel.temp_cnt == 0) {
+                            if (!dev->accel.temp_cnt) {
                                 dev->accel.temp_cnt = 8;
                                 mix_dat >>= 8;
                             }
@@ -540,10 +552,10 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                             }
                             break;
                         case 3:
-                            if ((svga->bpp == 15) || (svga->bpp == 16)) {
-                                READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), mix, dev->local);
+                            if (dev->bpp) {
+                                READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), mix);
                             } else {
-                                READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), mix, dev->local);
+                                READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), mix);
                             }
                             mix = (mix & rd_mask) == rd_mask;
                             break;
@@ -554,10 +566,10 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
 
                     if (((dev->accel.dx) >= clip_l) && ((dev->accel.dx) <= clip_r) && ((dev->accel.dy) >= clip_t) && ((dev->accel.dy) <= clip_b)) {
                         if (mach->accel.linedraw_opt & 0x02) {
-                            if ((svga->bpp == 15) || (svga->bpp == 16)) {
-                                READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), poly_src, dev->local);
+                            if (dev->bpp) {
+                                READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), poly_src);
                             } else {
-                                READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), poly_src, dev->local);
+                                READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), poly_src);
                             }
                             poly_src = ((poly_src & rd_mask) == rd_mask);
                             if (poly_src)
@@ -579,10 +591,10 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                                     if (mach_pixel_read(mach))
                                         src_dat = cpu_dat;
                                     else {
-                                        if ((svga->bpp == 15) || (svga->bpp == 16)) {
-                                            READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), mix, dev->local);
+                                        if (dev->bpp) {
+                                            READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), src_dat);
                                         } else {
-                                            READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), mix, dev->local);
+                                            READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), src_dat);
                                         }
                                         if (mono_src == 3) {
                                             src_dat = (src_dat & rd_mask) == rd_mask;
@@ -600,10 +612,10 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                                     break;
                             }
 
-                            if ((svga->bpp == 15) || (svga->bpp == 16)) {
-                                READ((mach->accel.ge_offset << 1) + ((dev->accel.dy) * (dev->pitch)) + (dev->accel.dx), dest_dat, dev->local);
+                            if (dev->bpp) {
+                                READ((mach->accel.ge_offset << 1) + ((dev->accel.dy) * (dev->pitch)) + (dev->accel.dx), dest_dat);
                             } else {
-                                READ((mach->accel.ge_offset << 2) + ((dev->accel.dy) * (dev->pitch)) + (dev->accel.dx), dest_dat, dev->local);
+                                READ((mach->accel.ge_offset << 2) + ((dev->accel.dy) * (dev->pitch)) + (dev->accel.dx), dest_dat);
                             }
                         }
 
@@ -645,17 +657,17 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                         if (mach->accel.dp_config & 0x10) {
                             if (mach->accel.linedraw_opt & 0x04) {
                                 if (dev->accel.sx < mach->accel.width) {
-                                    if ((svga->bpp == 15) || (svga->bpp == 16)) {
-                                        WRITE((mach->accel.ge_offset << 1) + ((dev->accel.dy) * (dev->pitch)) + (dev->accel.dx), dest_dat, dev->local);
+                                    if (dev->bpp) {
+                                        WRITE((mach->accel.ge_offset << 1) + ((dev->accel.dy) * (dev->pitch)) + (dev->accel.dx), dest_dat);
                                     } else {
-                                        WRITE((mach->accel.ge_offset << 2) + ((dev->accel.dy) * (dev->pitch)) + (dev->accel.dx), dest_dat, dev->local);
+                                        WRITE((mach->accel.ge_offset << 2) + ((dev->accel.dy) * (dev->pitch)) + (dev->accel.dx), dest_dat);
                                     }
                                 }
                             } else {
-                                if ((svga->bpp == 15) || (svga->bpp == 16)) {
-                                    WRITE((mach->accel.ge_offset << 1) + ((dev->accel.dy) * (dev->pitch)) + (dev->accel.dx), dest_dat, dev->local);
+                                if (dev->bpp) {
+                                    WRITE((mach->accel.ge_offset << 1) + ((dev->accel.dy) * (dev->pitch)) + (dev->accel.dx), dest_dat);
                                 } else {
-                                    WRITE((mach->accel.ge_offset << 2) + ((dev->accel.dy) * (dev->pitch)) + (dev->accel.dx), dest_dat, dev->local);
+                                    WRITE((mach->accel.ge_offset << 2) + ((dev->accel.dy) * (dev->pitch)) + (dev->accel.dx), dest_dat);
                                 }
                             }
                         }
@@ -666,10 +678,10 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                     else if ((mono_src != 1) && (dev->accel.sx >= mach->accel.width))
                         break;
 
-                    if (svga->bpp == 8)
-                        cpu_dat >>= 8;
-                    else
+                    if (dev->bpp)
                         cpu_dat >>= 16;
+                    else
+                        cpu_dat >>= 8;
 
                     switch (mach->accel.linedraw_opt & 0xe0) {
                         case 0x00:
@@ -726,7 +738,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                             mix = 1;
                             break;
                         case 1:
-                            if (dev->accel.temp_cnt == 0) {
+                            if (!dev->accel.temp_cnt) {
                                 dev->accel.temp_cnt = 8;
                                 mix_dat >>= 8;
                             }
@@ -751,10 +763,10 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                             }
                             break;
                         case 3:
-                            if ((svga->bpp == 15) || (svga->bpp == 16)) {
-                                READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), mix, dev->local);
+                            if (dev->bpp) {
+                                READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), mix);
                             } else {
-                                READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), mix, dev->local);
+                                READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), mix);
                             }
                             mix = (mix & rd_mask) == rd_mask;
                             break;
@@ -765,10 +777,10 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
 
                     if (((dev->accel.dx) >= clip_l) && ((dev->accel.dx) <= clip_r) && ((dev->accel.dy) >= clip_t) && ((dev->accel.dy) <= clip_b)) {
                         if (mach->accel.linedraw_opt & 0x02) {
-                            if ((svga->bpp == 15) || (svga->bpp == 16)) {
-                                READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), poly_src, dev->local);
+                            if (dev->bpp) {
+                                READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), poly_src);
                             } else {
-                                READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), poly_src, dev->local);
+                                READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), poly_src);
                             }
                             poly_src = ((poly_src & rd_mask) == rd_mask);
                             if (poly_src)
@@ -790,10 +802,10 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                                     if (mach_pixel_read(mach))
                                         src_dat = cpu_dat;
                                     else {
-                                        if ((svga->bpp == 15) || (svga->bpp == 16)) {
-                                            READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), mix, dev->local);
+                                        if (dev->bpp) {
+                                            READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), src_dat);
                                         } else {
-                                            READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), mix, dev->local);
+                                            READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), src_dat);
                                         }
                                         if (mono_src == 3) {
                                             src_dat = (src_dat & rd_mask) == rd_mask;
@@ -811,10 +823,10 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                                     break;
                             }
 
-                            if ((svga->bpp == 15) || (svga->bpp == 16)) {
-                                READ((mach->accel.ge_offset << 1) + ((dev->accel.dy) * (dev->pitch)) + (dev->accel.dx), dest_dat, dev->local);
+                            if (dev->bpp) {
+                                READ((mach->accel.ge_offset << 1) + ((dev->accel.dy) * (dev->pitch)) + (dev->accel.dx), dest_dat);
                             } else {
-                                READ((mach->accel.ge_offset << 2) + ((dev->accel.dy) * (dev->pitch)) + (dev->accel.dx), dest_dat, dev->local);
+                                READ((mach->accel.ge_offset << 2) + ((dev->accel.dy) * (dev->pitch)) + (dev->accel.dx), dest_dat);
                             }
                         }
 
@@ -856,17 +868,17 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                         if (mach->accel.dp_config & 0x10) {
                             if (mach->accel.linedraw_opt & 0x04) {
                                 if (dev->accel.sx < mach->accel.width) {
-                                    if ((svga->bpp == 15) || (svga->bpp == 16)) {
-                                        WRITE((mach->accel.ge_offset << 1) + ((dev->accel.dy) * (dev->pitch)) + (dev->accel.dx), dest_dat, dev->local);
+                                    if (dev->bpp) {
+                                        WRITE((mach->accel.ge_offset << 1) + ((dev->accel.dy) * (dev->pitch)) + (dev->accel.dx), dest_dat);
                                     } else {
-                                        WRITE((mach->accel.ge_offset << 2) + ((dev->accel.dy) * (dev->pitch)) + (dev->accel.dx), dest_dat, dev->local);
+                                        WRITE((mach->accel.ge_offset << 2) + ((dev->accel.dy) * (dev->pitch)) + (dev->accel.dx), dest_dat);
                                     }
                                 }
                             } else {
-                                if ((svga->bpp == 15) || (svga->bpp == 16)) {
-                                    WRITE((mach->accel.ge_offset << 1) + ((dev->accel.dy) * (dev->pitch)) + (dev->accel.dx), dest_dat, dev->local);
+                                if (dev->bpp) {
+                                    WRITE((mach->accel.ge_offset << 1) + ((dev->accel.dy) * (dev->pitch)) + (dev->accel.dx), dest_dat);
                                 } else {
-                                    WRITE((mach->accel.ge_offset << 2) + ((dev->accel.dy) * (dev->pitch)) + (dev->accel.dx), dest_dat, dev->local);
+                                    WRITE((mach->accel.ge_offset << 2) + ((dev->accel.dy) * (dev->pitch)) + (dev->accel.dx), dest_dat);
                                 }
                             }
                         }
@@ -877,10 +889,10 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                     else if ((mono_src != 1) && (dev->accel.sx >= mach->accel.width))
                         break;
 
-                    if (svga->bpp == 8)
-                        cpu_dat >>= 8;
-                    else
+                    if (dev->bpp)
                         cpu_dat >>= 16;
+                    else
+                        cpu_dat >>= 8;
 
                     if (mach->accel.linedraw_opt & 0x40) {
                         dev->accel.dy += mach->accel.stepy;
@@ -986,7 +998,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                 }
 
                 dev->accel.sy = 0;
-                if ((svga->bpp == 15) || (svga->bpp == 16))
+                if (dev->bpp)
                     dev->accel.dest = (mach->accel.ge_offset << 1) + (dev->accel.dy * (dev->pitch));
                 else
                     dev->accel.dest = (mach->accel.ge_offset << 2) + (dev->accel.dy * (dev->pitch));
@@ -1026,7 +1038,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                     mach_log("BitBLT: Src Indeterminate X: width = %d, coordinates: %d,%d px, end: %d px, stepx = %d, dpconfig = %04x, oddwidth = %d.\n", mach->accel.src_width, dev->accel.cx, dev->accel.cy, mach->accel.src_x_end, mach->accel.src_stepx, mach->accel.dp_config, mach->accel.src_width & 1);
                 }
                 mach->accel.sx = 0;
-                if ((svga->bpp == 15) || (svga->bpp == 16))
+                if (dev->bpp)
                     dev->accel.src = (mach->accel.ge_offset << 1) + (dev->accel.cy * (dev->pitch));
                 else
                     dev->accel.src = (mach->accel.ge_offset << 2) + (dev->accel.cy * (dev->pitch));
@@ -1078,7 +1090,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
             }
 
             if (frgd_sel == 5) {
-                if ((svga->bpp == 15) || (svga->bpp == 16)) {
+                if (dev->bpp) {
                     for (int x = 0; x <= mach->accel.patt_len; x += 2) {
                         mach->accel.color_pattern_word[x + (mach->accel.color_pattern_idx & 1)] = (mach->accel.patt_data[x & mach->accel.patt_len] & 0xff);
                         mach->accel.color_pattern_word[x + (mach->accel.color_pattern_idx & 1)] |= (mach->accel.patt_data[(x + 1) & mach->accel.patt_len] << 8);
@@ -1159,7 +1171,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                         }
                         break;
                     case 3:
-                        READ(dev->accel.src + ((dev->accel.cx)), mix, dev->local);
+                        READ(dev->accel.src + ((dev->accel.cx)), mix);
                         mix = (mix & rd_mask) == rd_mask;
                         break;
 
@@ -1169,7 +1181,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
 
                 if (((dev->accel.dx) >= clip_l) && ((dev->accel.dx) <= clip_r) && ((dev->accel.dy) >= clip_t) && ((dev->accel.dy) <= clip_b)) {
                     if (mach->accel.dp_config & 0x02) {
-                        READ(dev->accel.src + (dev->accel.cx), poly_src, dev->local);
+                        READ(dev->accel.src + (dev->accel.cx), poly_src);
                         poly_src = ((poly_src & rd_mask) == rd_mask);
                         if (poly_src)
                             mach->accel.poly_fill = !mach->accel.poly_fill;
@@ -1190,7 +1202,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                                 if (mach_pixel_read(mach))
                                     src_dat = cpu_dat;
                                 else {
-                                    READ(dev->accel.src + (dev->accel.cx), src_dat, dev->local);
+                                    READ(dev->accel.src + (dev->accel.cx), src_dat);
                                     if (mono_src == 3) {
                                         src_dat = (src_dat & rd_mask) == rd_mask;
                                     }
@@ -1198,7 +1210,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                                 break;
                             case 5:
                                 if (mix) {
-                                    if ((svga->bpp == 15) || (svga->bpp == 16)) {
+                                    if (dev->bpp) {
                                         src_dat = mach->accel.color_pattern_word[mach->accel.color_pattern_idx];
                                     } else {
                                         src_dat = mach->accel.color_pattern[mach->accel.color_pattern_idx];
@@ -1214,12 +1226,12 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
 
                     if ((svga->bpp == 24) && (mono_src == 1) && (frgd_sel == 5) && (mach->accel.patt_len_reg & 0x4000)) {
                         if (dev->accel.sy & 1) {
-                            READ(dev->accel.dest + dev->accel.dx - dev->ext_pitch, dest_dat, dev->local);
+                            READ(dev->accel.dest + dev->accel.dx - dev->ext_pitch, dest_dat);
                         } else {
-                            READ(dev->accel.dest + dev->accel.dx, dest_dat, dev->local);
+                            READ(dev->accel.dest + dev->accel.dx, dest_dat);
                         }
                     } else {
-                        READ(dev->accel.dest + dev->accel.dx, dest_dat, dev->local);
+                        READ(dev->accel.dest + dev->accel.dx, dest_dat);
                     }
 
                     switch (compare_mode) {
@@ -1260,20 +1272,20 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                     if (mach->accel.dp_config & 0x10) {
                         if ((svga->bpp == 24) && (mono_src == 1) && (frgd_sel == 5) && (mach->accel.patt_len_reg & 0x4000)) {
                             if (dev->accel.sy & 1) {
-                                WRITE(dev->accel.dest + dev->accel.dx - dev->ext_pitch, dest_dat, dev->local);
+                                WRITE(dev->accel.dest + dev->accel.dx - dev->ext_pitch, dest_dat);
                             } else {
-                                WRITE(dev->accel.dest + dev->accel.dx, dest_dat, dev->local);
+                                WRITE(dev->accel.dest + dev->accel.dx, dest_dat);
                             }
                         } else {
-                            WRITE(dev->accel.dest + dev->accel.dx, dest_dat, dev->local);
+                            WRITE(dev->accel.dest + dev->accel.dx, dest_dat);
                         }
                     }
                 }
 
-                if ((svga->bpp == 8) || (svga->bpp == 24))
-                    cpu_dat >>= 8;
-                else
+                if (dev->bpp)
                     cpu_dat >>= 16;
+                else
+                    cpu_dat >>= 8;
 
                 if ((mono_src == 3) || (frgd_sel == 3) || (bkgd_sel == 3)) {
                     dev->accel.cx += mach->accel.src_stepx;
@@ -1285,7 +1297,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                         else
                             dev->accel.cx -= mach->accel.src_width;
                         dev->accel.cy += (mach->accel.src_y_dir ? 1 : -1);
-                        if ((svga->bpp == 15) || (svga->bpp == 16))
+                        if (dev->bpp)
                             dev->accel.src = (mach->accel.ge_offset << 1) + (dev->accel.cy * (dev->pitch));
                         else
                             dev->accel.src = (mach->accel.ge_offset << 2) + (dev->accel.cy * (dev->pitch));
@@ -1313,7 +1325,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                         mach->accel.color_pattern_idx = 0;
                 }
 
-                if ((svga->bpp == 15) || (svga->bpp == 16)) {
+                if (dev->bpp) {
                     mach->accel.color_pattern_idx = (mach->accel.color_pattern_idx + mach->accel.stepx) & mach->accel.patt_len;
                     mach->accel.color_pattern_idx = (mach->accel.color_pattern_idx + mach->accel.stepx) & mach->accel.patt_len;
                 }
@@ -1330,7 +1342,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                     dev->accel.dy += mach->accel.stepy;
                     dev->accel.sy++;
 
-                    if ((svga->bpp == 15) || (svga->bpp == 16))
+                    if (dev->bpp)
                         dev->accel.dest = (mach->accel.ge_offset << 1) + (dev->accel.dy * (dev->pitch));
                     else {
                         dev->accel.dest = (mach->accel.ge_offset << 2) + (dev->accel.dy * (dev->pitch));
@@ -1455,10 +1467,10 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                                     break;
                             }
 
-                            if ((svga->bpp == 15) || (svga->bpp == 16)) {
-                                READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat, dev->local);
+                            if (dev->bpp) {
+                                READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                             } else {
-                                READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat, dev->local);
+                                READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                             }
 
                             switch (compare_mode) {
@@ -1496,10 +1508,10 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                                 }
                             }
                             if ((mach->accel.dp_config & 0x10) && (cmd_type == 3)) {
-                                if ((svga->bpp == 15) || (svga->bpp == 16)) {
-                                    WRITE((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat, dev->local);
+                                if (dev->bpp) {
+                                    WRITE((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                                 } else {
-                                    WRITE((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat, dev->local);
+                                    WRITE((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                                 }
                             }
                         } else
@@ -1508,10 +1520,10 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                         if (!count)
                             break;
 
-                        if (svga->bpp == 8)
-                            cpu_dat >>= 8;
-                        else
+                        if (dev->bpp)
                             cpu_dat >>= 16;
+                        else
+                            cpu_dat >>= 8;
 
                         if (mach->accel.err >= 0) {
                             dev->accel.cy += mach->accel.stepy;
@@ -1577,10 +1589,10 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                                     break;
                             }
 
-                            if ((svga->bpp == 15) || (svga->bpp == 16)) {
-                                READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat, dev->local);
+                            if (dev->bpp) {
+                                READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                             } else {
-                                READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat, dev->local);
+                                READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                             }
 
                             switch (compare_mode) {
@@ -1621,17 +1633,17 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                             if ((mach->accel.dp_config & 0x10) && (cmd_type == 3)) {
                                 if (mach->accel.linedraw_opt & 0x04) {
                                     if (dev->accel.sx < mach->accel.width) {
-                                        if ((svga->bpp == 15) || (svga->bpp == 16)) {
-                                            WRITE((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat, dev->local);
+                                        if (dev->bpp) {
+                                            WRITE((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                                         } else {
-                                            WRITE((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat, dev->local);
+                                            WRITE((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                                         }
                                     }
                                 } else {
-                                    if ((svga->bpp == 15) || (svga->bpp == 16)) {
-                                        WRITE((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat, dev->local);
+                                    if (dev->bpp) {
+                                        WRITE((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                                     } else {
-                                        WRITE((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat, dev->local);
+                                        WRITE((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                                     }
                                 }
                             }
@@ -1641,10 +1653,10 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                         if (dev->accel.sx >= mach->accel.width)
                             break;
 
-                        if (svga->bpp == 8)
-                            cpu_dat >>= 8;
-                        else
+                        if (dev->bpp)
                             cpu_dat >>= 16;
+                        else
+                            cpu_dat >>= 8;
 
                         if (mach->accel.err >= 0) {
                             dev->accel.cy += mach->accel.stepy;
@@ -1699,10 +1711,10 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                                     break;
                             }
 
-                            if ((svga->bpp == 15) || (svga->bpp == 16)) {
-                                READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat, dev->local);
+                            if (dev->bpp) {
+                                READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                             } else {
-                                READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat, dev->local);
+                                READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                             }
                             switch (compare_mode) {
                                 case 1:
@@ -1740,10 +1752,10 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                             }
 
                             if ((mach->accel.dp_config & 0x10) && (cmd_type == 3)) {
-                                if ((svga->bpp == 15) || (svga->bpp == 16)) {
-                                    WRITE((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat, dev->local);
+                                if (dev->bpp) {
+                                    WRITE((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                                 } else {
-                                    WRITE((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat, dev->local);
+                                    WRITE((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                                 }
                             }
                         } else
@@ -1752,10 +1764,10 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                         if (!count)
                             break;
 
-                        if (svga->bpp == 8)
-                            cpu_dat >>= 8;
-                        else
+                        if (dev->bpp)
                             cpu_dat >>= 16;
+                        else
+                            cpu_dat >>= 8;
 
                         if (mach->accel.err >= 0) {
                             dev->accel.cx += mach->accel.stepx;
@@ -1821,10 +1833,10 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                                     break;
                             }
 
-                            if ((svga->bpp == 15) || (svga->bpp == 16)) {
-                                READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat, dev->local);
+                            if (dev->bpp) {
+                                READ((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                             } else {
-                                READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat, dev->local);
+                                READ((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                             }
 
                             switch (compare_mode) {
@@ -1865,17 +1877,17 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                             if ((mach->accel.dp_config & 0x10) && (cmd_type == 3)) {
                                 if (mach->accel.linedraw_opt & 0x04) {
                                     if (dev->accel.sx < mach->accel.width) {
-                                        if ((svga->bpp == 15) || (svga->bpp == 16)) {
-                                            WRITE((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat, dev->local);
+                                        if (dev->bpp) {
+                                            WRITE((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                                         } else {
-                                            WRITE((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat, dev->local);
+                                            WRITE((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                                         }
                                     }
                                 } else {
-                                    if ((svga->bpp == 15) || (svga->bpp == 16)) {
-                                        WRITE((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat, dev->local);
+                                    if (dev->bpp) {
+                                        WRITE((mach->accel.ge_offset << 1) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                                     } else {
-                                        WRITE((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat, dev->local);
+                                        WRITE((mach->accel.ge_offset << 2) + ((dev->accel.cy) * (dev->pitch)) + (dev->accel.cx), dest_dat);
                                     }
                                 }
                             }
@@ -1885,10 +1897,10 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                         if (dev->accel.sx >= mach->accel.width)
                             break;
 
-                        if (svga->bpp == 8)
-                            cpu_dat >>= 8;
-                        else
+                        if (dev->bpp)
                             cpu_dat >>= 16;
+                        else
+                            cpu_dat >>= 8;
 
                         if (mach->accel.err >= 0) {
                             dev->accel.cx += mach->accel.stepx;
@@ -1960,7 +1972,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                     mach->accel.stepy = 0;
                 }
 
-                if ((svga->bpp == 15) || (svga->bpp == 16))
+                if (dev->bpp)
                     dev->accel.dest = (mach->accel.ge_offset << 1) + (dev->accel.dy * (dev->pitch));
                 else
                     dev->accel.dest = (mach->accel.ge_offset << 2) + (dev->accel.dy * (dev->pitch));
@@ -1997,7 +2009,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                 }
 
                 mach->accel.sx = 0;
-                if ((svga->bpp == 15) || (svga->bpp == 16))
+                if (dev->bpp)
                     dev->accel.src = (mach->accel.ge_offset << 1) + (dev->accel.cy * (dev->pitch));
                 else
                     dev->accel.src = (mach->accel.ge_offset << 2) + (dev->accel.cy * (dev->pitch));
@@ -2064,7 +2076,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                         mix = 1;
                         break;
                     case 1:
-                        if (dev->accel.temp_cnt == 0) {
+                        if (!dev->accel.temp_cnt) {
                             dev->accel.temp_cnt = 8;
                             mix_dat >>= 8;
                         }
@@ -2089,7 +2101,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                         }
                         break;
                     case 3:
-                        READ(dev->accel.src + (dev->accel.cx), mix, dev->local);
+                        READ(dev->accel.src + (dev->accel.cx), mix);
                         mix = (mix & rd_mask) == rd_mask;
                         break;
 
@@ -2112,7 +2124,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                             if (mach_pixel_read(mach))
                                 src_dat = cpu_dat;
                             else {
-                                READ(dev->accel.src + (dev->accel.cx), src_dat, dev->local);
+                                READ(dev->accel.src + (dev->accel.cx), src_dat);
                                 if (mono_src == 3) {
                                     src_dat = (src_dat & rd_mask) == rd_mask;
                                 }
@@ -2135,7 +2147,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                             break;
                     }
 
-                    READ(dev->accel.dest + (dev->accel.dx), dest_dat, dev->local);
+                    READ(dev->accel.dest + (dev->accel.dx), dest_dat);
 
                     switch (compare_mode) {
                         case 1:
@@ -2173,14 +2185,14 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                     }
 
                     if (mach->accel.dp_config & 0x10) {
-                        WRITE(dev->accel.dest + (dev->accel.dx), dest_dat, dev->local);
+                        WRITE(dev->accel.dest + (dev->accel.dx), dest_dat);
                     }
                 }
 
-                if ((svga->bpp == 8) || (svga->bpp == 24))
-                    cpu_dat >>= 8;
-                else
+                if (dev->bpp)
                     cpu_dat >>= 16;
+                else
+                    cpu_dat >>= 8;
 
                 dev->accel.cx += mach->accel.src_stepx;
                 mach->accel.sx++;
@@ -2191,7 +2203,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                     } else
                         dev->accel.cx -= mach->accel.src_width;
                     dev->accel.cy += (mach->accel.src_y_dir ? 1 : -1);
-                    if ((svga->bpp == 15) || (svga->bpp == 16))
+                    if (dev->bpp)
                         dev->accel.src = (mach->accel.ge_offset << 1) + (dev->accel.cy * (dev->pitch));
                     else
                         dev->accel.src = (mach->accel.ge_offset << 2) + (dev->accel.cy * (dev->pitch));
@@ -2218,7 +2230,7 @@ mach_accel_start(int cmd_type, int cpu_input, int count, uint32_t mix_dat, uint3
                 if (dev->accel.sx >= mach->accel.width) {
                     dev->accel.sx = 0;
                     dev->accel.dy += mach->accel.stepy;
-                    if ((svga->bpp == 15) || (svga->bpp == 16))
+                    if (dev->bpp)
                         dev->accel.dest = (mach->accel.ge_offset << 1) + (dev->accel.dy * (dev->pitch));
                     else
                         dev->accel.dest = (mach->accel.ge_offset << 2) + (dev->accel.dy * (dev->pitch));
@@ -2255,25 +2267,25 @@ mach_accel_out_pixtrans(mach_t *mach, ibm8514_t *dev, UNUSED(uint16_t port), uin
         case 0x000: /*8-bit size*/
             if (mono_src == 2) {
                 if ((frgd_sel != 2) && (bkgd_sel != 2)) {
-                    if ((mach->accel.dp_config & 0x1000) && dev->local)
+                    if ((mach->accel.dp_config & 0x1000) && (dev->local >= 2))
                         val = (val >> 8) | (val << 8);
-                    mach_accel_start(mach->accel.cmd_type, 1, 8, val | (val << 16), 0, mach, dev, len);
+                    mach_accel_start(mach->accel.cmd_type, 1, 8, val | (val << 16), 0, mach, dev);
                 } else
-                    mach_accel_start(mach->accel.cmd_type, 1, 1, -1, val | (val << 16), mach, dev, len);
+                    mach_accel_start(mach->accel.cmd_type, 1, 1, -1, val | (val << 16), mach, dev);
             } else
-                mach_accel_start(mach->accel.cmd_type, 1, 1, -1, val | (val << 16), mach, dev, len);
+                mach_accel_start(mach->accel.cmd_type, 1, 1, -1, val | (val << 16), mach, dev);
             break;
         case 0x200: /*16-bit size*/
             if (mono_src == 2) {
                 if ((frgd_sel != 2) && (bkgd_sel != 2)) {
                     if (mach->accel.dp_config & 0x1000)
                         val = (val >> 8) | (val << 8);
-                    mach_accel_start(mach->accel.cmd_type, 1, 16, val | (val << 16), 0, mach, dev, len);
+                    mach_accel_start(mach->accel.cmd_type, 1, 16, val | (val << 16), 0, mach, dev);
                 } else {
-                    mach_accel_start(mach->accel.cmd_type, 1, 2, -1, val | (val << 16), mach, dev, len);
+                    mach_accel_start(mach->accel.cmd_type, 1, 2, -1, val | (val << 16), mach, dev);
                 }
             } else {
-                mach_accel_start(mach->accel.cmd_type, 1, 2, -1, val | (val << 16), mach, dev, len);
+                mach_accel_start(mach->accel.cmd_type, 1, 2, -1, val | (val << 16), mach, dev);
             }
             break;
 
@@ -2287,9 +2299,10 @@ mach_out(uint16_t addr, uint8_t val, void *priv)
 {
     mach_t          *mach = (mach_t *) priv;
     svga_t          *svga = &mach->svga;
-    const ibm8514_t *dev  = &svga->dev8514;
+    ibm8514_t 		*dev  = &svga->dev8514;
     uint8_t          old;
     uint8_t          rs2;
+    uint8_t          rs3;
 
     if (((addr & 0xFFF0) == 0x3D0 || (addr & 0xFFF0) == 0x3B0) && !(svga->miscout & 1))
         addr ^= 0x60;
@@ -2312,7 +2325,7 @@ mach_out(uint16_t addr, uint8_t val, void *priv)
                         svga_recalctimings(svga);
                     break;
                 case 0xad:
-                    if (dev->local) {
+                    if (dev->local >= 2) {
                         if ((old ^ val) & 0x0c)
                             svga_recalctimings(svga);
                     }
@@ -2328,7 +2341,7 @@ mach_out(uint16_t addr, uint8_t val, void *priv)
                     if (mach->regs[0xbe] & 0x08) { /* Read/write bank mode */
                         mach->bank_r = (((mach->regs[0xb2] & 1) << 3) | ((mach->regs[0xb2] & 0xe0) >> 5));
                         mach->bank_w = ((mach->regs[0xb2] & 0x1e) >> 1);
-                        if (dev->local) {
+                        if (dev->local >= 2) {
                             mach->bank_r |= ((mach->regs[0xae] & 0x0c) << 2);
                             mach->bank_w |= ((mach->regs[0xae] & 3) << 4);
                         }
@@ -2336,7 +2349,7 @@ mach_out(uint16_t addr, uint8_t val, void *priv)
                             mach_log("Separate B2Bank = %02x, AEbank = %02x.\n", mach->regs[0xb2], mach->regs[0xae]);
                     } else { /* Single bank mode */
                         mach->bank_w = ((mach->regs[0xb2] & 0x1e) >> 1);
-                        if (dev->local) {
+                        if (dev->local >= 2) {
                             mach->bank_w |= ((mach->regs[0xae] & 3) << 4);
                         }
                         mach->bank_r = mach->bank_w;
@@ -2360,11 +2373,11 @@ mach_out(uint16_t addr, uint8_t val, void *priv)
                     ati_eeprom_write(&mach->eeprom, val & 8, val & 2, val & 1);
                     break;
                 case 0xb6:
-                    if ((old ^ val) & 0x10)
+                    if ((old ^ val) & 0x11)
                         svga_recalctimings(svga);
                     break;
                 case 0xb8:
-                    if (dev->local) {
+                    if (dev->local >= 2) {
                         if ((old ^ val) & 0x40)
                             svga_recalctimings(svga);
                     } else {
@@ -2387,11 +2400,12 @@ mach_out(uint16_t addr, uint8_t val, void *priv)
         case 0x2ec:
         case 0x2ed:
             rs2 = !!(mach->accel.ext_ge_config & 0x1000);
-            if (dev->local) {
-                if (mach->pci_bus)
-                    ati68860_ramdac_out((addr & 3) | (rs2 << 2), val, svga->ramdac, svga);
+            rs3 = !!(mach->accel.ext_ge_config & 0x2000);
+            if (dev->local >= 2) {
+                if (mach->pci_bus && !mach->ramdac_type)
+                    ati68860_ramdac_out((addr & 3) | (rs2 << 2) | (rs3 << 3), val, svga->ramdac, svga);
                 else
-                    svga_out(addr, val, svga);
+                    ati68875_ramdac_out(addr, rs2, rs3, val, svga->ramdac, svga);
             } else
                 svga_out(addr, val, svga);
             return;
@@ -2400,12 +2414,13 @@ mach_out(uint16_t addr, uint8_t val, void *priv)
         case 0x3C7:
         case 0x3C8:
         case 0x3C9:
-            rs2 = !!(mach->accel.ext_ge_config & 0x1000);
-            if (dev->local) {
-                if (mach->pci_bus)
-                    ati68860_ramdac_out((addr & 3) | (rs2 << 2), val, svga->ramdac, svga);
+            rs2 = !!(mach->regs[0xa0] & 0x20);
+            rs3 = !!(mach->regs[0xa0] & 0x40);
+            if (dev->local >= 2) {
+                if (mach->pci_bus && !mach->ramdac_type)
+                    ati68860_ramdac_out((addr & 3) | (rs2 << 2) | (rs3 << 3), val, svga->ramdac, svga);
                 else
-                    svga_out(addr, val, svga);
+                    ati68875_ramdac_out(addr, rs2, rs3, val, svga->ramdac, svga);
             } else
                 svga_out(addr, val, svga);
             return;
@@ -2424,9 +2439,11 @@ mach_out(uint16_t addr, uint8_t val, void *priv)
             svga->crtcreg = val & 0x3f;
             return;
         case 0x3D5:
-            if ((svga->crtcreg < 7) && (svga->crtc[0x11] & 0x80))
+            if (svga->crtcreg & 0x20)
                 return;
-            if ((svga->crtcreg == 7) && (svga->crtc[0x11] & 0x80))
+            if ((svga->crtcreg < 7) && (svga->crtc[0x11] & 0x80) && !(mach->regs[0xb4] & 0x80))
+                return;
+            if ((svga->crtcreg == 7) && (svga->crtc[0x11] & 0x80) && !(mach->regs[0xb4] & 0x80))
                 val = (svga->crtc[7] & ~0x10) | (val & 0x10);
 
             old                       = svga->crtc[svga->crtcreg];
@@ -2444,6 +2461,17 @@ mach_out(uint16_t addr, uint8_t val, void *priv)
             }
             break;
 
+        case 0x46e8:
+            io_removehandler(0x03c0, 32, mach_in, NULL, NULL, mach_out, NULL, NULL, mach);
+            mem_mapping_disable(&svga->mapping);
+            mem_mapping_disable(&mach->mmio_linear_mapping);
+            if (val & 8) {
+                io_sethandler(0x03c0, 32, mach_in, NULL, NULL, mach_out, NULL, NULL, mach);
+                mem_mapping_enable(&svga->mapping);
+                mach32_updatemapping(mach);
+            }
+            break;
+
         default:
             break;
     }
@@ -2455,9 +2483,10 @@ mach_in(uint16_t addr, void *priv)
 {
     mach_t          *mach = (mach_t *) priv;
     svga_t          *svga = &mach->svga;
-    const ibm8514_t *dev  = &svga->dev8514;
-    uint8_t          temp;
+    ibm8514_t 		*dev  = &svga->dev8514;
+    uint8_t          temp = 0xff;
     uint8_t          rs2;
+    uint8_t          rs3;
 
     if (((addr & 0xFFF0) == 0x3D0 || (addr & 0xFFF0) == 0x3B0) && !(svga->miscout & 1))
         addr ^= 0x60;
@@ -2476,7 +2505,7 @@ mach_in(uint16_t addr, void *priv)
                     break;
                 case 0xb0:
                     temp = mach->regs[0xb0] | 0x80;
-                    if (dev->local) { /*Mach32 VGA 1MB memory*/
+                    if (dev->local >= 2) { /*Mach32 VGA 1MB memory*/
                         temp |= 0x08;
                         temp &= ~0x10;
                     } else { /*ATI 28800 VGA 512kB memory*/
@@ -2501,40 +2530,40 @@ mach_in(uint16_t addr, void *priv)
         case 0x2ec:
         case 0x2ed:
             rs2 = !!(mach->accel.ext_ge_config & 0x1000);
-            if (dev->local) {
-                if (mach->pci_bus)
-                    return ati68860_ramdac_in((addr & 3) | (rs2 << 2), svga->ramdac, svga);
+            rs3 = !!(mach->accel.ext_ge_config & 0x2000);
+            if (dev->local >= 2) {
+                if (mach->pci_bus && !mach->ramdac_type)
+                    temp = ati68860_ramdac_in((addr & 3) | (rs2 << 2) | (rs3 << 3), svga->ramdac, svga);
                 else
-                    return svga_in(addr, svga);
-            }
-            return svga_in(addr, svga);
+                    temp = ati68875_ramdac_in(addr, rs2, rs3, svga->ramdac, svga);
+            } else
+                temp = svga_in(addr, svga);
+            break;
 
         case 0x3C6:
         case 0x3C7:
         case 0x3C8:
         case 0x3C9:
-            rs2 = !!(mach->accel.ext_ge_config & 0x1000);
-            if (dev->local) {
-                if (mach->pci_bus)
-                    return ati68860_ramdac_in((addr & 3) | (rs2 << 2), svga->ramdac, svga);
+            rs2 = !!(mach->regs[0xa0] & 0x20);
+            rs3 = !!(mach->regs[0xa0] & 0x40);
+            if (dev->local >= 2) {
+                if (mach->pci_bus && !mach->ramdac_type)
+                    temp = ati68860_ramdac_in((addr & 3) | (rs2 << 2) | (rs3 << 3), svga->ramdac, svga);
                 else
-                    return svga_in(addr, svga);
-            }
-            return svga_in(addr, svga);
+                    temp = ati68875_ramdac_in(addr, rs2, rs3, svga->ramdac, svga);
+            } else
+                temp = svga_in(addr, svga);
+            break;
 
         case 0x3D4:
             temp = svga->crtcreg;
             break;
         case 0x3D5:
-            temp = svga->crtc[svga->crtcreg];
-            break;
-        case 0x3DA:
-            svga->attrff = 0;
-            if (svga->cgastat & 0x01)
-                svga->cgastat &= ~0x38;
+            if (svga->crtcreg & 0x20)
+                temp = 0xff;
             else
-                svga->cgastat ^= 0x38;
-            return svga->cgastat;
+                temp = svga->crtc[svga->crtcreg];
+            break;
 
         default:
             temp = svga_in(addr, svga);
@@ -2546,63 +2575,18 @@ mach_in(uint16_t addr, void *priv)
 static void
 mach_recalctimings(svga_t *svga)
 {
-    const mach_t *mach = (mach_t *) svga->priv;
+    mach_t *mach = (mach_t *) svga->priv;
     ibm8514_t    *dev  = &svga->dev8514;
+    int clock_sel;
 
-    if (vga_on && !ibm8514_on) {
-        switch (((mach->regs[0xbe] & 0x10) >> 1) | ((mach->regs[0xb9] & 2) << 1) | ((svga->miscout & 0x0c) >> 2)) {
-            case 0x00:
-                svga->clock = (cpuclock * (double) (1ULL << 32)) / 42954000.0;
-                break;
-            case 0x01:
-                svga->clock = (cpuclock * (double) (1ULL << 32)) / 48771000.0;
-                break;
-            case 0x02:
-                mach_log("clock 2\n");
-                break;
-            case 0x03:
-                svga->clock = (cpuclock * (double) (1ULL << 32)) / 36000000.0;
-                break;
-            case 0x04:
-                svga->clock = (cpuclock * (double) (1ULL << 32)) / 50350000.0;
-                break;
-            case 0x05:
-                svga->clock = (cpuclock * (double) (1ULL << 32)) / 56640000.0;
-                break;
-            case 0x06:
-                mach_log("clock 2\n");
-                break;
-            case 0x07:
-                svga->clock = (cpuclock * (double) (1ULL << 32)) / 44900000.0;
-                break;
-            case 0x08:
-                svga->clock = (cpuclock * (double) (1ULL << 32)) / 30240000.0;
-                break;
-            case 0x09:
-                svga->clock = (cpuclock * (double) (1ULL << 32)) / 32000000.0;
-                break;
-            case 0x0A:
-                svga->clock = (cpuclock * (double) (1ULL << 32)) / 37500000.0;
-                break;
-            case 0x0B:
-                svga->clock = (cpuclock * (double) (1ULL << 32)) / 39000000.0;
-                break;
-            case 0x0C:
-                svga->clock = (cpuclock * (double) (1ULL << 32)) / 50350000.0;
-                break;
-            case 0x0D:
-                svga->clock = (cpuclock * (double) (1ULL << 32)) / 56644000.0;
-                break;
-            case 0x0E:
-                svga->clock = (cpuclock * (double) (1ULL << 32)) / 75000000.0;
-                break;
-            case 0x0F:
-                svga->clock = (cpuclock * (double) (1ULL << 32)) / 65000000.0;
-                break;
+    clock_sel = ((svga->miscout >> 2) & 3) | ((mach->regs[0xbe] & 0x10) >> 1) | ((mach->regs[0xb9] & 2) << 1);
 
-            default:
-                break;
-        }
+    if (dev->local >= 2) {
+        if (mach->regs[0xad] & 0x04)
+            svga->ma_latch |= 0x40000;
+
+        if (mach->regs[0xad] & 0x08)
+            svga->ma_latch |= 0x80000;
     }
 
     if (mach->regs[0xa3] & 0x10)
@@ -2611,32 +2595,6 @@ mach_recalctimings(svga_t *svga)
     if (mach->regs[0xb0] & 0x40)
         svga->ma_latch |= 0x20000;
 
-    if (dev->local) {
-        if (mach->regs[0xad] & 0x04)
-            svga->ma_latch |= 0x40000;
-
-        if (mach->regs[0xad] & 0x08)
-            svga->ma_latch |= 0x80000;
-
-        if (mach->regs[0xb8] & 0x40)
-            svga->clock *= 2;
-    } else {
-        switch (mach->regs[0xb8] & 0xc0) {
-            case 0x40:
-                svga->clock *= 2;
-                break;
-            case 0x80:
-                svga->clock *= 3;
-                break;
-            case 0xc0:
-                svga->clock *= 4;
-                break;
-
-            default:
-                break;
-        }
-    }
-
     if (mach->regs[0xa7] & 0x80)
         svga->clock *= 3;
 
@@ -2651,137 +2609,194 @@ mach_recalctimings(svga_t *svga)
         svga->gdcreg[5] |= 0x40;
     }
 
-    if (vga_on && !ibm8514_on) {
-        if (!svga->scrblank && (svga->crtc[0x17] & 0x80) && svga->attr_palette_enable) {
-            if ((svga->gdcreg[6] & 1) || (svga->attrregs[0x10] & 1)) {
-                switch (svga->gdcreg[5] & 0x60) {
-                    case 0x00:
-                        if (svga->seqregs[1] & 8) /*Low res (320)*/
-                            svga->render = svga_render_4bpp_lowres;
-                        else
-                            svga->render = svga_render_4bpp_highres;
-                        break;
-                    case 0x20:                    /*4 colours*/
-                        if (svga->seqregs[1] & 8) /*Low res (320)*/
-                            svga->render = svga_render_2bpp_lowres;
-                        else
-                            svga->render = svga_render_2bpp_highres;
-                        break;
-                    case 0x40:
-                    case 0x60: /*256+ colours*/
-                        switch (svga->bpp) {
-                            case 8:
-                                svga->map8 = svga->pallook;
-                                if (svga->lowres)
-                                    svga->render = svga_render_8bpp_lowres;
-                                else {
-                                    svga->render = svga_render_8bpp_highres;
-                                    svga->ma_latch <<= 1;
-                                    svga->rowoffset <<= 1;
-                                }
-                                break;
+    if (!svga->scrblank && (svga->crtc[0x17] & 0x80) && svga->attr_palette_enable) {
+        if (((svga->gdcreg[6] & 1) || (svga->attrregs[0x10] & 1))) {
+            mach_log("VGA ON.\n");
+            svga->clock = (uint64_t)((cpuclock * svga->getclock(clock_sel, svga->clock_gen)) / (double) (1ull << 32));
+            switch (svga->gdcreg[5] & 0x60) {
+                case 0x00:
+                    if (svga->seqregs[1] & 8) /*Low res (320)*/
+                        svga->render = svga_render_4bpp_lowres;
+                    else
+                        svga->render = svga_render_4bpp_highres;
+                    break;
+                case 0x20:                    /*4 colours*/
+                    if (svga->seqregs[1] & 8) /*Low res (320)*/
+                        svga->render = svga_render_2bpp_lowres;
+                    else
+                        svga->render = svga_render_2bpp_highres;
+                    break;
+                case 0x40:
+                case 0x60: /*256+ colours*/
+                    switch (svga->bpp) {
+                        case 8:
+                            svga->map8 = svga->pallook;
+                            if (svga->lowres)
+                                svga->render = svga_render_8bpp_lowres;
+                            else {
+                                svga->render = svga_render_8bpp_highres;
+                                svga->ma_latch <<= 1;
+                                svga->rowoffset <<= 1;
+                            }
+                            break;
 
-                            default:
-                                break;
-                        }
-                        break;
-
-                    default:
-                        break;
-                }
+                    }
+                    break;
             }
         }
-    } else if (dev->local) {
+    }
+
+    if (dev->local >= 2) {
         if (ibm8514_on) {
-            svga->hdisp_time = svga->hdisp = (dev->hdisp + 1) << 3;
-            dev->pitch                     = (dev->accel.advfunc_cntl & 4) ? 1024 : 640;
-            svga->htotal                   = (dev->htotal + 1);
-            svga->vtotal                   = (dev->vtotal + 1);
-            svga->vsyncstart               = (dev->vsyncstart + 1);
-            svga->rowcount                 = !!(dev->disp_cntl & 0x08);
-            svga->dispend                  = ((dev->vdisp >> 1) + 1);
-            svga->interlace                = dev->interlace;
-            svga->split                    = 0xffffff;
-            svga->vblankstart              = svga->dispend;
+            svga->clock = (uint64_t)((cpuclock * svga->getclock((mach->accel.clock_sel >> 2) & 0x0f, svga->clock_gen)) / (double) (1ull << 32));
+            dev->h_disp                     = (dev->hdisp + 1) << 3;
+            dev->h_total                    = (dev->htotal + 1);
+            dev->v_total                    = (dev->vtotal + 1);
+            dev->v_syncstart                = (dev->vsyncstart + 1);
+            dev->dispend                    = ((dev->vdisp >> 1) + 1);
+            dev->rowcount                   = !!(dev->disp_cntl & 0x08);
 
-            if (svga->dispend == 766) {
-                svga->dispend     = 768;
-                svga->vblankstart = svga->dispend;
-            }
+            if (dev->dispend == 766)
+                dev->dispend += 2;
 
-            if (svga->dispend == 598) {
-                svga->dispend     = 600;
-                svga->vblankstart = svga->dispend;
-            }
+            if (dev->dispend == 598)
+                dev->dispend += 2;
 
             if (dev->accel.advfunc_cntl & 4) {
-                if (dev->ibm_mode) {
-                    if (svga->hdisp == 8) {
-                        svga->hdisp      = 1024;
-                        svga->dispend    = 768;
-                        svga->vtotal     = 1536;
-                        svga->vsyncstart = 1536;
+                if (mach->shadow_set & 2) {
+                    if (dev->h_disp == 8) {
+                        dev->h_disp = 1024;
+                        dev->dispend = 768;
+                        dev->v_total = 1536;
+                        dev->v_syncstart = 1536;
                     }
                 }
-
-                if (svga->interlace) {
-                    svga->dispend >>= 1;
-                    svga->vsyncstart >>= 2;
-                    svga->vtotal >>= 2;
-                } else {
-                    svga->vsyncstart >>= 1;
-                    svga->vtotal >>= 1;
-                }
-
-                dev->pitch      = dev->ext_pitch;
-                svga->rowoffset = dev->ext_crt_pitch;
-
-                svga->clock = (cpuclock * (double) (1ULL << 32)) / 44900000.0;
+                mach_log("1024x768 clock mode, hdisp = %d, htotal = %d, vtotal = %d, vsyncstart = %d, interlace = %02x\n", dev->h_disp, dev->h_total, dev->v_total, dev->v_syncstart, dev->interlace);
             } else {
-                if (dev->ibm_mode) {
-                    if ((svga->hdisp == 1024) && !dev->internal_pitch) {
-                        svga->hdisp   = 640;
-                        svga->dispend = 480;
+                if (mach->shadow_set & 1) {
+                    if ((dev->h_disp == 1024) && !dev->internal_pitch) {
+                        dev->h_disp = 640;
+                        dev->dispend = 480;
                     }
                 }
+            }
 
-                if (svga->interlace) {
-                    svga->dispend >>= 1;
-                    svga->vsyncstart >>= 2;
-                    svga->vtotal >>= 2;
-                } else {
-                    svga->vsyncstart >>= 1;
-                    svga->vtotal >>= 1;
+            if (dev->interlace) {
+                dev->dispend >>= 1;
+                dev->v_syncstart >>= 2;
+                dev->v_total >>= 2;
+            } else {
+                dev->v_syncstart >>= 1;
+                dev->v_total >>= 1;
+            }
+            dev->pitch = dev->ext_pitch;
+            dev->rowoffset = dev->ext_crt_pitch;
+            if ((mach->accel.ext_ge_config & 0x800) || ((!(mach->accel.ext_ge_config & 0x8000) && !(mach->accel.ext_ge_config & 0x800)))) {
+                if ((mach->accel.ext_ge_config & 0x30) == 0x20) {
+                    if ((mach->accel.ext_ge_config & 0xc0) == 0x40)
+                        svga->bpp = 16;
+                    else
+                        svga->bpp = 15;
+                } else if ((mach->accel.ext_ge_config & 0x30) == 0x30) {
+                    if (mach->accel.ext_ge_config & 0x200)
+                        svga->bpp = 32;
+                    else
+                        svga->bpp = 24;
+                } else
+                    svga->bpp = 8;
+
+                mach_log("hv(%d,%d), pitch=%d, rowoffset=%d, gextconfig=%03x.\n", dev->h_disp, dev->dispend, dev->pitch, dev->rowoffset, mach->accel.ext_ge_config & 0xcec0);
+                switch (svga->bpp) {
+                    case 8:
+                        svga->render = ibm8514_render_8bpp;
+                        break;
+                    case 15:
+                        svga->render = ibm8514_render_15bpp;
+                        break;
+                    case 16:
+                        svga->render = ibm8514_render_16bpp;
+                        break;
+                    case 24:
+                        if (mach->accel.ext_ge_config & 0x400)
+                            svga->render = ibm8514_render_BGR;
+                        else
+                            svga->render = ibm8514_render_24bpp;
+                        break;
+                    case 32:
+                        if (mach->accel.ext_ge_config & 0x400)
+                            svga->render = ibm8514_render_ABGR8888;
+                        else
+                            svga->render = ibm8514_render_RGBA8888;
+                        break;
                 }
-
-                dev->pitch      = dev->ext_pitch;
-                svga->rowoffset = dev->ext_crt_pitch;
-
-                svga->clock = (cpuclock * (double) (1ULL << 32)) / 25175000.0;
             }
-            switch (svga->bpp) {
-                default:
-                case 8:
-                    svga->render = svga_render_8bpp_highres;
+            switch (mach->regs[0xb8] & 0xc0) {
+                case 0x40:
+                    svga->clock *= 2;
                     break;
-                case 15:
-                    svga->render = svga_render_15bpp_highres;
+                case 0x80:
+                    svga->clock *= 3;
                     break;
-                case 16:
-                    svga->render = svga_render_16bpp_highres;
-                    break;
-                case 24:
-                    svga->render = svga_render_24bpp_highres;
+                case 0xc0:
+                    svga->clock *= 4;
                     break;
             }
-            mach_log("BPP=%d, VRAM Mask=%08x, NormalPitch=%d, CRTPitch=%d, VSYNCSTART=%d, VTOTAL=%d, ROWCOUNT=%d, mode=%d, highres bit=%x, has_vga?=%d, override=%d.\n", svga->bpp, svga->vram_mask, dev->pitch, dev->ext_crt_pitch, svga->vsyncstart, svga->vtotal, svga->rowcount, dev->ibm_mode, dev->accel.advfunc_cntl & 4, ibm8514_has_vga, svga->override);
         }
-        mach_log("8514 enabled, hdisp=%d, vtotal=%d, htotal=%d, dispend=%d, rowoffset=%d, split=%d, vsyncstart=%d, split=%08x\n", svga->hdisp, svga->vtotal, svga->htotal, svga->dispend, svga->rowoffset, svga->split, svga->vsyncstart, svga->split);
+    } else {
+        if (ibm8514_on) {
+            svga->clock = (uint64_t)((cpuclock * svga->getclock((mach->accel.clock_sel >> 2) & 0x0f, svga->clock_gen)) / (double) (1ull << 32));
+            dev->h_disp                     = (dev->hdisp + 1) << 3;
+            dev->h_total                    = (dev->htotal + 1);
+            dev->v_total                    = (dev->vtotal + 1);
+            dev->v_syncstart                = (dev->vsyncstart + 1);
+            dev->rowcount                  = !!(dev->disp_cntl & 0x08);
+            dev->dispend                   = ((dev->vdisp >> 1) + 1);
+
+            if (dev->dispend == 766)
+                dev->dispend += 2;
+
+            if (dev->dispend == 598)
+                dev->dispend += 2;
+
+            if (dev->accel.advfunc_cntl & 4) {
+                if (mach->shadow_set & 2) {
+                    if (dev->h_disp == 8) {
+                        dev->h_disp = 1024;
+                        dev->dispend = 768;
+                        dev->v_total = 1536;
+                        dev->v_syncstart = 1536;
+                    }
+                }
+            } else {
+                if (mach->shadow_set & 1) {
+                    if ((dev->h_disp == 1024) && !dev->internal_pitch) {
+                        dev->h_disp = 640;
+                        dev->dispend = 480;
+                    }
+                }
+            }
+            if (dev->interlace) {
+                dev->dispend >>= 1;
+                dev->v_syncstart >>= 2;
+                dev->v_total >>= 2;
+            } else {
+                dev->v_syncstart >>= 1;
+                dev->v_total >>= 1;
+            }
+
+            dev->pitch = dev->ext_pitch;
+            dev->rowoffset = dev->ext_crt_pitch;
+            mach_log("hv(%d,%d), pitch=%d, rowoffset=%d, gextconfig=%03x.\n", dev->h_disp, dev->dispend, dev->pitch, dev->rowoffset, mach->accel.ext_ge_config & 0xcec0);
+            svga->map8 = dev->pallook;
+            svga->render = ibm8514_render_8bpp;
+            if (mach->regs[0xb8] & 0x40)
+                svga->clock *= 2;
+        }
     }
 }
 
 static void
-mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, uint32_t val, int len)
+mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, uint16_t val, int len)
 {
     int frgd_sel;
     int bkgd_sel;
@@ -2914,6 +2929,7 @@ mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, u
                         dev->accel.cmd_back = 0;
                 }
                 ibm8514_accel_start(-1, 0, -1, 0, svga, len);
+                svga_recalctimings(svga);
             }
             break;
         case 0x9ae9:
@@ -3035,23 +3051,23 @@ mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, u
                                 case 0x000: /*8-bit size*/
                                     if (mono_src == 2) {
                                         if ((frgd_sel != 2) && (bkgd_sel != 2)) {
-                                            mach_accel_start(mach->accel.cmd_type, 1, 8, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), 0, mach, dev, len);
+                                            mach_accel_start(mach->accel.cmd_type, 1, 8, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), 0, mach, dev);
                                         } else
-                                            mach_accel_start(mach->accel.cmd_type, 1, 1, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev, len);
+                                            mach_accel_start(mach->accel.cmd_type, 1, 1, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev);
                                     } else
-                                        mach_accel_start(mach->accel.cmd_type, 1, 1, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev, len);
+                                        mach_accel_start(mach->accel.cmd_type, 1, 1, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev);
                                     break;
                                 case 0x200: /*16-bit size*/
                                     if (mono_src == 2) {
                                         if ((frgd_sel != 2) && (bkgd_sel != 2)) {
                                             if (mach->accel.dp_config & 0x1000)
-                                                mach_accel_start(mach->accel.cmd_type, 1, 16, mach->accel.pix_trans[1] | (mach->accel.pix_trans[0] << 8), 0, mach, dev, len);
+                                                mach_accel_start(mach->accel.cmd_type, 1, 16, mach->accel.pix_trans[1] | (mach->accel.pix_trans[0] << 8), 0, mach, dev);
                                             else
-                                                mach_accel_start(mach->accel.cmd_type, 1, 16, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), 0, mach, dev, len);
+                                                mach_accel_start(mach->accel.cmd_type, 1, 16, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), 0, mach, dev);
                                         } else
-                                            mach_accel_start(mach->accel.cmd_type, 1, 2, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev, len);
+                                            mach_accel_start(mach->accel.cmd_type, 1, 2, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev);
                                     } else
-                                        mach_accel_start(mach->accel.cmd_type, 1, 2, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev, len);
+                                        mach_accel_start(mach->accel.cmd_type, 1, 2, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev);
                                     break;
 
                                 default:
@@ -3120,23 +3136,23 @@ mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, u
                                 case 0x000: /*8-bit size*/
                                     if (mono_src == 2) {
                                         if ((frgd_sel != 2) && (bkgd_sel != 2)) {
-                                            mach_accel_start(mach->accel.cmd_type, 1, 8, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), 0, mach, dev, len);
+                                            mach_accel_start(mach->accel.cmd_type, 1, 8, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), 0, mach, dev);
                                         } else
-                                            mach_accel_start(mach->accel.cmd_type, 1, 1, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev, len);
+                                            mach_accel_start(mach->accel.cmd_type, 1, 1, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev);
                                     } else
-                                        mach_accel_start(mach->accel.cmd_type, 1, 1, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev, len);
+                                        mach_accel_start(mach->accel.cmd_type, 1, 1, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev);
                                     break;
                                 case 0x200: /*16-bit size*/
                                     if (mono_src == 2) {
                                         if ((frgd_sel != 2) && (bkgd_sel != 2)) {
                                             if (mach->accel.dp_config & 0x1000)
-                                                mach_accel_start(mach->accel.cmd_type, 1, 16, mach->accel.pix_trans[1] | (mach->accel.pix_trans[0] << 8), 0, mach, dev, len);
+                                                mach_accel_start(mach->accel.cmd_type, 1, 16, mach->accel.pix_trans[1] | (mach->accel.pix_trans[0] << 8), 0, mach, dev);
                                             else
-                                                mach_accel_start(mach->accel.cmd_type, 1, 16, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), 0, mach, dev, len);
+                                                mach_accel_start(mach->accel.cmd_type, 1, 16, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), 0, mach, dev);
                                         } else
-                                            mach_accel_start(mach->accel.cmd_type, 1, 2, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev, len);
+                                            mach_accel_start(mach->accel.cmd_type, 1, 2, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev);
                                     } else
-                                        mach_accel_start(mach->accel.cmd_type, 1, 2, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev, len);
+                                        mach_accel_start(mach->accel.cmd_type, 1, 2, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev);
                                     break;
 
                                 default:
@@ -3221,7 +3237,7 @@ mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, u
                 }
                 mach_log("CLIPBOTTOM=%d, CLIPRIGHT=%d, bpp=%d, pitch=%d.\n", dev->accel.multifunc[3], dev->accel.multifunc[4], svga->bpp, dev->pitch);
                 if ((dev->accel.multifunc_cntl >> 12) == 5) {
-                    if (!dev->local || !dev->ext_crt_pitch)
+                    if (!dev->ext_crt_pitch || ((dev->local < 2)))
                         dev->ext_crt_pitch = 128;
                     svga_recalctimings(svga);
                 }
@@ -3243,7 +3259,7 @@ mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, u
                     dev->accel.clip_left = dev->accel.multifunc_cntl & 0x7ff;
                 }
                 if ((dev->accel.multifunc_cntl >> 12) == 5) {
-                    if (!dev->local || !dev->ext_crt_pitch)
+                    if (!dev->ext_crt_pitch || ((dev->local < 2)))
                         dev->ext_crt_pitch = 128;
                     svga_recalctimings(svga);
                 }
@@ -3309,7 +3325,7 @@ mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, u
                 dev->data_available  = 0;
                 dev->data_available2 = 0;
                 mach->accel.cmd_type = 1;
-                mach_accel_start(mach->accel.cmd_type, 0, -1, -1, 0, mach, dev, len);
+                mach_accel_start(mach->accel.cmd_type, 0, -1, -1, 0, mach, dev);
             }
             break;
         case 0x96ef:
@@ -3319,7 +3335,7 @@ mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, u
                 dev->data_available  = 0;
                 dev->data_available2 = 0;
                 mach->accel.cmd_type = 1;
-                mach_accel_start(mach->accel.cmd_type, 0, -1, -1, 0, mach, dev, len);
+                mach_accel_start(mach->accel.cmd_type, 0, -1, -1, 0, mach, dev);
             }
             break;
 
@@ -3378,7 +3394,7 @@ mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, u
                 dev->data_available2 = 0;
                 mach_log("BitBLT = %04x.\n", mach->accel.dp_config);
                 mach->accel.cmd_type = 2; /*Non-conforming BitBLT from dest_y_end register (0xaeee)*/
-                mach_accel_start(mach->accel.cmd_type, 0, -1, -1, 0, mach, dev, len);
+                mach_accel_start(mach->accel.cmd_type, 0, -1, -1, 0, mach, dev);
             }
             break;
         case 0xaeef:
@@ -3387,7 +3403,7 @@ mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, u
                 dev->data_available    = 0;
                 dev->data_available2   = 0;
                 mach->accel.cmd_type   = 2; /*Non-conforming BitBLT from dest_y_end register (0xaeee)*/
-                mach_accel_start(mach->accel.cmd_type, 0, -1, -1, 0, mach, dev, len);
+                mach_accel_start(mach->accel.cmd_type, 0, -1, -1, 0, mach, dev);
             }
             break;
 
@@ -3445,7 +3461,7 @@ mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, u
                 dev->data_available2 = 0;
                 mach->accel.cmd_type = 5; /*Horizontal Raster Draw from scan_to_x register (0xcaee)*/
                 mach_log("ScanToX = %04x.\n", mach->accel.dp_config);
-                mach_accel_start(mach->accel.cmd_type, 0, -1, -1, 0, mach, dev, len);
+                mach_accel_start(mach->accel.cmd_type, 0, -1, -1, 0, mach, dev);
             }
             break;
         case 0xcaef:
@@ -3454,7 +3470,7 @@ mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, u
                 dev->data_available   = 0;
                 dev->data_available2  = 0;
                 mach->accel.cmd_type  = 5; /*Horizontal Raster Draw from scan_to_x register (0xcaee)*/
-                mach_accel_start(mach->accel.cmd_type, 0, -1, -1, 0, mach, dev, len);
+                mach_accel_start(mach->accel.cmd_type, 0, -1, -1, 0, mach, dev);
             }
             break;
 
@@ -3578,7 +3594,7 @@ mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, u
                 mach->accel.cy_end_line                      = mach->accel.line_array[3];
                 if ((mach->accel.line_idx == 3) || (mach->accel.line_idx == 5)) {
                     mach->accel.cmd_type = (mach->accel.line_idx == 5) ? 4 : 3;
-                    mach_accel_start(mach->accel.cmd_type, 0, -1, -1, 0, mach, dev, len);
+                    mach_accel_start(mach->accel.cmd_type, 0, -1, -1, 0, mach, dev);
                     mach->accel.line_idx = (mach->accel.line_idx == 5) ? 4 : 2;
                     break;
                 }
@@ -3592,672 +3608,344 @@ mach_accel_out_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, u
 }
 
 static void
-mach_accel_out(uint16_t port, uint32_t val, mach_t *mach, int len)
+mach_accel_out(uint16_t port, uint8_t val, mach_t *mach)
 {
     svga_t    *svga = &mach->svga;
     ibm8514_t *dev  = &svga->dev8514;
 
-    mach_log("Port accel out = %04x, val = %04x, len = %d.\n", port, val, len);
+    mach_log("Port accel out = %04x, val = %04x.\n", port, val);
 
-    if (port & 0x8000) {
-        mach_accel_out_fifo(mach, svga, dev, port, val, len);
-    } else {
-        switch (port) {
-            case 0x2e8:
-                if (len == 1)
-                    dev->htotal = (dev->htotal & 0xff00) | val;
-                else {
-                    dev->htotal = val;
-                    svga_recalctimings(svga);
+    switch (port) {
+        case 0x2e8:
+        case 0x2e9:
+            WRITE8(port, dev->htotal, val);
+            svga_recalctimings(svga);
+            break;
+        case 0x6e8:
+            dev->hdisp = val;
+            mach_log("ATI 8514/A: H_DISP write 06E8 = %d\n", dev->hdisp + 1);
+            svga_recalctimings(svga);
+            break;
+
+        case 0xae8:
+            mach_log("ATI 8514/A: H_SYNC_STRT write 0AE8 = %d\n", val + 1);
+            svga_recalctimings(svga);
+            break;
+
+        case 0xee8:
+            mach_log("ATI 8514/A: H_SYNC_WID write 0EE8 = %d\n", val + 1);
+            svga_recalctimings(svga);
+            break;
+
+        case 0x12e8:
+        case 0x12e9:
+            WRITE8(port, dev->vtotal, val);
+            dev->vtotal &= 0x1fff;
+            svga_recalctimings(svga);
+            break;
+
+        case 0x16e8:
+        case 0x16e9:
+            WRITE8(port, dev->vdisp, val);
+            dev->vdisp &= 0x1fff;
+            svga_recalctimings(svga);
+            break;
+
+        case 0x1ae8:
+        case 0x1ae9:
+            WRITE8(port, dev->vsyncstart, val);
+            dev->vsyncstart &= 0x1fff;
+            svga_recalctimings(svga);
+            break;
+
+        case 0x1ee8:
+            dev->vsyncwidth = val;
+            mach_log("ATI 8514/A: V_SYNC_WID write 1EE8 = %02x\n", val);
+            svga_recalctimings(svga);
+            break;
+
+        case 0x22e8:
+            dev->disp_cntl = val & 0x7e;
+            dev->interlace = !!(val & 0x10);
+            mach_log("ATI 8514/A: DISP_CNTL write 22E8 = %02x, SCANMODULOS = %d\n", dev->disp_cntl, dev->scanmodulos);
+            svga_recalctimings(svga);
+            break;
+
+        case 0x42e8:
+            dev->subsys_stat &= ~val;
+            break;
+        case 0x42e9:
+            dev->subsys_cntl = val;
+            break;
+
+        case 0x4ae8:
+            if (!val)
+                break;
+            if (!dev->ext_crt_pitch || ((dev->local < 2)))
+                dev->ext_crt_pitch = 128;
+
+            dev->accel.advfunc_cntl = val & 0x0f;
+            ibm8514_on = val & 0x01;
+            if (!ibm8514_on && mach->old_mode) {
+                ibm8514_on = 1;
+                mach->old_mode = 0;
+            }
+            vga_on = !ibm8514_on;
+            mach_log("ATI 8514/A: VGA ON (0x4ae8) = %i, val = %02x, 8514 = %i, gdcreg6 = %02x\n", vga_on, val, ibm8514_on, svga->gdcreg[6] & 0x0c);
+            svga_recalctimings(svga);
+            break;
+
+        /*ATI Mach8/32 specific registers*/
+        case 0x2ee:
+            mach_log("2EE write val = %02x.\n", val);
+            break;
+        case 0x2ef:
+            mach_log("2EF write val = %02x.\n", val);
+            break;
+
+        case 0x6ee:
+            mach_log("6EE write val = %02x.\n", val);
+            break;
+        case 0x6ef:
+            mach_log("6EF write val = %02x.\n", val);
+            break;
+
+        case 0xaee:
+        case 0xaef:
+            WRITE8(port, mach->cursor_offset_lo, val);
+            break;
+
+        case 0xeee:
+        case 0xeef:
+            WRITE8(port, mach->cursor_offset_hi_reg, val);
+            mach->cursor_offset_hi = mach->cursor_offset_hi_reg & 0x0f;
+            dev->hwcursor.addr = ((mach->cursor_offset_lo | (mach->cursor_offset_hi << 16))) << 2;
+            dev->hwcursor.ena = !!(mach->cursor_offset_hi_reg & 0x8000);
+            break;
+
+        case 0x12ee:
+        case 0x12ef:
+            WRITE8(port, mach->cursor_x, val);
+            dev->hwcursor.x = mach->cursor_x & 0x7ff;
+            break;
+
+        case 0x16ee:
+        case 0x16ef:
+            WRITE8(port, mach->cursor_y, val);
+            dev->hwcursor.y = mach->cursor_y & 0xfff;
+            break;
+
+        case 0x1aee:
+        case 0x1aef:
+            WRITE8(port, mach->cursor_col_b, val);
+            mach->cursor_col_0 = mach->cursor_col_b & 0xff;
+            mach->cursor_col_1 = (mach->cursor_col_b >> 8) & 0xff;
+            break;
+
+        case 0x1eee:
+        case 0x1eef:
+            WRITE8(port, mach->cursor_vh_offset, val);
+            dev->hwcursor.xoff = mach->cursor_vh_offset & 0x3f;
+            dev->hwcursor.yoff = (mach->cursor_vh_offset >> 8) & 0x3f;
+            break;
+
+        case 0x22ee:
+            if (mach->pci_bus) {
+                mach->pci_cntl_reg = val;
+                mach32_updatemapping(mach);
+            }
+            break;
+
+        case 0x26ee:
+            mach_log("CRT Pitch = %d, original val = %d.\n", val << 3, val);
+            dev->ext_crt_pitch = val;
+            dev->internal_pitch = val;
+            if (svga->bpp > 8) {
+                if (svga->bpp == 24)
+                    dev->ext_crt_pitch *= 3;
+                else if (svga->bpp == 32)
+                    dev->ext_crt_pitch <<= 2;
+                else
+                    dev->ext_crt_pitch <<= 1;
+            }
+            if (dev->local >= 2) {
+                if (!ibm8514_on) {
+                    ibm8514_on = 1;
+                    svga->adv_flags |= FLAG_ATI;
                 }
-                break;
-            case 0x2e9:
-                if (len != 1) {
-                    dev->htotal = (dev->htotal & 0xff) | (val << 8);
-                    mach_log("ATI 8514/A: H_TOTAL write 02E8 = %d\n", dev->htotal + 1);
-                    svga_recalctimings(svga);
-                }
-                break;
+            }
+            svga_recalctimings(svga);
+            break;
 
-            case 0x6e8:
-                dev->hdisp = val;
-                mach_log("ATI 8514/A: H_DISP write 06E8 = %d\n", dev->hdisp + 1);
-                svga_recalctimings(svga);
-                break;
+        case 0x32ee:
+        case 0x32ef:
+            WRITE8(port, mach->local_cntl, val);
+            mach32_updatemapping(mach);
+            break;
 
-            case 0xae8:
-                mach_log("ATI 8514/A: H_SYNC_STRT write 0AE8 = %d\n", val + 1);
-                svga_recalctimings(svga);
-                break;
+        case 0x36ee:
+        case 0x36ef:
+            WRITE8(port, mach->misc, val);
+            mach->misc &= 0xfff0;
+            break;
 
-            case 0xee8:
-                mach_log("ATI 8514/A: H_SYNC_WID write 0EE8 = %d\n", val + 1);
-                svga_recalctimings(svga);
-                break;
+        case 0x3aee:
+        case 0x3aef:
+            WRITE8(port, mach->cursor_col_0_rg, val);
+            mach->ext_cur_col_0_g = mach->cursor_col_0_rg & 0xff;
+            mach->ext_cur_col_0_r = (mach->cursor_col_0_rg >> 8) & 0xff;
+            break;
 
-            case 0x12e8:
-                if (len == 1)
-                    dev->vtotal = (dev->vtotal & 0x1f00) | val;
-                else {
-                    dev->vtotal = val & 0x1fff;
-                    svga_recalctimings(svga);
-                }
-                break;
-            case 0x12e9:
-                if (len == 1) {
-                    dev->vtotal = (dev->vtotal & 0xff) | ((val & 0x1f) << 8);
-                    mach_log("ATI 8514/A: V_TOTAL write 12E8 = %d\n", dev->vtotal);
-                    svga_recalctimings(svga);
-                }
-                break;
+        case 0x3eee:
+        case 0x3eef:
+            WRITE8(port, mach->cursor_col_1_rg, val);
+            mach->ext_cur_col_1_g = mach->cursor_col_1_rg & 0xff;
+            mach->ext_cur_col_1_r = (mach->cursor_col_1_rg >> 8) & 0xff;
+            break;
 
-            case 0x16e8:
-                if (len == 1)
-                    dev->vdisp = (dev->vdisp & 0x1f00) | val;
-                else {
-                    dev->vdisp = val & 0x1fff;
-                    svga_recalctimings(svga);
-                }
-                break;
-            case 0x16e9:
-                if (len == 1) {
-                    dev->vdisp = (dev->vdisp & 0xff) | ((val & 0x1f) << 8);
-                    mach_log("ATI 8514/A: V_DISP write 16E8 = %d\n", dev->vdisp);
-                    svga_recalctimings(svga);
-                }
-                break;
+        case 0x42ee:
+            mach->accel.test2[0] = val;
+            break;
+        case 0x42ef:
+            mach->accel.test2[1] = val;
+            break;
 
-            case 0x1ae8:
-                if (len == 1)
-                    dev->vsyncstart = (dev->vsyncstart & 0x1f00) | val;
-                else {
-                    dev->vsyncstart = val & 0x1fff;
-                    svga_recalctimings(svga);
-                }
-                break;
-            case 0x1ae9:
-                if (len == 1) {
-                    dev->vsyncstart = (dev->vsyncstart & 0xff) | ((val & 0x1f) << 8);
-                    mach_log("ATI 8514/A: V_SYNC_STRT write 1AE8 = %d\n", dev->vsyncstart);
-                    svga_recalctimings(svga);
-                }
-                break;
+        case 0x46ee:
+            mach->accel.test3[0] = val;
+            break;
+        case 0x46ef:
+            mach->accel.test3[1] = val;
+            break;
 
-            case 0x1ee8:
-                dev->vsyncwidth = val;
-                mach_log("ATI 8514/A: V_SYNC_WID write 1EE8 = %02x\n", val);
-                svga_recalctimings(svga);
-                break;
+        case 0x4aee:
+        case 0x4aef:
+        	if (!(port & 1))
+                mach->old_mode = mach->accel.clock_sel & 0x01;
 
-            case 0x22e8:
-                dev->disp_cntl = val & 0x7e;
-                dev->interlace = !!(val & 0x10);
-                mach_log("ATI 8514/A: DISP_CNTL write 22E8 = %02x, SCANMODULOS = %d\n", dev->disp_cntl, dev->scanmodulos);
-                svga_recalctimings(svga);
-                break;
-
-            case 0x42e8:
-                if (len == 1) {
-                    dev->subsys_stat &= ~val;
-                } else {
-                    dev->subsys_stat &= ~(val & 0xff);
-                    dev->subsys_cntl = (val >> 8);
-                    mach_log("CNTL = %02x.\n", val >> 8);
-                }
-                break;
-            case 0x42e9:
-                if (len == 1) {
-                    dev->subsys_cntl = val;
-                    mach_log("CNTL = %02x.\n", val);
-                }
-                break;
-
-            case 0x4ae8:
-                mach_log("ATI 8514/A: VGA ON (0x4ae8) = %i, val = %02x\n", vga_on, val);
-                if (!val)
-                    break;
-                if (!dev->local || !dev->ext_crt_pitch)
-                    dev->ext_crt_pitch = 128;
-                dev->accel.advfunc_cntl = val & 7;
-                ibm8514_on              = (dev->accel.advfunc_cntl & 1);
-                vga_on                  = !ibm8514_on;
-                dev->ibm_mode           = 1;
+            WRITE8(port, mach->accel.clock_sel, val);
+            ibm8514_on = mach->accel.clock_sel & 0x01;
+            vga_on = !ibm8514_on;
+            if (dev->local >= 2) {
                 if (ibm8514_on)
                     svga->adv_flags |= FLAG_ATI;
                 else
                     svga->adv_flags &= ~FLAG_ATI;
-                svga_recalctimings(svga);
-                break;
+            }
+            mach_log("ATI 8514/A: VGA ON (0x4aee) = %i, Ext = %i, val = %04x\n", vga_on, ibm8514_on, mach->accel.clock_sel);
+            svga_recalctimings(svga);
+            break;
 
-            /*ATI Mach8/32 specific registers*/
-            case 0x6ee:
-                mach_log("6EE write val = %02x, len = %d.\n", val, len);
-                break;
+        case 0x52ee:
+        case 0x52ef:
+            WRITE8(port, mach->accel.scratch0, val);
+            break;
 
-            case 0x6ef:
-                mach_log("6EF write val = %02x, len = %d.\n", val, len);
-                break;
+        case 0x56ee:
+        case 0x56ef:
+            WRITE8(port, mach->accel.scratch1, val);
+            break;
 
-            case 0xaee:
-                if (len == 1)
-                    mach->cursor_offset_lo = (mach->cursor_offset_lo & 0xff00) | val;
-                else {
-                    mach_log("AEE val=%02x.\n", val);
-                    mach->cursor_offset_lo = val;
-                    svga->hwcursor.addr    = mach->cursor_offset_lo << 2;
-                }
-                break;
-            case 0xaef:
-                if (len == 1) {
-                    mach->cursor_offset_lo = (mach->cursor_offset_lo & 0x00ff) | (val << 8);
-                    svga->hwcursor.addr    = mach->cursor_offset_lo << 2;
-                }
-                break;
+        case 0x5aee:
+        case 0x5aef:
+            WRITE8(port, mach->shadow_set, val);
+            svga_recalctimings(svga);
+            break;
 
-            case 0xeee:
-                mach->cursor_offset_hi = val & 0x0f;
-                if (len != 1) {
-                    svga->hwcursor.addr = ((mach->cursor_offset_lo | (mach->cursor_offset_hi << 16))) << 2;
-                    svga->hwcursor.ena  = !!(val & 0x8000);
-                }
-                mach_log("EEE val=%08x.\n", svga->hwcursor.addr);
-                break;
-            case 0xeef:
-                if (len == 1) {
-                    svga->hwcursor.addr = ((mach->cursor_offset_lo | (mach->cursor_offset_hi << 16))) << 2;
-                    svga->hwcursor.ena  = !!(val & 0x80);
-                }
-                break;
+        case 0x5eee:
+        case 0x5eef:
+            WRITE8(port, mach->memory_aperture, val);
+            mach_log("Memory Aperture = %04x.\n", mach->memory_aperture);
+            if (!mach->pci_bus)
+                mach->linear_base = (mach->memory_aperture & 0xff00) << 12;
 
-            case 0x12ee:
-                if (len == 1) {
-                    svga->hwcursor.x = (svga->hwcursor.x & 0x700) | val;
-                } else {
-                    svga->hwcursor.x = val & 0x7ff;
-                    mach_log("X = %03x.\n", val);
-                }
-                break;
-            case 0x12ef:
-                if (len == 1) {
-                    svga->hwcursor.x = (svga->hwcursor.x & 0x0ff) | ((val & 0x07) << 8);
-                }
-                break;
+            mach32_updatemapping(mach);
+            break;
 
-            case 0x16ee:
-                if (len == 1) {
-                    svga->hwcursor.y = (svga->hwcursor.y & 0xf00) | val;
-                } else {
-                    svga->hwcursor.y = val & 0xfff;
-                }
-                break;
-            case 0x16ef:
-                if (len == 1) {
-                    svga->hwcursor.y = (svga->hwcursor.y & 0x0ff) | ((val & 0x0f) << 8);
-                }
-                break;
+        case 0x62ee:
+            mach_log("62EE write val = %04x, len = %d.\n", val, len);
+            break;
 
-            case 0x1aee:
-                if (len != 1) {
-                    mach->cursor_col_0 = val & 0xff;
-                    mach->cursor_col_1 = (val >> 8) & 0xff;
-                } else
-                    mach->cursor_col_0 = val;
-                break;
-            case 0x1aef:
-                if (len == 1)
-                    mach->cursor_col_1 = val;
-                break;
+        case 0x66ee:
+            mach_log("66EE write val = %04x, len = %d.\n", val, len);
+            break;
 
-            case 0x1eee:
-                if (len != 1) {
-                    svga->hwcursor.xoff = val & 0x3f;
-                    svga->hwcursor.yoff = (val >> 8) & 0x3f;
-                } else
-                    svga->hwcursor.xoff = val & 0x3f;
-                break;
-            case 0x1eef:
-                if (len == 1)
-                    svga->hwcursor.yoff = val & 0x3f;
-                break;
+        case 0x6aee:
+        case 0x6aef:
+            WRITE8(port, mach->accel.max_waitstates, val);
+            break;
 
-            case 0x2aee:
-                mach_log("2AEE write val = %04x\n", val);
-                if (len == 1)
-                    mach->accel.crt_offset_lo = (mach->accel.crt_offset_lo & 0xff00) | val;
-                else
-                    mach->accel.crt_offset_lo = val;
-                break;
-            case 0x2aef:
-                if (len == 1)
-                    mach->accel.crt_offset_lo = (mach->accel.crt_offset_lo & 0x00ff) | (val << 8);
-                break;
+        case 0x6eee:
+        case 0x6eef:
+            WRITE8(port, mach->accel.ge_offset_lo, val);
+            dev->accel.ge_offset = mach->accel.ge_offset_lo;
+            break;
 
-            case 0x2eee:
-                mach_log("2EEE write val = %04x\n", val);
-                if (len == 1)
-                    mach->accel.crt_offset_hi = (mach->accel.crt_offset_hi & 0xff00) | val;
-                else
-                    mach->accel.crt_offset_hi = val;
-                break;
-            case 0x2eef:
-                if (len == 1)
-                    mach->accel.crt_offset_hi = (mach->accel.crt_offset_hi & 0x00ff) | (val << 8);
-                break;
+        case 0x72ee:
+        case 0x72ef:
+            WRITE8(port, mach->accel.ge_offset_hi, val);
+            dev->accel.ge_offset = mach->accel.ge_offset_lo | (mach->accel.ge_offset_hi << 16);
+            break;
 
-            case 0x26ee:
-                mach_log("CRT Pitch = %d, original val = %d.\n", val << 3, val);
-                dev->ext_crt_pitch  = val;
-                dev->internal_pitch = val;
-                if (svga->bpp > 8) {
-                    if (svga->bpp == 24)
-                        dev->ext_crt_pitch *= 3;
-                    else
+        case 0x76ee:
+            mach_log("76EE write val=%d shifted, normal=%d.\n", val << 3, val);
+            dev->ext_pitch = val << 3;
+            mach->old_mode = 1;
+            svga_recalctimings(svga);
+            break;
+
+        case 0x7aee:
+        case 0x7aef:
+            WRITE8(port, mach->accel.ext_ge_config, val);
+            if (dev->local >= 2) {
+                dev->ext_crt_pitch = dev->internal_pitch;
+                switch (mach->accel.ext_ge_config & 0x30) {
+                    case 0:
+                    case 0x10:
+                        dev->bpp = 0;
+                        break;
+                    case 0x20:
+                        dev->bpp = 1;
                         dev->ext_crt_pitch <<= 1;
-                }
-                if (dev->local) {
-                    if (!ibm8514_on) {
-                        ibm8514_on ^= 1;
-                        svga->adv_flags |= FLAG_ATI;
-                    }
-                }
-                svga_recalctimings(svga);
-                break;
-
-            case 0x32ee:
-                if (len == 1) {
-                    mach->local_cntl = (mach->local_cntl & 0xff00) | val;
-                } else {
-                    mach->local_cntl = val;
-                    mach32_updatemapping(mach);
-                }
-                break;
-
-            case 0x32ef:
-                if (len == 1) {
-                    mach->local_cntl = (mach->local_cntl & 0x00ff) | (val << 8);
-                    mach32_updatemapping(mach);
-                }
-                break;
-
-            case 0x36ee:
-                if (len == 1) {
-                    mach->misc = (mach->misc & 0xff00) | (val);
-                } else {
-                    mach->misc = val;
-                }
-                break;
-            case 0x36ef:
-                if (len == 1) {
-                    mach->misc = (mach->misc & 0x00ff) | (val << 8);
-                }
-                break;
-
-            case 0x3aee:
-                if (len == 1) {
-                    mach->ext_cur_col_0_g = val;
-                } else {
-                    mach->ext_cur_col_0_g = val & 0xff;
-                    mach->ext_cur_col_0_r = (val >> 8) & 0xff;
-                }
-                break;
-            case 0x3aef:
-                if (len == 1) {
-                    mach->ext_cur_col_0_r = val;
-                }
-                break;
-
-            case 0x3eee:
-                if (len == 1) {
-                    mach->ext_cur_col_1_g = val;
-                } else {
-                    mach->ext_cur_col_1_g = val & 0xff;
-                    mach->ext_cur_col_1_r = (val >> 8) & 0xff;
-                }
-                break;
-            case 0x3eef:
-                if (len == 1) {
-                    mach->ext_cur_col_1_r = val;
-                }
-                break;
-
-            case 0x42ee:
-                mach->accel.test2[0] = val;
-                break;
-            case 0x42ef:
-                mach->accel.test2[1] = val;
-                break;
-
-            case 0x46ee:
-                mach->accel.test3[0] = val;
-                break;
-            case 0x46ef:
-                mach->accel.test3[1] = val;
-                break;
-
-            case 0x4aee:
-                if (len == 1)
-                    mach->accel.clock_sel = (mach->accel.clock_sel & 0xff00) | val;
-                else {
-                    mach->accel.clock_sel = val;
-                    ibm8514_on            = (mach->accel.clock_sel & 1);
-                    vga_on                = !ibm8514_on;
-                    dev->ibm_mode         = 0;
-                    if (ibm8514_on)
-                        svga->adv_flags |= FLAG_ATI;
-                    else
-                        svga->adv_flags &= ~FLAG_ATI;
-                    mach_log("ATI 8514/A: VGA ON (0x4aee) = %i, val = %04x\n", vga_on, val);
-                    svga_recalctimings(svga);
-                }
-                break;
-            case 0x4aef:
-                if (len == 1) {
-                    mach->accel.clock_sel = (mach->accel.clock_sel & 0x00ff) | (val << 8);
-                    ibm8514_on            = (mach->accel.clock_sel & 1);
-                    vga_on                = !ibm8514_on;
-                    dev->ibm_mode         = 0;
-                    if (ibm8514_on)
-                        svga->adv_flags |= FLAG_ATI;
-                    else
-                        svga->adv_flags &= ~FLAG_ATI;
-                    mach_log("ATI 8514/A: VGA ON (0x4aef) = %i, val = %04x\n", vga_on, mach->accel.clock_sel);
-                    svga_recalctimings(svga);
-                }
-                break;
-
-            case 0x52ee:
-                if (len == 1)
-                    mach->accel.scratch0 = (mach->accel.scratch0 & 0xff00) | val;
-                else
-                    mach->accel.scratch0 = val;
-                break;
-            case 0x52ef:
-                if (len == 1)
-                    mach->accel.scratch0 = (mach->accel.scratch0 & 0x00ff) | (val << 8);
-                break;
-
-            case 0x56ee:
-                if (len == 1)
-                    mach->accel.scratch1 = (mach->accel.scratch1 & 0xff00) | val;
-                else
-                    mach->accel.scratch1 = val;
-                break;
-            case 0x56ef:
-                if (len == 1)
-                    mach->accel.scratch1 = (mach->accel.scratch1 & 0x00ff) | (val << 8);
-                break;
-
-            case 0x5aee:
-                mach_log("Shadow set = %04x\n", val);
-                break;
-            case 0x5aef:
-                mach_log("Shadow + 1 set = %02x\n", val);
-                break;
-
-            case 0x5eee:
-                mach_log("Memory Aperture = %04x, len = %d.\n", val, len);
-                if (len == 1) {
-                    mach->memory_aperture = (mach->memory_aperture & 0xff00) | val;
-                } else {
-                    mach->memory_aperture = val;
-                    if (!mach->pci_bus)
-                        mach->linear_base = (mach->memory_aperture & 0xff00) << 12;
-
-                    mach32_updatemapping(mach);
-                }
-                break;
-
-            case 0x5eef:
-                if (len == 1) {
-                    mach->memory_aperture = (mach->memory_aperture & 0x00ff) | (val << 8);
-                    if (!mach->pci_bus)
-                        mach->linear_base = (mach->memory_aperture & 0xff00) << 12;
-
-                    mach32_updatemapping(mach);
-                }
-                break;
-
-            case 0x62ee:
-                mach_log("62EE write val = %04x, len = %d.\n", val, len);
-                break;
-
-            case 0x66ee:
-                mach_log("66EE write val = %04x, len = %d.\n", val, len);
-                break;
-
-            case 0x6aee:
-                mach_log("6AEE write val = %04x.\n", val & 0x400);
-                if (len == 1)
-                    mach->accel.max_waitstates = (mach->accel.max_waitstates & 0xff00) | val;
-                else {
-                    mach->accel.max_waitstates = val;
-                }
-                break;
-            case 0x6aef:
-                if (len == 1)
-                    mach->accel.max_waitstates = (mach->accel.max_waitstates & 0x00ff) | (val << 8);
-                break;
-
-            case 0x6eee:
-                mach_log("6EEE write val = %04x\n", val);
-                if (len == 1)
-                    mach->accel.ge_offset_lo = (mach->accel.ge_offset_lo & 0xff00) | val;
-                else {
-                    mach->accel.ge_offset_lo = val;
-                    dev->accel.ge_offset     = mach->accel.ge_offset_lo;
-                }
-                break;
-            case 0x6eef:
-                if (len == 1) {
-                    mach->accel.ge_offset_lo = (mach->accel.ge_offset_lo & 0x00ff) | (val << 8);
-                    dev->accel.ge_offset     = mach->accel.ge_offset_lo;
-                }
-                break;
-
-            case 0x72ee:
-                mach_log("72EE write val = %04x\n", val);
-                if (len == 1)
-                    mach->accel.ge_offset_hi = (mach->accel.ge_offset_hi & 0xff00) | val;
-                else {
-                    mach->accel.ge_offset_hi = val;
-                    dev->accel.ge_offset     = mach->accel.ge_offset_lo | (mach->accel.ge_offset_hi << 16);
-                }
-                break;
-            case 0x72ef:
-                if (len == 1) {
-                    mach->accel.ge_offset_hi = (mach->accel.ge_offset_hi & 0x00ff) | (val << 8);
-                    dev->accel.ge_offset     = mach->accel.ge_offset_lo | (mach->accel.ge_offset_hi << 16);
-                }
-                break;
-
-            case 0x76ee:
-                mach_log("76EE write val=%d shifted, normal=%d.\n", val << 3, val);
-                dev->ext_pitch = val << 3;
-                svga_recalctimings(svga);
-                break;
-
-            case 0x7aee:
-                mach_log("7AEE write val = %04x, len = %d.\n", val, len);
-                if (len == 1)
-                    mach->accel.ext_ge_config = (mach->accel.ext_ge_config & 0xff00) | val;
-                else {
-                    mach->accel.ext_ge_config = val;
-                    dev->ext_crt_pitch        = dev->internal_pitch;
-                    switch (mach->accel.ext_ge_config & 0x30) {
-                        case 0:
-                        case 0x10:
-                            svga->bpp = 8;
-                            break;
-                        case 0x20:
-                            if ((mach->accel.ext_ge_config & 0xc0) == 0x40)
-                                svga->bpp = 16;
-                            else
-                                svga->bpp = 15;
-
-                            dev->ext_crt_pitch <<= 1;
-                            break;
-                        case 0x30:
-                            svga->bpp = 24;
+                        break;
+                    case 0x30:
+                        dev->bpp = 0;
+                        if (mach->accel.ext_ge_config & 0x200)
+                            dev->ext_crt_pitch <<= 2;
+                        else
                             dev->ext_crt_pitch *= 3;
-                            break;
-
-                        default:
-                            break;
-                    }
-                    if (mach->accel.ext_ge_config & 0x800) {
-                        svga_recalctimings(svga);
-                    }
-                    if (!(mach->accel.ext_ge_config & 0x8000) && !(mach->accel.ext_ge_config & 0x800))
-                        svga_recalctimings(svga);
+                        break;
                 }
-                break;
-            case 0x7aef:
-                mach_log("7AEF write val = %02x.\n", val);
-                if (len == 1) {
-                    mach->accel.ext_ge_config = (mach->accel.ext_ge_config & 0x00ff) | (val << 8);
-                    dev->ext_crt_pitch        = dev->internal_pitch;
-                    switch (mach->accel.ext_ge_config & 0x30) {
-                        case 0:
-                        case 0x10:
-                            svga->bpp = 8;
-                            break;
-                        case 0x20:
-                            if ((mach->accel.ext_ge_config & 0xc0) == 0x40)
-                                svga->bpp = 16;
-                            else
-                                svga->bpp = 15;
+                svga_set_ramdac_type(svga, !!(mach->accel.ext_ge_config & 0x4000));
+                mach_log("Passthrough override = %04x.\n", val & 0x1000);
+                svga_recalctimings(svga);
+            }
+            mach_log("7AEE write val = %04x.\n", mach->accel.ext_ge_config);
+            break;
 
-                            dev->ext_crt_pitch <<= 1;
-                            break;
-                        case 0x30:
-                            svga->bpp = 24;
-                            dev->ext_crt_pitch *= 3;
-                            break;
+        case 0x7eee:
+        case 0x7eef:
+            WRITE8(port, mach->accel.eeprom_control, val);
+            ati_eeprom_write(&mach->eeprom, mach->accel.eeprom_control & 4, mach->accel.eeprom_control & 2, mach->accel.eeprom_control & 1);
+            mach_log("[%04X]: 7EEE+%d VGA ON = %d, Ext = %i, val = %04x, pagesel = %04x.\n", CS, port & 1, vga_on, ibm8514_on, mach->accel.eeprom_control & 0x10ff, (mach->accel.eeprom_control & 0xf0) << 7);
+            svga_recalctimings(svga);
+            break;
 
-                        default:
-                            break;
-                    }
-                    if (mach->accel.ext_ge_config & 0x800) {
-                        svga_recalctimings(svga);
-                    }
-                    if (!(mach->accel.ext_ge_config & 0x8000) && !(mach->accel.ext_ge_config & 0x800))
-                        svga_recalctimings(svga);
-                }
-                break;
-
-            case 0x7eee:
-                mach->accel.eeprom_control = val;
-                break;
-
-            default:
-                break;
-        }
+		default:
+			break;
     }
 }
 
-static uint32_t
-mach_accel_in(uint16_t port, mach_t *mach, int len)
+static uint16_t
+mach_accel_in_fifo(mach_t *mach, svga_t *svga, ibm8514_t *dev, uint16_t port, int len)
 {
-    svga_t         *svga   = &mach->svga;
-    ibm8514_t      *dev    = &svga->dev8514;
-    const uint16_t *vram_w = (uint16_t *) svga->vram;
-    uint16_t        temp   = 0;
-    int             cmd;
-    int             vpos      = dev->displine + svga->y_add;
-    int             vblankend = svga->vblankstart + svga->crtc[0x16];
+	uint16_t  	   *vram_w = (uint16_t *) dev->vram;
+    uint16_t   		temp = 0;
+    int        		cmd;
     int             frgd_sel;
     int             bkgd_sel;
     int             mono_src;
 
     switch (port) {
-        case 0x2e8:
-            if (dev->local) {
-                vpos = svga->displine + svga->y_add;
-                if (vblankend > svga->vtotal) {
-                    vblankend -= svga->vtotal;
-                    if (vpos >= svga->vblankstart || vpos <= vblankend)
-                        temp |= 2;
-                } else {
-                    if (vpos >= svga->vblankstart && vpos <= vblankend)
-                        temp |= 2;
-                }
-            } else {
-                vpos = dev->displine + svga->y_add;
-                if (vblankend > dev->vtotal) {
-                    vblankend -= dev->vtotal;
-                    if (vpos >= svga->vblankstart || vpos <= vblankend)
-                        temp |= 2;
-                } else {
-                    if (vpos >= svga->vblankstart && vpos <= vblankend)
-                        temp |= 2;
-                }
-            }
-            break;
-
-        case 0x6e8:
-            temp = dev->hdisp;
-            break;
-
-        case 0x22e8:
-            temp = dev->disp_cntl;
-            break;
-
-        case 0x26e8:
-            if (len == 1)
-                temp = dev->htotal & 0xff;
-            else
-                temp = dev->htotal;
-            break;
-        case 0x26e9:
-            if (len == 1)
-                temp = dev->htotal >> 8;
-            break;
-
-        case 0x2ee8:
-            temp = dev->subsys_cntl;
-            break;
-
-        case 0x42e8:
-            if (dev->local) {
-                vpos = svga->displine + svga->y_add;
-                if (vblankend > svga->vtotal) {
-                    vblankend -= svga->vtotal;
-                    if (vpos >= svga->vblankstart || vpos <= vblankend)
-                        dev->subsys_stat |= 1;
-                } else {
-                    if (vpos >= svga->vblankstart && vpos <= vblankend)
-                        dev->subsys_stat |= 1;
-                }
-            } else {
-                vpos = dev->displine + svga->y_add;
-                if (vblankend > dev->vtotal) {
-                    vblankend -= dev->vtotal;
-                    if (vpos >= svga->vblankstart || vpos <= vblankend)
-                        dev->subsys_stat |= 1;
-                } else {
-                    if (vpos >= svga->vblankstart && vpos <= vblankend)
-                        dev->subsys_stat |= 1;
-                }
-            }
-
-            if (len != 1) {
-                temp = dev->subsys_stat | 0xa0 | 0x8000;
-            } else {
-                temp = dev->subsys_stat | 0xa0;
-            }
-            break;
-
-        case 0x4ae8:
-            temp = dev->accel.advfunc_cntl;
-            break;
-
-        case 0x42e9:
-            if (len == 1) {
-                temp = dev->subsys_stat >> 8;
-                temp |= 0x80;
-            }
-            break;
-
         case 0x82e8:
         case 0xc2e8:
             if (len != 1) {
@@ -4353,14 +4041,13 @@ mach_accel_in(uint16_t port, mach_t *mach, int len)
                 if (mach_pixel_read(mach)) {
                     cmd = -1;
                     if (len == 1) {
-                        READ_PIXTRANS_BYTE_IO(dev->accel.dx, 1, dev->local)
-
+                        READ_PIXTRANS_BYTE_IO(dev->accel.dx, 1)
                         temp = mach->accel.pix_trans[1];
                     } else {
                         if (mach->accel.cmd_type == 3) {
-                            READ_PIXTRANS_WORD(dev->accel.cx, 0, dev->local)
+                            READ_PIXTRANS_WORD(dev->accel.cx, 0)
                         } else {
-                            READ_PIXTRANS_WORD(dev->accel.dx, 0, dev->local)
+                            READ_PIXTRANS_WORD(dev->accel.dx, 0)
                         }
                         mach_accel_out_pixtrans(mach, dev, port, temp, len);
                     }
@@ -4368,19 +4055,14 @@ mach_accel_in(uint16_t port, mach_t *mach, int len)
             } else {
                 if (ibm8514_cpu_dest(svga)) {
                     cmd = (dev->accel.cmd >> 13);
-                    if (len == 1) {
-                        ; // READ_PIXTRANS_BYTE_IO(0)
-                    } else {
-                        READ_PIXTRANS_WORD(dev->accel.cx, 0, dev->local)
+                    if (len != 1) {
+                        READ_PIXTRANS_WORD(dev->accel.cx, 0)
                         if (dev->accel.input && !dev->accel.odd_in && !dev->accel.sx) {
                             temp &= ~0xff00;
-                            if (dev->local)
-                                temp |= (svga->vram[(dev->accel.newdest_in + dev->accel.cur_x) & svga->vram_mask] << 8);
-                            else
-                                temp |= (dev->vram[(dev->accel.newdest_in + dev->accel.cur_x) & dev->vram_mask] << 8);
+                            temp |= (dev->vram[(dev->accel.newdest_in + dev->accel.cur_x) & dev->vram_mask] << 8);
                         }
                         if (dev->subsys_stat & 1) {
-                            dev->force_busy     = 1;
+                            dev->force_busy = 1;
                             dev->data_available = 1;
                         }
                     }
@@ -4395,9 +4077,9 @@ mach_accel_in(uint16_t port, mach_t *mach, int len)
                 if (mach_pixel_read(mach)) {
                     if (len == 1) {
                         cmd = -1;
-                        READ_PIXTRANS_BYTE_IO(dev->accel.dx, 0, dev->local)
+                        READ_PIXTRANS_BYTE_IO(dev->accel.dx, 0)
 
-                        temp     = mach->accel.pix_trans[0];
+                        temp = mach->accel.pix_trans[0];
                         frgd_sel = (mach->accel.dp_config >> 13) & 7;
                         bkgd_sel = (mach->accel.dp_config >> 7) & 3;
                         mono_src = (mach->accel.dp_config >> 5) & 3;
@@ -4406,26 +4088,23 @@ mach_accel_in(uint16_t port, mach_t *mach, int len)
                             case 0x000: /*8-bit size*/
                                 if (mono_src == 2) {
                                     if ((frgd_sel != 2) && (bkgd_sel != 2)) {
-                                        mach_accel_start(mach->accel.cmd_type, 1, 8, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), 0, mach, dev, len);
+                                        mach_accel_start(mach->accel.cmd_type, 1, 8, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), 0, mach, dev);
                                     } else
-                                        mach_accel_start(mach->accel.cmd_type, 1, 1, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev, len);
+                                        mach_accel_start(mach->accel.cmd_type, 1, 1, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev);
                                 } else
-                                    mach_accel_start(mach->accel.cmd_type, 1, 1, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev, len);
+                                    mach_accel_start(mach->accel.cmd_type, 1, 1, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev);
                                 break;
                             case 0x200: /*16-bit size*/
                                 if (mono_src == 2) {
                                     if ((frgd_sel != 2) && (bkgd_sel != 2)) {
                                         if (mach->accel.dp_config & 0x1000)
-                                            mach_accel_start(mach->accel.cmd_type, 1, 16, mach->accel.pix_trans[1] | (mach->accel.pix_trans[0] << 8), 0, mach, dev, len);
+                                            mach_accel_start(mach->accel.cmd_type, 1, 16, mach->accel.pix_trans[1] | (mach->accel.pix_trans[0] << 8), 0, mach, dev);
                                         else
-                                            mach_accel_start(mach->accel.cmd_type, 1, 16, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), 0, mach, dev, len);
+                                            mach_accel_start(mach->accel.cmd_type, 1, 16, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), 0, mach, dev);
                                     } else
-                                        mach_accel_start(mach->accel.cmd_type, 1, 2, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev, len);
+                                        mach_accel_start(mach->accel.cmd_type, 1, 2, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev);
                                 } else
-                                    mach_accel_start(mach->accel.cmd_type, 1, 2, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev, len);
-                                break;
-
-                            default:
+                                    mach_accel_start(mach->accel.cmd_type, 1, 2, -1, mach->accel.pix_trans[0] | (mach->accel.pix_trans[1] << 8), mach, dev);
                                 break;
                         }
                     }
@@ -4465,196 +4144,10 @@ mach_accel_in(uint16_t port, mach_t *mach, int len)
                     case 0x0a:
                         temp = dev->accel.multifunc[0x0a];
                         break;
-
-                    default:
-                        break;
                 }
             }
             break;
 
-            /*ATI Mach8/32 specific registers*/
-        case 0x12ee:
-            if (len == 1)
-                temp = mach->config1 & 0xff;
-            else
-                temp = mach->config1;
-            break;
-        case 0x12ef:
-            if (len == 1)
-                temp = mach->config1 >> 8;
-            break;
-
-        case 0x16ee:
-            if (len == 1)
-                temp = mach->config2 & 0xff;
-            else
-                temp = mach->config2;
-            break;
-        case 0x16ef:
-            if (len == 1)
-                temp = mach->config2 >> 8;
-            break;
-
-        case 0x32ee:
-            if (len == 1)
-                temp = mach->local_cntl & 0xff;
-            else
-                temp = mach->local_cntl;
-            break;
-        case 0x32ef:
-            if (len == 1)
-                temp = mach->local_cntl >> 8;
-            break;
-
-        case 0x36ee:
-            if (len == 1)
-                temp = mach->misc & 0xff;
-            else
-                temp = mach->misc;
-            break;
-        case 0x36ef:
-            if (len == 1)
-                temp = mach->misc >> 8;
-            break;
-
-        case 0x42ee:
-            temp = mach->accel.test2[0];
-            break;
-        case 0x42ef:
-            temp = mach->accel.test2[1];
-            break;
-
-        case 0x46ee:
-            temp = mach->accel.test3[0];
-            break;
-        case 0x46ef:
-            temp = mach->accel.test3[1];
-            break;
-
-        case 0x4aee:
-            if (len == 1)
-                temp = mach->accel.clock_sel & 0xff;
-            else
-                temp = mach->accel.clock_sel;
-            break;
-        case 0x4aef:
-            if (len == 1)
-                temp = mach->accel.clock_sel >> 8;
-            break;
-
-        case 0x52ee:
-            if (len == 1)
-                temp = mach->accel.scratch0 & 0xff;
-            else
-                temp = mach->accel.scratch0;
-            break;
-        case 0x52ef:
-            if (len == 1)
-                temp = mach->accel.scratch0 >> 8;
-            break;
-
-        case 0x56ee:
-            if (len == 1)
-                temp = mach->accel.scratch1 & 0xff;
-            else
-                temp = mach->accel.scratch1;
-            break;
-        case 0x56ef:
-            if (len == 1)
-                temp = mach->accel.scratch1 >> 8;
-            break;
-
-        case 0x5eee:
-            if (mach->pci_bus)
-                mach->memory_aperture = (mach->memory_aperture & ~0xfff0) | ((mach->linear_base >> 20) << 4);
-
-            if (len == 1)
-                temp = mach->memory_aperture & 0xff;
-            else
-                temp = mach->memory_aperture;
-            break;
-        case 0x5eef:
-            if (len == 1)
-                temp = mach->memory_aperture >> 8;
-            break;
-
-        case 0x62ee:
-            temp = mach->accel.clip_overrun;
-            if (len != 1) {
-                if (mach->force_busy)
-                    temp |= 0x2000;
-                mach->force_busy = 0;
-                if (ati_eeprom_read(&mach->eeprom))
-                    temp |= 0x4000;
-            }
-            mach_log("[%04X:%08X]: 62EE: Temp = %04x, len = %d\n\n", CS, cpu_state.pc, temp, len);
-            break;
-        case 0x62ef:
-            if (len == 1) {
-                if (mach->force_busy2)
-                    temp |= 0x20;
-                mach->force_busy2 = 0;
-                if (ati_eeprom_read(&mach->eeprom))
-                    temp |= 0x40;
-            }
-            mach_log("[%04X:%08X]: 62EF: Temp = %04x, len = %d\n\n", CS, cpu_state.pc, temp, len);
-            break;
-
-        case 0x6aee:
-            if (len == 1)
-                temp = mach->accel.max_waitstates & 0xff;
-            else
-                temp = mach->accel.max_waitstates;
-            break;
-        case 0x6aef:
-            if (len == 1)
-                temp = mach->accel.max_waitstates >> 8;
-            break;
-
-        case 0x72ee:
-            if (len == 1)
-                temp = dev->accel.clip_left & 0xff;
-            else
-                temp = dev->accel.clip_left;
-            break;
-        case 0x72ef:
-            if (len == 1)
-                temp = dev->accel.clip_left >> 8;
-            break;
-
-        case 0x76ee:
-            if (len == 1)
-                temp = dev->accel.clip_top & 0xff;
-            else
-                temp = dev->accel.clip_top;
-            break;
-        case 0x76ef:
-            if (len == 1)
-                temp = dev->accel.clip_top >> 8;
-            break;
-
-        case 0x7aee:
-            if (len == 1)
-                temp = dev->accel.multifunc[4] & 0xff;
-            else
-                temp = dev->accel.multifunc[4];
-            break;
-        case 0x7aef:
-            if (len == 1)
-                temp = dev->accel.multifunc[4] >> 8;
-            break;
-
-        case 0x7eee:
-            if (len == 1)
-                temp = dev->accel.multifunc[3] & 0xff;
-            else
-                temp = dev->accel.multifunc[3];
-            break;
-        case 0x7eef:
-            if (len == 1)
-                temp = dev->accel.multifunc[3] >> 8;
-            break;
-
         case 0x82ee:
             temp = mach->accel.patt_data_idx;
             break;
@@ -4664,6 +4157,8 @@ mach_accel_in(uint16_t port, mach_t *mach, int len)
                 temp = mach->accel.ext_ge_config & 0xff;
             else
                 temp = mach->accel.ext_ge_config;
+
+            mach_log("ExtGE Read = %04x, len=%d.\n", temp, len);
             break;
         case 0x8eef:
             if (len == 1)
@@ -4671,7 +4166,18 @@ mach_accel_in(uint16_t port, mach_t *mach, int len)
             break;
 
         case 0x92ee:
-            temp = mach->accel.eeprom_control;
+            if (len == 1)
+                temp = mach->accel.eeprom_control & 0xff;
+            else
+                temp = mach->accel.eeprom_control;
+
+            mach_log("EEPROM cntl read=%04x, len=%d.\n", temp, len);
+            break;
+        case 0x92ef:
+            if (len == 1)
+                temp = mach->accel.eeprom_control >> 8;
+
+            mach_log("EEPROM cntl read+1=%02x, len=%d.\n", temp, len);
             break;
 
         case 0x96ee:
@@ -4754,40 +4260,39 @@ mach_accel_in(uint16_t port, mach_t *mach, int len)
 
         case 0xceee:
             if (len == 1)
-                temp = svga->vc & 0xff;
+                temp = dev->vc & 0xff;
             else
-                temp = svga->vc & 0x7ff;
+                temp = dev->vc & 0x7ff;
             break;
         case 0xceef:
             if (len == 1)
-                temp = (svga->vc >> 8) & 7;
+                temp = (dev->vc >> 8) & 7;
             break;
 
         case 0xdaee:
             if (len != 1) {
                 temp = mach->accel.src_x;
-                if (dev->local) {
+                if (dev->local >= 2)
                     temp &= 0x7ff;
-                }
             } else
-                temp = dev->accel.destx_distp & 0xff;
+                temp = mach->accel.src_x & 0xff;
             break;
         case 0xdaef:
             if (len == 1)
-                temp = dev->accel.destx_distp >> 8;
+                temp = mach->accel.src_x >> 8;
             break;
 
         case 0xdeee:
             if (len != 1) {
                 temp = mach->accel.src_y;
-                if (dev->local)
+                if (dev->local >= 2)
                     temp &= 0x7ff;
             } else
-                temp = dev->accel.desty_axstp & 0xff;
+                temp = mach->accel.src_y & 0xff;
             break;
         case 0xdeef:
             if (len == 1)
-                temp = dev->accel.desty_axstp >> 8;
+                temp = mach->accel.src_y >> 8;
             break;
 
         case 0xfaee:
@@ -4811,13 +4316,207 @@ mach_accel_in(uint16_t port, mach_t *mach, int len)
                     temp = 0x22;
             }
             break;
+    }
 
-        default:
+    mach_log("Port FIFO IN=%04x, temp=%04x, len=%d.\n", port, temp, len);
+    return temp;
+}
+
+static uint8_t
+mach_accel_in(uint16_t port, mach_t *mach)
+{
+    svga_t         *svga   = &mach->svga;
+    ibm8514_t      *dev    = &svga->dev8514;
+    uint8_t         temp   = 0;
+    int             vpos      = 0;
+    int             vblankend = svga->vblankstart + svga->crtc[0x16];
+
+    switch (port) {
+        case 0x2e8:
+            vpos = dev->vc & 0x7ff;
+            if (vblankend > dev->v_total) {
+                vblankend -= dev->v_total;
+                if (vpos >= svga->vblankstart || vpos <= vblankend)
+                    temp |= 2;
+            } else {
+                if (vpos >= svga->vblankstart && vpos <= vblankend)
+                    temp |= 2;
+            }
             break;
+
+        case 0x6e8:
+            temp = dev->hdisp;
+            break;
+
+        case 0x22e8:
+            temp = dev->disp_cntl;
+            break;
+
+        case 0x26e8:
+        case 0x26e9:
+            READ8(port, dev->htotal);
+            break;
+
+        case 0x2ee8:
+            temp = dev->subsys_cntl;
+            break;
+        case 0x2ee9:
+            temp = 0xff;
+            break;
+
+        case 0x42e8:
+       	case 0x42e9:
+            vpos = dev->vc & 0x7ff;
+            if (vblankend > dev->v_total) {
+                vblankend -= dev->v_total;
+                if (vpos >= svga->vblankstart || vpos <= vblankend)
+                    dev->subsys_stat |= 1;
+            } else {
+                 if (vpos >= svga->vblankstart && vpos <= vblankend)
+                    dev->subsys_stat |= 1;
+            }
+
+            if (port & 1) {
+                temp = 0x80;
+            } else {
+                temp = dev->subsys_stat | 0x80;
+                if (mach->accel.ext_ge_config & 0x08) {
+                    temp |= ((mach->accel.ext_ge_config & 0x07) << 4);
+                } else
+                    temp |= 0x20;
+            }
+            break;
+
+        case 0x4ae8:
+            temp = dev->accel.advfunc_cntl;
+            break;
+
+            /*ATI Mach8/32 specific registers*/
+        case 0x12ee:
+        case 0x12ef:
+            READ8(port, mach->config1);
+            break;
+
+        case 0x16ee:
+        case 0x16ef:
+            READ8(port, mach->config2);
+            break;
+
+        case 0x22ee:
+            if (mach->pci_bus) {
+                temp = mach->pci_cntl_reg;
+            }
+            break;
+
+        case 0x32ee:
+        case 0x32ef:
+            READ8(port, mach->local_cntl);
+            break;
+
+        case 0x36ee:
+        case 0x36ef:
+            READ8(port, mach->misc);
+
+            if (!(port & 1)) {
+                switch (mach->memory) {
+                	case 512:
+                		temp &= ~0x0c;
+                		break;
+                    case 1024:
+                        temp |= 0x04;
+                        break;
+                    case 2048:
+                        temp |= 0x08;
+                        break;
+                    case 4096:
+                        temp |= 0x0c;
+                        break;
+                }
+            }
+            break;
+
+        case 0x42ee:
+            temp = mach->accel.test2[0];
+            break;
+        case 0x42ef:
+            temp = mach->accel.test2[1];
+            break;
+
+        case 0x46ee:
+            temp = mach->accel.test3[0];
+            break;
+        case 0x46ef:
+            temp = mach->accel.test3[1];
+            break;
+
+        case 0x4aee:
+        case 0x4aef:
+            READ8(port, mach->accel.clock_sel);
+            break;
+
+        case 0x52ee:
+        case 0x52ef:
+            READ8(port, mach->accel.scratch0);
+            break;
+
+        case 0x56ee:
+        case 0x56ef:
+            READ8(port, mach->accel.scratch1);
+            break;
+
+        case 0x5eee:
+        case 0x5eef:
+            if (mach->pci_bus)
+                mach->memory_aperture = (mach->memory_aperture & ~0xfff0) | ((mach->linear_base >> 20) << 4);
+            if ((port & 1) && ibm8514_on) {
+                ibm8514_on = 0;
+                vga_on = 1;
+            }
+            READ8(port, mach->memory_aperture);
+            break;
+
+        case 0x62ee:
+            temp = mach->accel.clip_overrun;
+            mach_log("ClipOverrun = %02x.\n", temp);
+            break;
+        case 0x62ef:
+            if (mach->force_busy)
+                temp |= 0x20;
+            mach->force_busy = 0;
+            if (ati_eeprom_read(&mach->eeprom))
+                temp |= 0x40;
+            mach_log("Mach busy temp=%02x.\n", temp);
+            break;
+
+        case 0x6aee:
+        case 0x6aef:
+            READ8(port, mach->accel.max_waitstates);
+            break;
+
+        case 0x72ee:
+        case 0x72ef:
+            READ8(port, dev->accel.clip_left);
+            break;
+
+        case 0x76ee:
+        case 0x76ef:
+            READ8(port, dev->accel.clip_top);
+            break;
+
+        case 0x7aee:
+        case 0x7aef:
+            READ8(port, dev->accel.multifunc[4]);
+            break;
+
+        case 0x7eee:
+        case 0x7eef:
+            READ8(port, dev->accel.multifunc[3]);
+            break;
+
+		default:
+			break;
     }
-    if (port != 0x9ae8 && port != 0x9ae9 && port != 0x62ee && port != 0x9aee) {
-        mach_log("Port accel in = %04x, temp = %04x, len = %d, mode = %d.\n", port, temp, len, dev->ibm_mode);
-    }
+    mach_log("Port accel in = %04x, temp = %04x.\n", port, temp);
     return temp;
 }
 
@@ -4825,28 +4524,92 @@ static void
 mach_accel_outb(uint16_t port, uint8_t val, void *priv)
 {
     mach_t *mach = (mach_t *) priv;
-    mach_accel_out(port, val, mach, 1);
+    svga_t *svga = &mach->svga;
+
+    if (port & 0x8000)
+        mach_accel_out_fifo(mach, svga, &svga->dev8514, port, val, 1);
+    else
+        mach_accel_out(port, val, mach);
 }
 
 static void
 mach_accel_outw(uint16_t port, uint16_t val, void *priv)
 {
     mach_t *mach = (mach_t *) priv;
-    mach_accel_out(port, val, mach, 2);
+    svga_t *svga = &mach->svga;
+
+    if (port & 0x8000)
+        mach_accel_out_fifo(mach, svga, &svga->dev8514, port, val, 2);
+    else {
+        mach_accel_out(port, val, mach);
+        mach_accel_out(port + 1, (val >> 8), mach);
+    }
 }
 
+static void
+mach_accel_outl(uint16_t port, uint32_t val, void *priv)
+{
+    mach_t *mach = (mach_t *) priv;
+    svga_t *svga = &mach->svga;
+
+    if (port & 0x8000) {
+        mach_accel_out_fifo(mach, svga, &svga->dev8514, port, val & 0xffff, 2);
+        mach_accel_out_fifo(mach, svga, &svga->dev8514, port + 2, val >> 16, 2);
+    } else {
+        mach_accel_out(port, val, mach);
+        mach_accel_out(port + 1, (val >> 8), mach);
+        mach_accel_out(port + 2, (val >> 16), mach);
+        mach_accel_out(port + 3, (val >> 24), mach);
+    }
+}
 static uint8_t
 mach_accel_inb(uint16_t port, void *priv)
 {
     mach_t *mach = (mach_t *) priv;
-    return mach_accel_in(port, mach, 1);
+    svga_t *svga = &mach->svga;
+    uint8_t temp;
+
+    if (port & 0x8000)
+        temp = mach_accel_in_fifo(mach, svga, &svga->dev8514, port, 1);
+    else
+        temp = mach_accel_in(port, mach);
+
+    return temp;
 }
 
 static uint16_t
 mach_accel_inw(uint16_t port, void *priv)
 {
     mach_t *mach = (mach_t *) priv;
-    return mach_accel_in(port, mach, 2);
+    svga_t *svga = &mach->svga;
+    uint16_t temp;
+
+    if (port & 0x8000)
+        temp = mach_accel_in_fifo(mach, svga, &svga->dev8514, port, 2);
+    else {
+        temp = mach_accel_in(port, mach);
+        temp |= (mach_accel_in(port + 1, mach) << 8);
+    }
+    return temp;
+}
+
+static uint32_t
+mach_accel_inl(uint16_t port, void *priv)
+{
+    mach_t *mach = (mach_t *) priv;
+    svga_t *svga = &mach->svga;
+    uint32_t temp;
+
+    if (port & 0x8000) {
+        temp = mach_accel_in_fifo(mach, svga, &svga->dev8514, port, 2);
+        temp = (mach_accel_in_fifo(mach, svga, &svga->dev8514, port + 2, 2) << 16);
+    } else {
+        temp = mach_accel_in(port, mach);
+        temp |= (mach_accel_in(port + 1, mach) << 8);
+        temp |= (mach_accel_in(port + 2, mach) << 16);
+        temp |= (mach_accel_in(port + 3, mach) << 24);
+    }
+    return temp;
 }
 
 static void
@@ -4855,7 +4618,8 @@ mach32_ap_writeb(uint32_t addr, uint8_t val, void *priv)
     mach_t *mach       = (mach_t *) priv;
     uint8_t port_dword = addr & 0xfc;
 
-    if ((addr >= ((mach->ap_size << 20) - 0x200)) && (addr < (mach->ap_size << 20))) {
+    if (((mach->local_cntl & 0x20) || (mach->pci_cntl_reg & 0x80)) &&
+        ((addr >= ((mach->ap_size << 20) - 0x200)) && (addr < (mach->ap_size << 20)))) {
         if (addr & 0x100) {
             mach_log("Port WORDB Write=%04x.\n", 0x02ee + (port_dword << 8));
             mach_accel_outb(0x02ee + (addr & 1) + (port_dword << 8), val, mach);
@@ -4875,7 +4639,8 @@ mach32_ap_writew(uint32_t addr, uint16_t val, void *priv)
     mach_t *mach       = (mach_t *) priv;
     uint8_t port_dword = addr & 0xfc;
 
-    if ((addr >= ((mach->ap_size << 20) - 0x200)) && (addr < (mach->ap_size << 20))) {
+    if (((mach->local_cntl & 0x20) || (mach->pci_cntl_reg & 0x80)) &&
+        ((addr >= ((mach->ap_size << 20) - 0x200)) && (addr < (mach->ap_size << 20)))) {
         if (addr & 0x100) {
             mach_log("Port WORDW Write=%04x.\n", 0x02ee + (port_dword << 8));
             mach_accel_outw(0x02ee + (port_dword << 8), val, mach);
@@ -4895,7 +4660,8 @@ mach32_ap_writel(uint32_t addr, uint32_t val, void *priv)
     mach_t *mach       = (mach_t *) priv;
     uint8_t port_dword = addr & 0xfc;
 
-    if ((addr >= ((mach->ap_size << 20) - 0x200)) && (addr < (mach->ap_size << 20))) {
+    if (((mach->local_cntl & 0x20) || (mach->pci_cntl_reg & 0x80)) &&
+        ((addr >= ((mach->ap_size << 20) - 0x200)) && (addr < (mach->ap_size << 20)))) {
         if (addr & 0x100) {
             mach_log("Port WORDL Write=%04x.\n", 0x02ee + (port_dword << 8));
             mach_accel_outw(0x02ee + (port_dword << 8), val & 0xffff, mach);
@@ -4918,7 +4684,8 @@ mach32_ap_readb(uint32_t addr, void *priv)
     uint8_t temp;
     uint8_t port_dword = addr & 0xfc;
 
-    if ((addr >= ((mach->ap_size << 20) - 0x200)) && (addr < (mach->ap_size << 20))) {
+    if (((mach->local_cntl & 0x20) || (mach->pci_cntl_reg & 0x80)) &&
+        ((addr >= ((mach->ap_size << 20) - 0x200)) && (addr < (mach->ap_size << 20)))) {
         if (addr & 0x100) {
             temp = mach_accel_inb(0x02ee + (addr & 1) + (port_dword << 8), mach);
         } else {
@@ -4937,7 +4704,8 @@ mach32_ap_readw(uint32_t addr, void *priv)
     uint16_t temp;
     uint8_t  port_dword = addr & 0xfc;
 
-    if ((addr >= ((mach->ap_size << 20) - 0x200)) && (addr < (mach->ap_size << 20))) {
+    if (((mach->local_cntl & 0x20) || (mach->pci_cntl_reg & 0x80)) &&
+        ((addr >= ((mach->ap_size << 20) - 0x200)) && (addr < (mach->ap_size << 20)))) {
         if (addr & 0x100) {
             temp = mach_accel_inw(0x02ee + (port_dword << 8), mach);
         } else {
@@ -4956,7 +4724,8 @@ mach32_ap_readl(uint32_t addr, void *priv)
     uint32_t temp;
     uint8_t  port_dword = addr & 0xfc;
 
-    if ((addr >= ((mach->ap_size << 20) - 0x200)) && (addr < (mach->ap_size << 20))) {
+    if (((mach->local_cntl & 0x20) || (mach->pci_cntl_reg & 0x80)) &&
+        ((addr >= ((mach->ap_size << 20) - 0x200)) && (addr < (mach->ap_size << 20)))) {
         if (addr & 0x100) {
             temp = mach_accel_inw(0x02ee + (port_dword << 8), mach);
             temp |= (mach_accel_inw(0x02ee + (port_dword << 8) + 4, mach) << 8);
@@ -4984,7 +4753,6 @@ mach32_updatemapping(mach_t *mach)
     if (mach->regs[0xbd] & 4) {
         mem_mapping_set_addr(&svga->mapping, 0xa0000, 0x20000);
         svga->banked_mask = 0xffff;
-        mach_log("Bit 2 of BD.\n");
     } else {
         switch (svga->gdcreg[6] & 0x0c) {
             case 0x0: /*128k at A0000*/
@@ -5020,14 +4788,6 @@ mach32_updatemapping(mach_t *mach)
             mach->ap_size = 4;
             mem_mapping_set_addr(&mach->mmio_linear_mapping, mach->linear_base, mach->ap_size << 20);
         }
-        /*Force IBM/ATI mode on when the MMIO registers are loaded.*/
-        if (mach->local_cntl & 0x20) {
-            if (!ibm8514_on) {
-                ibm8514_on ^= 1;
-                svga->adv_flags |= FLAG_ATI;
-                svga_recalctimings(svga);
-            }
-        }
     } else {
         mach->ap_size = 4;
         mem_mapping_disable(&mach->mmio_linear_mapping);
@@ -5037,39 +4797,36 @@ mach32_updatemapping(mach_t *mach)
 static void
 mach32_hwcursor_draw(svga_t *svga, int displine)
 {
-    const mach_t *mach = (mach_t *) svga->priv;
+    mach_t *mach 	   = (mach_t *) svga->priv;
+    ibm8514_t *dev 	   = &svga->dev8514;
     uint16_t      dat;
     int           comb;
-    int           offset = svga->hwcursor_latch.x - svga->hwcursor_latch.xoff;
+    int           offset = dev->hwcursor_latch.x - dev->hwcursor_latch.xoff;
     uint32_t      color0;
     uint32_t      color1;
 
     if (svga->bpp == 8) {
-        color0 = svga->pallook[mach->cursor_col_0];
-        color1 = svga->pallook[mach->cursor_col_1];
-        mach_log("8BPP: Offset = %x, XOFF = %02x, YOFF = %02x.\n", offset, svga->hwcursor_latch.xoff, svga->hwcursor_latch.yoff);
+        color0 = dev->pallook[mach->cursor_col_0];
+        color1 = dev->pallook[mach->cursor_col_1];
     } else if (svga->bpp == 15) {
         color0 = video_15to32[((mach->ext_cur_col_0_r << 16) | (mach->ext_cur_col_0_g << 8) | mach->cursor_col_0) & 0xffff];
         color1 = video_15to32[((mach->ext_cur_col_1_r << 16) | (mach->ext_cur_col_1_g << 8) | mach->cursor_col_1) & 0xffff];
-        mach_log("15BPP: Offset = %x, XOFF = %02x, YOFF = %02x.\n", offset, svga->hwcursor_latch.xoff, svga->hwcursor_latch.yoff);
     } else if (svga->bpp == 16) {
         color0 = video_16to32[((mach->ext_cur_col_0_r << 16) | (mach->ext_cur_col_0_g << 8) | mach->cursor_col_0) & 0xffff];
         color1 = video_16to32[((mach->ext_cur_col_1_r << 16) | (mach->ext_cur_col_1_g << 8) | mach->cursor_col_1) & 0xffff];
-        mach_log("16BPP: Offset = %x, XOFF = %02x, YOFF = %02x.\n", offset, svga->hwcursor_latch.xoff, svga->hwcursor_latch.yoff);
     } else {
         color0 = ((mach->ext_cur_col_0_r << 16) | (mach->ext_cur_col_0_g << 8) | mach->cursor_col_0);
         color1 = ((mach->ext_cur_col_1_r << 16) | (mach->ext_cur_col_1_g << 8) | mach->cursor_col_1);
-        mach_log("24BPP: Offset = %x, XOFF = %02x, YOFF = %02x.\n", offset, svga->hwcursor_latch.xoff, svga->hwcursor_latch.yoff);
     }
 
-    if (svga->interlace && svga->hwcursor_oddeven)
-        svga->hwcursor_latch.addr += 16;
+    if (dev->interlace && dev->hwcursor_oddeven)
+        dev->hwcursor_latch.addr += 16;
 
     for (int x = 0; x < 64; x += 8) {
-        dat = svga->vram[svga->hwcursor_latch.addr & svga->vram_mask] | (svga->vram[(svga->hwcursor_latch.addr + 1) & svga->vram_mask] << 8);
+        dat = dev->vram[dev->hwcursor_latch.addr & dev->vram_mask] | (dev->vram[(dev->hwcursor_latch.addr + 1) & dev->vram_mask] << 8);
         for (int xx = 0; xx < 8; xx++) {
             comb = (dat >> (xx << 1)) & 0x03;
-            if (offset >= svga->hwcursor_latch.x) {
+            if (offset >= dev->hwcursor_latch.x) {
                 switch (comb) {
                     case 0:
                         (svga->monitor->target_buffer->line[displine])[offset + svga->x_add] = color0;
@@ -5087,250 +4844,286 @@ mach32_hwcursor_draw(svga_t *svga, int displine)
             }
             offset++;
         }
-        svga->hwcursor_latch.addr += 2;
+        dev->hwcursor_latch.addr += 2;
     }
-    if (svga->interlace && !svga->hwcursor_oddeven)
-        svga->hwcursor_latch.addr += 16;
+    if (dev->interlace && !dev->hwcursor_oddeven)
+        dev->hwcursor_latch.addr += 16;
 }
 
 static void
 mach_io_remove(mach_t *mach)
 {
-    io_removehandler(0x01ce, 2,
-                     mach_in, NULL, NULL,
-                     mach_out, NULL, NULL, mach);
-    io_removehandler(0x03c0, 32,
-                     mach_in, NULL, NULL,
-                     mach_out, NULL, NULL, mach);
+    io_removehandler(0x2e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x6e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xae8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xee8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x12e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x16e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x1ae8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x1ee8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x22e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x26e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x2ee8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x42e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x4ae8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x52e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x56e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x5ae8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x5ee8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x82e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x86e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x8ae8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x8ee8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x92e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x96e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x9ae8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x9ee8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xa2e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xa6e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xaae8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xaee8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xb2e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xb6e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xbae8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xbee8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xe2e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
 
-    io_removehandler(0x2e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x2ea, 0x0004, mach_in, NULL, NULL, mach_out, NULL, NULL, mach);
-    io_removehandler(0x6e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xae8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xee8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x12e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x16e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x1ae8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x1ee8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x22e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x26e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x2ee8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x42e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x4ae8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x52e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x56e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x5ae8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x5ee8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x82e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x86e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x8ae8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x8ee8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x92e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x96e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x9ae8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x9ee8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xa2e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xa6e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xaae8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xaee8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xb2e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xb6e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xbae8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xbee8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xe2e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
+    io_removehandler(0xc2e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xc6e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xcae8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xcee8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xd2e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xd6e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xdae8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xdee8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xe6e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xeae8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xeee8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xf2e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xf6e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xfae8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xfee8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
 
-    io_removehandler(0xc2e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xc6e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xcae8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xcee8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xd2e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xd6e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xdae8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xdee8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xe6e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xeae8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xeee8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xf2e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xf6e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xfae8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xfee8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-
-    io_removehandler(0x06ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x0aee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x0eee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x12ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x16ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x1aee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x1eee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x26ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x2aee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x2eee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x32ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x36ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x3aee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x3eee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x42ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x46ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x4aee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x52ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x56ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x5aee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x5eee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x62ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x66ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x6aee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x6eee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x72ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x76ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x7aee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x7eee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x82ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x8eee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x92ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x96ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0x9aee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xa2ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xa6ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xaaee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xaeee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xb2ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xb6ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xbaee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xbeee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xc2ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xc6ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xcaee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xceee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xd2ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xd6ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xdaee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xdeee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xe2ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xe6ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xeeee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xf2ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xfaee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_removehandler(0xfeee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
+    io_removehandler(0x02ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x06ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x0aee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x0eee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x12ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x16ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x1aee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x1eee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x22ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x26ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x2aee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x2eee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x32ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x36ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x3aee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x3eee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x42ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x46ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x4aee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x52ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x56ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x5aee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x5eee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x62ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x66ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x6aee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x6eee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x72ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x76ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x7aee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x7eee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x82ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x8eee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x92ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x96ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0x9aee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xa2ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xa6ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xaaee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xaeee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xb2ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xb6ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xbaee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xbeee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xc2ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xc6ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xcaee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xceee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xd2ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xd6ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xdaee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xdeee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xe2ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xe6ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xeeee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xf2ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xfaee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_removehandler(0xfeee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
 }
 
 static void
 mach_io_set(mach_t *mach)
 {
-    io_sethandler(0x01ce, 2,
-                  mach_in, NULL, NULL,
-                  mach_out, NULL, NULL, mach);
-    io_sethandler(0x03c0, 32,
-                  mach_in, NULL, NULL,
-                  mach_out, NULL, NULL, mach);
+    io_sethandler(0x2e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x6e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xae8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xee8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x12e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x16e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x1ae8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x1ee8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x22e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x26e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x2ee8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x42e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x4ae8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x52e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x56e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x5ae8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x5ee8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x82e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x86e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x8ae8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x8ee8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x92e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x96e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x9ae8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x9ee8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xa2e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xa6e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xaae8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xaee8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xb2e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xb6e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xbae8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xbee8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xe2e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
 
-    io_sethandler(0x2e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x2ea, 0x0004, mach_in, NULL, NULL, mach_out, NULL, NULL, mach);
-    io_sethandler(0x6e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xae8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xee8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x12e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x16e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x1ae8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x1ee8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x22e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x26e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x2ee8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x42e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x4ae8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x52e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x56e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x5ae8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x5ee8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x82e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x86e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x8ae8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x8ee8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x92e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x96e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x9ae8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x9ee8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xa2e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xa6e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xaae8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xaee8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xb2e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xb6e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xbae8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xbee8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xe2e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
+    io_sethandler(0xc2e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xc6e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xcae8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xcee8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xd2e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xd6e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xdae8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xdee8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xe6e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xeae8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xeee8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xf2e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xf6e8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xfae8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xfee8, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
 
-    io_sethandler(0xc2e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xc6e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xcae8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xcee8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xd2e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xd6e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xdae8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xdee8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xe6e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xeae8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xeee8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xf2e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xf6e8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xfae8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xfee8, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
+    io_sethandler(0x02ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x06ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x0aee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x0eee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x12ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x16ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x1aee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x1eee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x22ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x26ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x2aee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x2eee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x32ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x36ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x3aee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x3eee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x42ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x46ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x4aee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x52ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x56ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x5aee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x5eee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x62ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x66ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x6aee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x6eee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x72ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x76ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x7aee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x7eee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x82ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x8eee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x92ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x96ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0x9aee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xa2ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xa6ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xaaee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xaeee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xb2ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xb6ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xbaee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xbeee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xc2ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xc6ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xcaee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xceee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xd2ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xd6ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xdaee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xdeee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xe2ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xe6ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xeeee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xf2ee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xfaee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+    io_sethandler(0xfeee, 0x0002, mach_accel_inb, mach_accel_inw, mach_accel_inl, mach_accel_outb, mach_accel_outw, mach_accel_outl, mach);
+}
 
-    io_sethandler(0x06ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x0aee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x0eee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x12ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x16ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x1aee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x1eee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x26ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x2aee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x2eee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x32ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x36ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x3aee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x3eee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x42ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x46ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x4aee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x52ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x56ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x5aee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x5eee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x62ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x66ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x6aee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x6eee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x72ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x76ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x7aee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x7eee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x82ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x8eee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x92ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x96ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0x9aee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xa2ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xa6ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xaaee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xaeee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xb2ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xb6ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xbaee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xbeee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xc2ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xc6ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xcaee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xceee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xd2ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xd6ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xdaee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xdeee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xe2ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xe6ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xeeee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xf2ee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xfaee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
-    io_sethandler(0xfeee, 0x0002, mach_accel_inb, mach_accel_inw, NULL, mach_accel_outb, mach_accel_outw, NULL, mach);
+static uint8_t
+mach_mca_read(int port, void *p)
+{
+    mach_t *mach = (mach_t *) p;
+    mach_log("[%04X]: MCA read port = %x, val = %02x.\n", CS, port & 7, mach->pos_regs[port & 7]);
+    return mach->pos_regs[port & 7];
+}
+
+static void
+mach_mca_write(int port, uint8_t val, void *priv)
+{
+    mach_t *mach = (mach_t *) priv;
+    svga_t *svga = &mach->svga;
+
+    if (port < 0x102)
+        return;
+
+    mach->pos_regs[port & 7] = val;
+    mach_log("[%04X]: MCA write port = %x, val = %02x, biosaddr = %05x.\n", CS, port & 7, mach->pos_regs[port & 7], (((mach->pos_regs[3] & 0x3e) << 0x0c) >> 1) + 0xc0000);
+    mem_mapping_disable(&mach->bios_rom.mapping);
+    mem_mapping_disable(&mach->bios_rom2.mapping);
+    if (mach->pos_regs[2] & 0x01) {
+        mem_mapping_enable(&mach->bios_rom.mapping);
+        mem_mapping_enable(&mach->bios_rom2.mapping);
+    }
+}
+
+static uint8_t
+mach_mca_feedb(void *p)
+{
+    mach_t *mach = (mach_t *) p;
+    mach_log("FeedB = %x.\n", mach->pos_regs[2] & 0x01);
+    return mach->pos_regs[2] & 0x01;
+}
+
+static void
+mach_mca_reset(void *p)
+{
+    mach_t *mach = (mach_t *) p;
+
+    mem_mapping_disable(&mach->bios_rom.mapping);
+    mem_mapping_disable(&mach->bios_rom2.mapping);
+    mach_log("MCA reset.\n");
+    ibm8514_on = 0;
+    vga_on = 1;
+    mach_mca_write(0x102, 0, mach);
 }
 
 static uint8_t
@@ -5339,6 +5132,9 @@ mach32_pci_read(UNUSED(int func), int addr, void *priv)
     const mach_t *mach = (mach_t *) priv;
     uint8_t       ret  = 0x00;
 
+    if ((addr >= 0x30) && (addr <= 0x33) && !mach->has_bios)
+        return ret;
+
     switch (addr) {
         case 0x00:
             ret = 0x02; /*ATI*/
@@ -5413,14 +5209,18 @@ static void
 mach32_pci_write(UNUSED(int func), int addr, uint8_t val, void *priv)
 {
     mach_t *mach = (mach_t *) priv;
+    if ((addr >= 0x30) && (addr <= 0x33) && !mach->has_bios)
+        return;
 
     switch (addr) {
         case PCI_REG_COMMAND:
             mach->pci_regs[PCI_REG_COMMAND] = val & 0x27;
-            mach_io_remove(mach);
             if (val & PCI_COMMAND_IO) {
-                mach_io_set(mach);
-            }
+                io_removehandler(0x03c0, 32, mach_in, NULL, NULL, mach_out, NULL, NULL, mach);
+                io_sethandler(0x03c0, 32, mach_in, NULL, NULL, mach_out, NULL, NULL, mach);
+            } else
+                io_removehandler(0x03c0, 32, mach_in, NULL, NULL, mach_out, NULL, NULL, mach);
+
             mach32_updatemapping(mach);
             break;
 
@@ -5462,7 +5262,6 @@ mach8_init(const device_t *info)
     mach_t    *mach;
     svga_t    *svga;
     ibm8514_t *dev;
-    uint32_t   memory;
 
     mach = malloc(sizeof(mach_t));
     memset(mach, 0x00, sizeof(mach_t));
@@ -5472,30 +5271,42 @@ mach8_init(const device_t *info)
 
     mach->pci_bus    = !!(info->flags & DEVICE_PCI);
     mach->vlb_bus    = !!(info->flags & DEVICE_VLB);
-    dev->local       = info->local;
-    dev->vram_size   = (1024 << 10);
-    dev->vram        = calloc(dev->vram_size, 1);
-    dev->changedvram = calloc(dev->vram_size >> 12, 1);
-    dev->vram_mask   = dev->vram_size - 1;
-    dev->map8        = svga->pallook;
-    memory           = device_get_config_int("memory");
+    mach->mca_bus    = !!(info->flags & DEVICE_MCA);
+    dev->type        = info->flags;
+    dev->local       = info->local & 0xff;
+    mach->has_bios   = !(info->local & 0xff00);
+    mach->memory     = device_get_config_int("memory");
+    mach->ramdac_type = mach->pci_bus ? device_get_config_int("ramdac") : 1;
 
-    if (dev->local) {
-        if (mach->vlb_bus)
+    if (dev->local >= 2) {
+        if (mach->pci_bus) {
+            if (mach->has_bios) {
+                rom_init(&mach->bios_rom,
+                         BIOS_MACH32_PCI_ROM_PATH,
+                         0xc0000, 0x8000, 0x7fff,
+                         0, MEM_MAPPING_EXTERNAL);
+            }
+        }
+        else if (mach->vlb_bus)
             rom_init(&mach->bios_rom,
                      BIOS_MACH32_VLB_ROM_PATH,
                      0xc0000, 0x8000, 0x7fff,
                      0, MEM_MAPPING_EXTERNAL);
-        else if (mach->pci_bus)
+        else if (mach->mca_bus) {
             rom_init(&mach->bios_rom,
-                     BIOS_MACH32_PCI_ROM_PATH,
+                     BIOS_MACH32_MCA_ROM_PATH,
                      0xc0000, 0x8000, 0x7fff,
                      0, MEM_MAPPING_EXTERNAL);
-        else
+            rom_init(&mach->bios_rom2,
+                     BIOS_MACH32_MCA_ROM_PATH,
+                     0xc8000, 0x1000, 0x0fff,
+                     0x8000, MEM_MAPPING_EXTERNAL);
+        } else {
             rom_init(&mach->bios_rom,
                      BIOS_MACH32_ISA_ROM_PATH,
                      0xc0000, 0x8000, 0x7fff,
                      0, MEM_MAPPING_EXTERNAL);
+        }
     } else {
         rom_init(&mach->bios_rom,
                  BIOS_MACH8_ROM_PATH,
@@ -5503,75 +5314,108 @@ mach8_init(const device_t *info)
                  0, MEM_MAPPING_EXTERNAL);
     }
 
-    svga_init(info, svga, mach, dev->local ? (memory << 10) : (512 << 10), /*default: 512kB for Mach8, 2MB for Mach32*/
-              mach_recalctimings,
-              mach_in, mach_out,
-              dev->local ? mach32_hwcursor_draw : NULL,
-              NULL);
-
-    if (dev->local) {
-        switch (memory) {
-            case 1024:
-                mach->misc |= 0x04;
-                break;
-            case 2048:
-                mach->misc |= 0x08;
-                break;
-            case 4096:
-                mach->misc |= 0x0c;
-                break;
-
-            default:
-                break;
-        }
-        svga->hwcursor.cur_ysize = 64;
-        mach->config1            = 0x20;
-        mach->config2            = 0x08;
-        /*Fake the RAMDAC to give the VLB/MCA variants full 24-bit support until said RAMDAC is implemented.*/
+    if (dev->local >= 2) {
+        svga_init(info, svga, mach, mach->memory << 10, /*default: 2MB for Mach32*/
+                      mach_recalctimings,
+                      mach_in, mach_out,
+                      mach32_hwcursor_draw,
+                      NULL);
+        dev->hwcursor.cur_ysize = 64;
+        dev->vram_size   = svga->vram_max;
+        dev->vram        = svga->vram;
+        dev->changedvram = svga->changedvram;
+        dev->vram_mask   = svga->vram_mask;
+        mach->config1 = 0x20;
+        mach->config2 = 0x08;
+        if (mach->pci_bus && !mach->ramdac_type)
+            svga->ramdac = device_add(&ati68860_ramdac_device);
+        else
+            svga->ramdac = device_add(&ati68875_ramdac_device);
         if (mach->vlb_bus) {
             video_inform(VIDEO_FLAG_TYPE_8514, &timing_mach32_vlb);
-            mach->config1 |= 0x0c;
+            if (!is486)
+                mach->config1 |= 0x0a;
+            else
+                mach->config1 |= 0x0c;
             mach->config1 |= 0x0400;
+            svga->clock_gen = device_add(&ati18811_0_device);
+        } else if (mach->mca_bus) {
+            video_inform(VIDEO_FLAG_TYPE_8514, &timing_mach32_mca);
+            if (is286 && !is386)
+                mach->config1 |= 0x04;
+            else
+                mach->config1 |= 0x06;
+            mach->config1 |= 0x0400;
+            mach->config1 |= 0x1000;
+            svga->clock_gen = device_add(&ati18811_1_device);
         } else if (mach->pci_bus) {
             video_inform(VIDEO_FLAG_TYPE_8514, &timing_mach32_pci);
             mach->config1 |= 0x0e;
-            mach->config1 |= 0x0a00;
+            if (mach->ramdac_type)
+                mach->config1 |= 0x0400;
+            else
+                mach->config1 |= 0x0a00;
             mach->config2 |= 0x2000;
-            svga->ramdac = device_add(&ati68860_ramdac_device);
+            svga->clock_gen = device_add(&ati18811_0_device);
         } else {
             video_inform(VIDEO_FLAG_TYPE_8514, &timing_gfxultra_isa);
+            mach->config1 |= 0x0400;
+            svga->clock_gen = device_add(&ati18811_0_device);
         }
         mem_mapping_add(&mach->mmio_linear_mapping, 0, 0, mach32_ap_readb, mach32_ap_readw, mach32_ap_readl, mach32_ap_writeb, mach32_ap_writew, mach32_ap_writel, NULL, MEM_MAPPING_EXTERNAL, mach);
         mem_mapping_disable(&mach->mmio_linear_mapping);
     } else {
+        svga_init(info, svga, mach, (512 << 10), /*default: 512kB VGA for 28800-6 + 1MB for Mach8*/
+                      mach_recalctimings,
+                      mach_in, mach_out,
+                      NULL,
+                      NULL);
+        dev->vram_size   = (1024 << 10);
+        dev->vram        = calloc(dev->vram_size, 1);
+        dev->changedvram = calloc(dev->vram_size >> 12, 1);
+        dev->vram_mask   = dev->vram_size - 1;
         video_inform(VIDEO_FLAG_TYPE_8514, &timing_gfxultra_isa);
-        mach->config1  = 0x02 | 0x20 | 0x80;
-        mach->config2  = 0x02;
-        dev->ext_pitch = 1024;
+        mach->config1 = 0x02 | 0x20 | 0x80;
+        mach->config2 = 0x02;
+        svga->clock_gen = device_add(&ati18810_device);
     }
+    dev->bpp = 0;
+    svga->getclock  = ics2494_getclock;
 
+    dev->ext_pitch = 1024;
+    dev->ext_crt_pitch = 0x80;
     svga->force_old_addr = 1;
-    svga->miscout        = 1;
-    svga->bpp            = 8;
-    svga->packed_chain4  = 1;
-    ibm8514_enabled      = 1;
-    ibm8514_has_vga      = 1;
-    dev->ibm_mode        = 1;
-    dev->rowoffset       = 128;
+    svga->miscout = 1;
+    svga->bpp = 8;
+    svga->packed_chain4 = 1;
+    dev->rowoffset = 0x80;
+    io_sethandler(0x01ce, 2,  mach_in, NULL, NULL, mach_out, NULL, NULL, mach);
+    io_sethandler(0x03c0, 32, mach_in, NULL, NULL, mach_out, NULL, NULL, mach);
+    io_sethandler(0x02ea, 4,  mach_in, NULL, NULL, mach_out, NULL, NULL, mach);
+    io_sethandler(0x46e8, 1,  mach_in, NULL, NULL, mach_out, NULL, NULL, mach);
     mach_io_set(mach);
 
-    if (dev->local) {
+    if (dev->local >= 2) {
         svga->decode_mask     = (4 << 20) - 1;
         mach->cursor_col_1    = 0xff;
         mach->ext_cur_col_1_r = 0xff;
         mach->ext_cur_col_1_g = 0xff;
-        dev->ext_crt_pitch    = 128;
         if (mach->vlb_bus)
             ati_eeprom_load(&mach->eeprom, "mach32_vlb.nvr", 1);
-        else if (mach->pci_bus) {
-            ati_eeprom_load(&mach->eeprom, "mach32_pci.nvr", 1);
+        else if (mach->mca_bus) {
+            ati_eeprom_load(&mach->eeprom, "mach32_mca.nvr", 1);
             mem_mapping_disable(&mach->bios_rom.mapping);
-            pci_add_card(PCI_ADD_VIDEO, mach32_pci_read, mach32_pci_write, mach, &mach->pci_slot);
+            mem_mapping_disable(&mach->bios_rom2.mapping);
+            mach->pos_regs[0] = 0x89;
+            mach->pos_regs[1] = 0x80;
+            mca_add(mach_mca_read, mach_mca_write, mach_mca_feedb, mach_mca_reset, mach);
+        } else if (mach->pci_bus) {
+            ati_eeprom_load(&mach->eeprom, "mach32_pci.nvr", 1);
+			if (mach->has_bios) {
+	            mem_mapping_disable(&mach->bios_rom.mapping);
+                pci_add_card(PCI_ADD_NORMAL, mach32_pci_read, mach32_pci_write, mach, &mach->pci_slot);
+			} else
+                pci_add_card(PCI_ADD_VIDEO, mach32_pci_read, mach32_pci_write, mach, &mach->pci_slot);
             mach->pci_regs[PCI_REG_COMMAND] = 0x83;
             mach->pci_regs[0x30]            = 0x00;
             mach->pci_regs[0x32]            = 0x0c;
@@ -5603,6 +5447,12 @@ mach32_vlb_available(void)
     return rom_present(BIOS_MACH32_VLB_ROM_PATH);
 }
 
+static int
+mach32_mca_available(void)
+{
+    return rom_present(BIOS_MACH32_MCA_ROM_PATH);
+}
+
 static int
 mach32_pci_available(void)
 {
@@ -5616,7 +5466,7 @@ mach_close(void *priv)
     svga_t    *svga = &mach->svga;
     ibm8514_t *dev  = &svga->dev8514;
 
-    if (dev) {
+    if (dev && (dev->local < 2)) {
         free(dev->vram);
         free(dev->changedvram);
     }
@@ -5640,7 +5490,7 @@ mach_force_redraw(void *priv)
     mach_t *mach = (mach_t *) priv;
     svga_t *svga = &mach->svga;
 
-    svga->fullchange = changeframecount;
+    svga->fullchange = svga->monitor->mon_changeframecount;
 }
 
 // clang-format off
@@ -5677,11 +5527,64 @@ static const device_config_t mach32_config[] = {
     }
 };
 
+// clang-format off
+static const device_config_t mach32_pci_config[] = {
+    {
+        .name = "ramdac",
+        .description = "RAMDAC type",
+        .type = CONFIG_SELECTION,
+        .default_int = 0,
+        .selection = {
+            {
+                .description = "ATI 68860",
+                .value = 0
+            },
+            {
+                .description = "ATI 68875",
+                .value = 1
+            },
+            {
+                .description = ""
+            }
+        }
+    },
+    {
+        .name = "memory",
+        .description = "Memory size",
+        .type = CONFIG_SELECTION,
+        .default_int = 2048,
+        .selection = {
+            {
+                .description = "512 KB",
+                .value = 512
+            },
+            {
+                .description = "1 MB",
+                .value = 1024
+            },
+            {
+                .description = "2 MB",
+                .value = 2048
+            },
+            {
+                .description = "4 MB",
+                .value = 4096
+            },
+            {
+                .description = ""
+            }
+        }
+    },
+    {
+        .type = CONFIG_END
+    }
+};
+
 const device_t mach8_isa_device = {
     .name = "ATI Mach8 (ISA)",
     .internal_name = "mach8_isa",
     .flags = DEVICE_ISA,
-    .local = 0,
+    .local = 1,
     .init = mach8_init,
     .close = mach_close,
     .reset = NULL,
@@ -5695,7 +5598,7 @@ const device_t mach32_isa_device = {
     .name = "ATI Mach32 (ISA)",
     .internal_name = "mach32_isa",
     .flags = DEVICE_ISA,
-    .local = 1,
+    .local = 2,
     .init = mach8_init,
     .close = mach_close,
     .reset = NULL,
@@ -5709,7 +5612,7 @@ const device_t mach32_vlb_device = {
     .name = "ATI Mach32 (VLB)",
     .internal_name = "mach32_vlb",
     .flags = DEVICE_VLB,
-    .local = 1,
+    .local = 2,
     .init = mach8_init,
     .close = mach_close,
     .reset = NULL,
@@ -5719,16 +5622,45 @@ const device_t mach32_vlb_device = {
     .config = mach32_config
 };
 
+const device_t mach32_mca_device = {
+    .name = "ATI Mach32 (MCA)",
+    .internal_name = "mach32_mca",
+    .flags = DEVICE_MCA,
+    .local = 2,
+    .init = mach8_init,
+    .close = mach_close,
+    .reset = NULL,
+    { .available = mach32_mca_available },
+    .speed_changed = mach_speed_changed,
+    .force_redraw = mach_force_redraw,
+    .config = mach32_config
+};
+
 const device_t mach32_pci_device = {
     .name = "ATI Mach32 (PCI)",
     .internal_name = "mach32_pci",
     .flags = DEVICE_PCI,
-    .local = 1,
+    .local = 2,
     .init = mach8_init,
     .close = mach_close,
     .reset = NULL,
     { .available = mach32_pci_available },
     .speed_changed = mach_speed_changed,
     .force_redraw = mach_force_redraw,
-    .config = mach32_config
+    .config = mach32_pci_config
 };
+
+const device_t mach32_onboard_pci_device = {
+    .name = "ATI Mach32 (PCI) On-Board",
+    .internal_name = "mach32_pci_onboard",
+    .flags = DEVICE_PCI,
+    .local = 2 | 0x100,
+    .init = mach8_init,
+    .close = mach_close,
+    .reset = NULL,
+    { .available = NULL },
+    .speed_changed = mach_speed_changed,
+    .force_redraw = mach_force_redraw,
+    .config = mach32_pci_config
+};
+
diff --git a/src/video/vid_et4000w32.c b/src/video/vid_et4000w32.c
index 791b65912..10e9695b4 100644
--- a/src/video/vid_et4000w32.c
+++ b/src/video/vid_et4000w32.c
@@ -500,7 +500,8 @@ et4000w32p_recalctimings(svga_t *svga)
     switch (svga->bpp) {
         case 15:
         case 16:
-            svga->hdisp >>= 1;
+            if ((svga->gdcreg[6] & 1) || (svga->attrregs[0x10] & 1))
+                svga->hdisp >>= 1;
             if (et4000->type <= ET4000W32P_REVC) {
                 if (et4000->type == ET4000W32P_REVC) {
                     if (svga->hdisp != 1024)
@@ -513,7 +514,7 @@ et4000w32p_recalctimings(svga_t *svga)
             svga->hdisp /= 3;
             if (et4000->type <= ET4000W32P_REVC)
                 et4000->adjust_cursor = 2;
-            if (et4000->type == ET4000W32P_DIAMOND && (svga->hdisp == 640 / 2 || svga->hdisp == 1232)) {
+            if ((et4000->type == ET4000W32P_DIAMOND) && ((svga->hdisp == (640 / 2)) || (svga->hdisp == 1232))) {
                 svga->hdisp = 640;
             }
             break;
@@ -1072,8 +1073,7 @@ et4000w32p_mmu_read(uint32_t addr, void *priv)
                 case 0x8e:
                     if (et4000->type >= ET4000W32P_REVC)
                         return et4000->acl.internal.pixel_depth;
-                    else
-                        return et4000->acl.internal.vbus;
+                    return et4000->acl.internal.vbus;
                 case 0x8f:
                     return et4000->acl.internal.xy_dir;
                 case 0x90:
diff --git a/src/video/vid_oak_oti.c b/src/video/vid_oak_oti.c
index 932e13913..cdc518c0c 100644
--- a/src/video/vid_oak_oti.c
+++ b/src/video/vid_oak_oti.c
@@ -84,8 +84,9 @@ oti_out(uint16_t addr, uint8_t val, void *priv)
             if (!oti->chip_id) {
                 oti->enable_register = val & 1;
                 return;
-            } else
-                break;
+            }
+            svga_out(addr, val, svga);
+            return;
 
         case 0x3c6:
         case 0x3c7:
diff --git a/src/video/vid_svga.c b/src/video/vid_svga.c
index 0e83c034a..0b75e4b09 100644
--- a/src/video/vid_svga.c
+++ b/src/video/vid_svga.c
@@ -111,10 +111,48 @@ void
 svga_out(uint16_t addr, uint8_t val, void *priv)
 {
     svga_t *svga = (svga_t *) priv;
+    ibm8514_t *dev = &svga->dev8514;
+    xga_t *xga = &svga->xga;
     uint8_t o;
     uint8_t index;
 
     switch (addr) {
+        case 0x2ea:
+            dev->dac_mask = val;
+            break;
+        case 0x2eb:
+        case 0x2ec:
+            dev->dac_pos    = 0;
+            dev->dac_status = addr & 0x03;
+            dev->dac_addr   = (val + (addr & 0x01)) & 0xff;
+            break;
+        case 0x2ed:
+            svga->fullchange = svga->monitor->mon_changeframecount;
+            switch (dev->dac_pos) {
+                case 0:
+                    dev->dac_r = val;
+                    dev->dac_pos++;
+                    break;
+                case 1:
+                    dev->dac_g = val;
+                    dev->dac_pos++;
+                    break;
+                case 2:
+                    index                 = dev->dac_addr & 0xff;
+                    dev->dac_b = val;
+                    svga->vgapal[index].r = dev->dac_r;
+                    svga->vgapal[index].g = dev->dac_g;
+                    svga->vgapal[index].b = dev->dac_b;
+                    if (svga->ramdac_type == RAMDAC_8BIT)
+                        dev->pallook[index] = makecol32(svga->vgapal[index].r, svga->vgapal[index].g, svga->vgapal[index].b);
+                    else
+                        dev->pallook[index] = makecol32(video_6to8[svga->vgapal[index].r & 0x3f], video_6to8[svga->vgapal[index].g & 0x3f], video_6to8[svga->vgapal[index].b & 0x3f]);
+                    dev->dac_pos  = 0;
+                    dev->dac_addr = (dev->dac_addr + 1) & 0xff;
+                    break;
+            }
+            break;
+
         case 0x3c0:
         case 0x3c1:
             if (!svga->attrff) {
@@ -167,10 +205,12 @@ svga_out(uint16_t addr, uint8_t val, void *priv)
             svga_recalctimings(svga);
             break;
         case 0x3c3:
-            if (xga_enabled) {
-                svga->xga.on = (val & 0x01) ? 0 : 1;
-                vga_on       = !svga->xga.on;
-            }
+            if (xga_enabled)
+                xga->on = (val & 0x01) ? 0 : 1;
+            if (ibm8514_enabled)
+                ibm8514_on = (val & 0x01) ? 0 : 1;
+
+            vga_on = val & 0x01;
             break;
         case 0x3c4:
             svga->seqaddr = val;
@@ -210,19 +250,15 @@ svga_out(uint16_t addr, uint8_t val, void *priv)
                     break;
             }
             break;
-        case 0x2ea:
         case 0x3c6:
             svga->dac_mask = val;
             break;
-        case 0x2eb:
-        case 0x2ec:
         case 0x3c7:
         case 0x3c8:
             svga->dac_pos    = 0;
             svga->dac_status = addr & 0x03;
             svga->dac_addr   = (val + (addr & 0x01)) & 255;
             break;
-        case 0x2ed:
         case 0x3c9:
             if (svga->adv_flags & FLAG_RAMDAC_SHIFT)
                 val <<= 2;
@@ -238,9 +274,10 @@ svga_out(uint16_t addr, uint8_t val, void *priv)
                     break;
                 case 2:
                     index                 = svga->dac_addr & 255;
+					svga->dac_b 		  = val;
                     svga->vgapal[index].r = svga->dac_r;
                     svga->vgapal[index].g = svga->dac_g;
-                    svga->vgapal[index].b = val;
+                    svga->vgapal[index].b = svga->dac_b;
                     if (svga->ramdac_type == RAMDAC_8BIT)
                         svga->pallook[index] = makecol32(svga->vgapal[index].r, svga->vgapal[index].g, svga->vgapal[index].b);
                     else
@@ -317,10 +354,51 @@ uint8_t
 svga_in(uint16_t addr, void *priv)
 {
     svga_t *svga = (svga_t *) priv;
+    ibm8514_t *dev = &svga->dev8514;
     uint8_t index;
     uint8_t ret = 0xff;
 
     switch (addr) {
+        case 0x2ea:
+            ret = dev->dac_mask;
+            break;
+        case 0x2eb:
+            ret = dev->dac_status;
+            break;
+        case 0x2ec:
+            ret = dev->dac_addr;
+            break;
+        case 0x2ed:
+            index = (dev->dac_addr - 1) & 0xff;
+            switch (dev->dac_pos) {
+                case 0:
+                    dev->dac_pos++;
+                    if (svga->ramdac_type == RAMDAC_8BIT)
+                        ret = svga->vgapal[index].r;
+                    else
+                        ret = svga->vgapal[index].r & 0x3f;
+                    break;
+                case 1:
+                    dev->dac_pos++;
+                    if (svga->ramdac_type == RAMDAC_8BIT)
+                        ret = svga->vgapal[index].g;
+                    else
+                        ret = svga->vgapal[index].g & 0x3f;
+                    break;
+                case 2:
+                    dev->dac_pos  = 0;
+                    dev->dac_addr = (dev->dac_addr + 1) & 0xff;
+                    if (svga->ramdac_type == RAMDAC_8BIT)
+                        ret = svga->vgapal[index].b;
+                    else
+                        ret = svga->vgapal[index].b & 0x3f;
+                    break;
+
+              	default:
+              		break;
+            }
+            break;
+
         case 0x3c0:
             ret = svga->attraddr | svga->attr_palette_enable;
             break;
@@ -333,25 +411,24 @@ svga_in(uint16_t addr, void *priv)
             else
                 ret = 0x10;
             break;
+        case 0x3c3:
+            ret = vga_on;
+            break;
         case 0x3c4:
             ret = svga->seqaddr;
             break;
         case 0x3c5:
             ret = svga->seqregs[svga->seqaddr & 0x0f];
             break;
-        case 0x2ea:
         case 0x3c6:
             ret = svga->dac_mask;
             break;
-        case 0x2eb:
         case 0x3c7:
             ret = svga->dac_status;
             break;
-        case 0x2ec:
         case 0x3c8:
             ret = svga->dac_addr;
             break;
-        case 0x2ed:
         case 0x3c9:
             index = (svga->dac_addr - 1) & 255;
             switch (svga->dac_pos) {
@@ -432,16 +509,27 @@ svga_in(uint16_t addr, void *priv)
 void
 svga_set_ramdac_type(svga_t *svga, int type)
 {
+    ibm8514_t *dev = &svga->dev8514;
+
     if (svga->ramdac_type != type) {
         svga->ramdac_type = type;
 
         for (int c = 0; c < 256; c++) {
-            if (svga->ramdac_type == RAMDAC_8BIT)
-                svga->pallook[c] = makecol32(svga->vgapal[c].r, svga->vgapal[c].g, svga->vgapal[c].b);
-            else
-                svga->pallook[c] = makecol32((svga->vgapal[c].r & 0x3f) * 4,
-                                             (svga->vgapal[c].g & 0x3f) * 4,
-                                             (svga->vgapal[c].b & 0x3f) * 4);
+            if (ibm8514_on) {
+                if (svga->ramdac_type == RAMDAC_8BIT)
+                    dev->pallook[c] = makecol32(svga->vgapal[c].r, svga->vgapal[c].g, svga->vgapal[c].b);
+                else
+                    dev->pallook[c] = makecol32((svga->vgapal[c].r & 0x3f) * 4,
+                                                 (svga->vgapal[c].g & 0x3f) * 4,
+                                                 (svga->vgapal[c].b & 0x3f) * 4);
+            } else {
+                if (svga->ramdac_type == RAMDAC_8BIT)
+                    svga->pallook[c] = makecol32(svga->vgapal[c].r, svga->vgapal[c].g, svga->vgapal[c].b);
+                else
+                    svga->pallook[c] = makecol32((svga->vgapal[c].r & 0x3f) * 4,
+                                                 (svga->vgapal[c].g & 0x3f) * 4,
+                                                 (svga->vgapal[c].b & 0x3f) * 4);
+            }
         }
     }
 }
@@ -449,6 +537,8 @@ svga_set_ramdac_type(svga_t *svga, int type)
 void
 svga_recalctimings(svga_t *svga)
 {
+    ibm8514_t *dev = &svga->dev8514;
+    xga_t *xga = &svga->xga;
     double crtcconst;
     double _dispontime;
     double _dispofftime;
@@ -495,7 +585,7 @@ svga_recalctimings(svga_t *svga)
 
     svga->htotal = svga->crtc[0];
     /* +5 has been verified by Sergi to be correct - +6 must have been an off by one error. */
-    svga->htotal += 5; /*+6 is required for Tyrian*/
+    svga->htotal += 5; /*+5 is required for Tyrian*/
 
     svga->rowoffset = svga->crtc[0x13];
 
@@ -612,23 +702,18 @@ svga_recalctimings(svga_t *svga)
     } else
         svga->monitor->mon_overscan_x = 16;
 
-    if (vga_on) {
-        if (svga->recalctimings_ex) {
-            svga->recalctimings_ex(svga);
-        }
-    } else {
-        if (ibm8514_enabled) {
-            if (svga->dev8514.local) {
-                if (svga->recalctimings_ex) {
-                    svga->recalctimings_ex(svga);
-                }
-            } else
-                ibm8514_recalctimings(svga);
-        }
-        if (xga_enabled)
-            xga_recalctimings(svga);
+    if (svga->recalctimings_ex) {
+        svga->recalctimings_ex(svga);
     }
 
+    if (ibm8514_enabled) {
+        if (!dev->local)
+            ibm8514_recalctimings(svga);
+    }
+
+    if (xga_enabled)
+        xga_recalctimings(svga);
+
     if (svga->hdisp >= 2048)
         svga->monitor->mon_overscan_x = 0;
 
@@ -640,9 +725,9 @@ svga_recalctimings(svga_t *svga)
 
     crtcconst = svga->clock * svga->char_width;
 
-    if (ibm8514_on && !svga->dev8514.local) {
-        disptime    = svga->dev8514.h_total;
-        _dispontime = svga->dev8514.h_disp;
+    if (ibm8514_on) {
+        disptime    = dev->h_total;
+        _dispontime = dev->h_disp;
     } else {
         disptime    = svga->htotal;
         _dispontime = svga->hdisp_time;
@@ -727,21 +812,19 @@ svga_poll(void *priv)
 {
     svga_t    *svga = (svga_t *) priv;
     ibm8514_t *dev  = &svga->dev8514;
+   	xga_t     *xga  = &svga->xga;
     uint32_t   x;
     uint32_t   blink_delay;
     int        wx;
     int        wy;
     int        ret;
     int        old_ma;
-    int        linecountff = 0;
 
-    if (!vga_on && ibm8514_enabled && ibm8514_on) {
-        if (!dev->local) {
-            ibm8514_poll(dev, svga);
-            return;
-        }
-    } else if (!vga_on && xga_enabled && svga->xga.on) {
-        xga_poll(&svga->xga, svga);
+    if (ibm8514_enabled && ibm8514_on) {
+        ibm8514_poll(dev, svga);
+        return;
+    } else if (xga_enabled && xga->on) {
+        xga_poll(xga, svga);
         return;
     }
 
@@ -835,14 +918,8 @@ svga_poll(void *priv)
         if ((svga->sc == (svga->crtc[11] & 31)) || (svga->sc == svga->rowcount))
             svga->con = 0;
         if (svga->dispon) {
-            /*Real IBM 8514/A or compatibility mode doesn't have linedbl, so skip those.*/
-            if (dev->local && ibm8514_on) {
-                svga->linedbl     = 0;
-                svga->linecountff = 0;
-                linecountff       = 1;
-            }
-            if (svga->linedbl && !svga->linecountff && !linecountff) {
-                svga->linecountff = 1;
+            if (svga->linedbl && !svga->linecountff) {
+            	svga->linecountff = 1;
                 svga->ma          = svga->maback;
             } else if (svga->sc == svga->rowcount) {
                 svga->linecountff = 0;
@@ -949,17 +1026,11 @@ svga_poll(void *priv)
             svga->monitor->mon_changeframecount = svga->interlace ? 3 : 2;
             svga->vslines                       = 0;
 
-            if ((dev->local && vga_on) || !dev->local) {
-                if (svga->interlace && svga->oddeven)
-                    svga->ma = svga->maback = svga->ma_latch + (svga->rowoffset << 1) + ((svga->crtc[5] & 0x60) >> 5);
-                else
-                    svga->ma = svga->maback = svga->ma_latch + ((svga->crtc[5] & 0x60) >> 5);
-            } else if (dev->local && ibm8514_on) {
-                if (svga->interlace && svga->oddeven)
-                    svga->ma = svga->maback = svga->ma_latch + (svga->rowoffset << 1);
-                else
-                    svga->ma = svga->maback = svga->ma_latch;
-            }
+            if (svga->interlace && svga->oddeven)
+                svga->ma = svga->maback = svga->ma_latch + (svga->rowoffset << 1) + ((svga->crtc[5] & 0x60) >> 5);
+            else
+                svga->ma = svga->maback = svga->ma_latch + ((svga->crtc[5] & 0x60) >> 5);
+
             svga->ca     = ((svga->crtc[0xe] << 8) | svga->crtc[0xf]) + ((svga->crtc[0xb] & 0x60) >> 5) + svga->ca_adj;
             svga->ma     = (svga->ma << 2);
             svga->maback = (svga->maback << 2);
@@ -974,24 +1045,22 @@ svga_poll(void *priv)
             svga->dispon   = 1;
             svga->displine = (svga->interlace && svga->oddeven) ? 1 : 0;
 
-            if (!ibm8514_on) {
-                svga->scrollcache = (svga->attrregs[0x13] & 0x0f);
-                if (!(svga->gdcreg[6] & 1) && !(svga->attrregs[0x10] & 1)) { /*Text mode*/
-                    if (svga->seqregs[1] & 1)
-                        svga->scrollcache &= 0x07;
-                    else {
-                        svga->scrollcache++;
-                        if (svga->scrollcache > 8)
-                            svga->scrollcache = 0;
-                    }
-                } else if ((svga->render == svga_render_2bpp_lowres) || (svga->render == svga_render_2bpp_highres) || (svga->render == svga_render_4bpp_lowres) || (svga->render == svga_render_4bpp_highres))
+            svga->scrollcache = (svga->attrregs[0x13] & 0x0f);
+            if (!(svga->gdcreg[6] & 1) && !(svga->attrregs[0x10] & 1)) { /*Text mode*/
+                if (svga->seqregs[1] & 1)
                     svga->scrollcache &= 0x07;
-                else
-                    svga->scrollcache = (svga->scrollcache & 0x06) >> 1;
+                else {
+                    svga->scrollcache++;
+                    if (svga->scrollcache > 8)
+                        svga->scrollcache = 0;
+                }
+            } else if ((svga->render == svga_render_2bpp_lowres) || (svga->render == svga_render_2bpp_highres) || (svga->render == svga_render_4bpp_lowres) || (svga->render == svga_render_4bpp_highres))
+                svga->scrollcache &= 0x07;
+            else
+                svga->scrollcache = (svga->scrollcache & 0x06) >> 1;
 
-                if ((svga->seqregs[1] & 8) || (svga->render == svga_render_8bpp_lowres))
-                    svga->scrollcache <<= 1;
-            }
+            if ((svga->seqregs[1] & 8) || (svga->render == svga_render_8bpp_lowres))
+                svga->scrollcache <<= 1;
 
             svga->x_add = (svga->monitor->mon_overscan_x >> 1) - svga->scrollcache;
 
@@ -1152,7 +1221,7 @@ static __inline void
 svga_write_common(uint32_t addr, uint8_t val, uint8_t linear, void *priv)
 {
     svga_t *svga = (svga_t *) priv;
-
+    xga_t  *xga  = &svga->xga;
     int     writemask2 = svga->writemask;
     int     reset_wm   = 0;
     latch_t vall;
@@ -1167,23 +1236,23 @@ svga_write_common(uint32_t addr, uint8_t val, uint8_t linear, void *priv)
 
     if (!linear) {
         if (xga_enabled) {
-            if (((svga->xga.op_mode & 7) >= 4) && (svga->xga.aperture_cntl >= 1)) {
+            if (((xga->op_mode & 7) >= 4) && (xga->aperture_cntl >= 1)) {
                 if (val == 0xa5) { /*Memory size test of XGA*/
-                    svga->xga.test    = val;
-                    svga->xga.a5_test = 1;
+                    xga->test    = val;
+                    xga->a5_test = 1;
                     return;
                 } else if (val == 0x5a) {
-                    svga->xga.test = val;
+                    xga->test = val;
                     return;
                 } else if ((val == 0x12) || (val == 0x34)) {
-                    addr += svga->xga.write_bank;
-                    svga->xga.vram[addr & svga->xga.vram_mask] = val;
-                    svga->xga.linear_endian_reverse            = 1;
+                    addr += xga->write_bank;
+                    xga->vram[addr & xga->vram_mask] = val;
+                    xga->linear_endian_reverse            = 1;
                     return;
                 }
             } else {
-                svga->xga.on = 0;
-                vga_on       = !svga->xga.on;
+                xga->on = 0;
+                vga_on = 1;
             }
         }
         addr = svga_decode_addr(svga, addr, 1);
@@ -1368,6 +1437,7 @@ static __inline uint8_t
 svga_read_common(uint32_t addr, uint8_t linear, void *priv)
 {
     svga_t  *svga       = (svga_t *) priv;
+    xga_t  *xga         = &svga->xga;
     uint32_t latch_addr = 0;
     int      readplane  = svga->readplane;
     uint8_t  count;
@@ -1381,22 +1451,22 @@ svga_read_common(uint32_t addr, uint8_t linear, void *priv)
 
     if (!linear) {
         if (xga_enabled) {
-            if (((svga->xga.op_mode & 7) >= 4) && (svga->xga.aperture_cntl >= 1)) {
-                if (svga->xga.test == 0xa5) { /*Memory size test of XGA*/
-                    svga->xga.on = 1;
-                    vga_on       = !svga->xga.on;
-                    return svga->xga.test;
-                } else if (svga->xga.test == 0x5a) {
-                    svga->xga.on = 1;
-                    vga_on       = !svga->xga.on;
-                    return svga->xga.test;
+            if (((xga->op_mode & 7) >= 4) && (xga->aperture_cntl >= 1)) {
+                if (xga->test == 0xa5) { /*Memory size test of XGA*/
+                    xga->on = 1;
+                    vga_on = 0;
+                    return xga->test;
+                } else if (xga->test == 0x5a) {
+                    xga->on = 1;
+                    vga_on = 0;
+                    return xga->test;
                 } else if ((addr == 0xa0000) || (addr == 0xa0010)) {
-                    addr += svga->xga.read_bank;
-                    return svga->xga.vram[addr & svga->xga.vram_mask];
+                    addr += xga->read_bank;
+                    return xga->vram[addr & xga->vram_mask];
                 }
             } else {
-                svga->xga.on = 0;
-                vga_on       = !svga->xga.on;
+                xga->on = 0;
+                vga_on = 1;
             }
         }
         addr = svga_decode_addr(svga, addr, 0);
@@ -1617,7 +1687,7 @@ svga_writeb_linear(uint32_t addr, uint8_t val, void *priv)
         return;
     addr &= svga->vram_mask;
     svga->changedvram[addr >> 12] = svga->monitor->mon_changeframecount;
-    *&svga->vram[addr]            = val;
+    svga->vram[addr]              = val;
 }
 
 void
@@ -1752,7 +1822,7 @@ svga_readb_linear(uint32_t addr, void *priv)
     if (addr >= svga->vram_max)
         return 0xff;
 
-    return *&svga->vram[addr & svga->vram_mask];
+    return svga->vram[addr & svga->vram_mask];
 }
 
 uint16_t
diff --git a/src/video/vid_table.c b/src/video/vid_table.c
index 6bd18c91a..dcc95b4b0 100644
--- a/src/video/vid_table.c
+++ b/src/video/vid_table.c
@@ -154,6 +154,7 @@ video_cards[] = {
     { &vga_device                                    },
     { &v7_vga_1024i_device                           },
     { &wy700_device                                  },
+    { &mach32_mca_device,       VIDEO_FLAG_TYPE_8514 },
     { &gd5426_mca_device                             },
     { &gd5428_mca_device                             },
     { &et4000_mca_device                             },
@@ -343,9 +344,9 @@ video_reset(int card)
     monitor_index_global = 0;
     loadfont("roms/video/mda/mda.rom", 0);
 
-    if (!(card == VID_NONE)
+    if ((card != VID_NONE)
         && !machine_has_flags(machine, MACHINE_VIDEO_ONLY)
-        && gfxcard[1] != 0
+        && (gfxcard[1] != 0)
         && device_is_valid(video_card_getdevice(gfxcard[1]), machine)) {
         video_monitor_init(1);
         monitor_index_global = 1;
@@ -354,7 +355,7 @@ video_reset(int card)
     }
 
     /* Do not initialize internal cards here. */
-    if (!(card == VID_NONE) && !(card == VID_INTERNAL) && !machine_has_flags(machine, MACHINE_VIDEO_ONLY)) {
+    if ((card != VID_NONE) && (card != VID_INTERNAL) && !machine_has_flags(machine, MACHINE_VIDEO_ONLY)) {
         vid_table_log("VIDEO: initializing '%s'\n", video_cards[card].device->name);
 
         video_prepare();
@@ -369,14 +370,24 @@ video_reset(int card)
 void
 video_post_reset(void)
 {
-    if (gfxcard[0] != VID_NONE) {
-        if (ibm8514_enabled) {
+    if (gfxcard[0] == VID_INTERNAL)
+        ibm8514_has_vga = (video_get_type_monitor(0) == VIDEO_FLAG_TYPE_8514);
+    else if (gfxcard[0] != VID_NONE)
+        ibm8514_has_vga = (video_card_get_flags(gfxcard[0]) == VIDEO_FLAG_TYPE_8514);
+    else
+        ibm8514_has_vga = 0;
+
+    if (ibm8514_has_vga)
+        ibm8514_enabled = 1;
+
+    if (ibm8514_enabled) {
+        if (!ibm8514_has_vga)
             ibm8514_device_add();
-        }
-        if (xga_enabled)
-            xga_device_add();
     }
 
+    if (xga_enabled)
+        xga_device_add();
+
     /* Reset the graphics card (or do nothing if it was already done
        by the machine's init function). */
     video_reset(gfxcard[0]);
diff --git a/src/video/vid_tgui9440.c b/src/video/vid_tgui9440.c
index e5f605a9d..a1fbe09e5 100644
--- a/src/video/vid_tgui9440.c
+++ b/src/video/vid_tgui9440.c
@@ -6,7 +6,7 @@
  *
  *          This file is part of the 86Box distribution.
  *
- *          Trident TGUI9400CXi and TGUI9440 emulation.
+ *          Trident TGUI9400CXi and TGUI9440/96x0 emulation.
  *
  *          TGUI9400CXi has extended write modes, controlled by extended
  *          GDC registers :
@@ -223,11 +223,10 @@ tgui_update_irqs(tgui_t *tgui)
     if (!tgui->pci)
         return;
 
-    if (!(tgui->oldctrl1 & 0x40)) {
+    if (!(tgui->oldctrl1 & 0x40))
         pci_set_irq(tgui->pci_slot, PCI_INTA, &tgui->irq_state);
-    } else {
+    else
         pci_clear_irq(tgui->pci_slot, PCI_INTA, &tgui->irq_state);
-    }
 }
 
 static void
@@ -500,7 +499,7 @@ tgui_out(uint16_t addr, uint8_t val, void *priv)
                         svga->fullchange = 3;
                         svga->ma_latch   = ((svga->crtc[0xc] << 8) | svga->crtc[0xd]) + ((svga->crtc[8] & 0x60) >> 5);
                     } else {
-                        svga->fullchange = changeframecount;
+                        svga->fullchange = svga->monitor->mon_changeframecount;
                         svga_recalctimings(svga);
                     }
                 }
@@ -629,7 +628,6 @@ void
 tgui_recalctimings(svga_t *svga)
 {
     tgui_t *tgui       = (tgui_t *) svga->priv;
-    uint8_t ger22lower = tgui->accel.ger22 & 0xff;
     uint8_t ger22upper = (tgui->accel.ger22 >> 8);
 
     if (!svga->rowoffset)
@@ -638,13 +636,8 @@ tgui_recalctimings(svga_t *svga)
     if (svga->crtc[0x29] & 0x10)
         svga->rowoffset |= 0x100;
 
-    if ((tgui->type >= TGUI_9440) && (svga->bpp >= 24)) {
-        if ((tgui->accel.bpp == 0) && (ger22lower != 14) && (svga->bpp == 24))
-            svga->hdisp = (svga->crtc[1] + 1) * 8;
-        if ((tgui->accel.bpp == 3) && (ger22lower == 14) && (svga->bpp == 32) && (tgui->type == TGUI_9440))
-            svga->rowoffset <<= 1;
-        // pclog("Accelbpp = %d, ger22lower = %02x, ger22upper = %02x, bpp = %d, rowoffset = %d.\n", tgui->accel.bpp, ger22lower, ger22upper, svga->bpp, svga->rowoffset);
-    }
+    if ((tgui->type >= TGUI_9440) && (svga->bpp >= 24))
+        svga->hdisp = (svga->crtc[1] + 1) * 8;
 
     if ((svga->crtc[0x1e] & 0xA0) == 0xA0)
         svga->ma_latch |= 0x10000;
@@ -750,14 +743,14 @@ tgui_recalctimings(svga_t *svga)
                     else if ((svga->dispend == 600) && (svga->hdisp == 800) && svga->interlace)
                         svga->hdisp = 1600;
 
+                    if (ger22upper & 0x80) {
+                        svga->htotal <<= 1;
+                        svga->hdisp <<= 1;
+                        svga->hdisp_time <<= 1;
+                    }
                     switch (svga->hdisp) {
                         case 640:
-                            if (ger22upper & 0x01)
-                                svga->rowoffset = 0x50;
-                            break;
-                        case 1600:
-                            if (svga->rowoffset != 0x100)
-                                svga->rowoffset = 0x100;
+                            svga->rowoffset = 80;
                             break;
                     }
                 }
@@ -780,9 +773,8 @@ tgui_recalctimings(svga_t *svga)
             case 32:
                 svga->render = svga_render_32bpp_highres;
                 if (tgui->type >= TGUI_9660) {
-                    if (svga->hdisp == 1024) {
+                    if (svga->hdisp == 1024)
                         svga->rowoffset <<= 1;
-                    }
                 }
                 break;
         }
@@ -1079,15 +1071,20 @@ tgui_ext_linear_read(uint32_t addr, void *priv)
     if (addr >= svga->vram_max)
         return 0xff;
 
-    addr &= ~0xf;
+    addr &= svga->vram_mask;
+    addr &= (tgui->ext_gdc_regs[0] & EXT_CTRL_LATCH_COPY) ? ~0x0f : ~0x07;
     addr = dword_remap(svga, addr);
 
-    for (uint8_t c = 0; c < 16; c++) {
-        tgui->copy_latch[c] = svga->vram[addr + c];
-        addr += ((c & 3) == 3) ? 13 : 1;
+    if (tgui->ext_gdc_regs[0] & EXT_CTRL_LATCH_COPY) {
+        for (int c = 0; c < 16; c++) {
+            tgui->copy_latch[c] = svga->vram[addr];
+            addr += (c & 3) ? 1 : 13;
+            addr &= svga->vram_mask;
+        }
+        return svga->vram[addr];
     }
 
-    return svga->vram[addr & svga->vram_mask];
+    return svga_read_linear(addr, svga);
 }
 
 static uint8_t
@@ -1105,7 +1102,8 @@ tgui_ext_linear_write(uint32_t addr, uint8_t val, void *priv)
 {
     svga_t *svga = (svga_t *) priv;
     tgui_t *tgui = (tgui_t *) svga->priv;
-    int     c;
+    int c;
+    int bpp = (tgui->ext_gdc_regs[0] & EXT_CTRL_16BIT);
     uint8_t fg[2] = { tgui->ext_gdc_regs[4], tgui->ext_gdc_regs[5] };
     uint8_t bg[2] = { tgui->ext_gdc_regs[1], tgui->ext_gdc_regs[2] };
     uint8_t mask  = tgui->ext_gdc_regs[7];
@@ -1116,62 +1114,53 @@ tgui_ext_linear_write(uint32_t addr, uint8_t val, void *priv)
     if (addr >= svga->vram_max)
         return;
     addr &= svga->vram_mask;
-    addr &= (tgui->ext_gdc_regs[0] & 8) ? ~0xf : ~0x7;
+    addr &= (tgui->ext_gdc_regs[0] & EXT_CTRL_LATCH_COPY) ? ~0x0f : ~0x07;
 
     addr                          = dword_remap(svga, addr);
-    svga->changedvram[addr >> 12] = changeframecount;
+    svga->changedvram[addr >> 12] = svga->monitor->mon_changeframecount;
 
-    switch (tgui->ext_gdc_regs[0] & 0xf) {
-        /*8-bit mono->colour expansion, unmasked*/
-        case 2:
-            for (c = 7; c >= 0; c--) {
-                if (mask & (1 << c))
-                    *(uint8_t *) &svga->vram[addr] = (val & (1 << c)) ? fg[0] : bg[0];
-                addr += (c == 4) ? 13 : 1;
+    if (tgui->ext_gdc_regs[0] & EXT_CTRL_LATCH_COPY) {
+        for (c = 0; c < 16; c++) {
+            svga->vram[addr] = tgui->copy_latch[c];
+            addr += ((c & 3) == 3) ? 13 : 1;
+            addr &= svga->vram_mask;
+        }
+    } else if (tgui->ext_gdc_regs[0] & (EXT_CTRL_MONO_EXPANSION | EXT_CTRL_MONO_TRANSPARENT)) {
+        if (tgui->ext_gdc_regs[0] & EXT_CTRL_MONO_TRANSPARENT) {
+            if (bpp) {
+                for (c = 7; c >= 0; c--) {
+                    if ((val & mask) & (1 << c))
+                        svga->vram[addr] = fg[(c & 1) ^ 1];
+                    addr += (c & 3) ? 1 : 13;
+                    addr &= svga->vram_mask;
+                }
+            } else {
+                for (c = 7; c >= 0; c--) {
+                    if ((val & mask) & (1 << c))
+                        svga->vram[addr] = tgui->ext_gdc_regs[4];
+                    addr += (c == 4) ? 13 : 1;
+                    addr &= svga->vram_mask;
+                }
             }
-            break;
-
-        /*16-bit mono->colour expansion, unmasked*/
-        case 3:
-            for (c = 7; c >= 0; c--) {
-                if (mask & (1 << c))
-                    *(uint8_t *) &svga->vram[addr] = (val & (1 << c)) ? fg[(c & 1) ^ 1] : bg[(c & 1) ^ 1];
-                addr += (c == 4) ? 13 : 1;
+        } else {
+            if (bpp) {
+                for (c = 7; c >= 0; c--) {
+                    if (mask & (1 << c))
+                        svga->vram[addr] = (val & (1 << c)) ? fg[(c & 1) ^ 1] : bg[(c & 1) ^ 1];
+                    addr += (c & 3) ? 1 : 13;
+                    addr &= svga->vram_mask;
+                }
+            } else {
+                for (c = 7; c >= 0; c--) {
+                    if (mask & (1 << c))
+                        svga->vram[addr] = (val & (1 << c)) ? tgui->ext_gdc_regs[4] : tgui->ext_gdc_regs[1];
+                    addr += (c == 4) ? 13 : 1;
+                    addr &= svga->vram_mask;
+                }
             }
-            break;
-
-        /*8-bit mono->colour expansion, masked*/
-        case 6:
-            for (c = 7; c >= 0; c--) {
-                if ((val & mask) & (1 << c))
-                    *(uint8_t *) &svga->vram[addr] = fg[0];
-                addr += (c == 4) ? 13 : 1;
-            }
-            break;
-
-        /*16-bit mono->colour expansion, masked*/
-        case 7:
-            for (c = 7; c >= 0; c--) {
-                if ((val & mask) & (1 << c))
-                    *(uint8_t *) &svga->vram[addr] = fg[(c & 1) ^ 1];
-                addr += (c == 4) ? 13 : 1;
-            }
-            break;
-
-        case 0x8:
-        case 0x9:
-        case 0xa:
-        case 0xb:
-        case 0xc:
-        case 0xd:
-        case 0xe:
-        case 0xf:
-            for (c = 0; c < 16; c++) {
-                *(uint8_t *) &svga->vram[addr] = tgui->copy_latch[c];
-                addr += ((c & 3) == 3) ? 13 : 1;
-            }
-            break;
-    }
+        }
+    } else
+        svga_write_linear(addr, val, svga);
 }
 
 static void
@@ -1179,7 +1168,8 @@ tgui_ext_linear_writew(uint32_t addr, uint16_t val, void *priv)
 {
     svga_t  *svga = (svga_t *) priv;
     tgui_t  *tgui = (tgui_t *) svga->priv;
-    int      c;
+    int  c;
+    int bpp = (tgui->ext_gdc_regs[0] & EXT_CTRL_16BIT);
     uint8_t  fg[2] = { tgui->ext_gdc_regs[4], tgui->ext_gdc_regs[5] };
     uint8_t  bg[2] = { tgui->ext_gdc_regs[1], tgui->ext_gdc_regs[2] };
     uint16_t mask  = (tgui->ext_gdc_regs[7] << 8) | tgui->ext_gdc_regs[8];
@@ -1190,70 +1180,81 @@ tgui_ext_linear_writew(uint32_t addr, uint16_t val, void *priv)
     if (addr >= svga->vram_max)
         return;
     addr &= svga->vram_mask;
-    addr &= ~0xf;
+    addr &= (tgui->ext_gdc_regs[0] & EXT_CTRL_LATCH_COPY) ? ~0x0f : ~0x07;
 
     addr                          = dword_remap(svga, addr);
-    svga->changedvram[addr >> 12] = changeframecount;
+    svga->changedvram[addr >> 12] = svga->monitor->mon_changeframecount;
 
     val = (val >> 8) | (val << 8);
 
-    switch (tgui->ext_gdc_regs[0] & 0xf) {
-        /*8-bit mono->colour expansion, unmasked*/
-        case 2:
-            for (c = 15; c >= 0; c--) {
-                if (mask & (1 << c))
-                    *(uint8_t *) &svga->vram[addr] = (val & (1 << c)) ? fg[0] : bg[0];
-                addr += (c & 3) ? 1 : 13;
-            }
-            break;
+    if (tgui->ext_gdc_regs[0] & EXT_CTRL_LATCH_COPY) {
+        for (c = 0; c < 16; c++) {
+            svga->vram[addr] = tgui->copy_latch[c];
+            addr += (c & 3) ? 1 : 13;
+            addr &= svga->vram_mask;
+        }
+    } else if (tgui->ext_gdc_regs[0] & (EXT_CTRL_MONO_EXPANSION | EXT_CTRL_MONO_TRANSPARENT)) {
+        if (tgui->ext_gdc_regs[0] & EXT_CTRL_MONO_TRANSPARENT) {
+            if (bpp) {
+                for (c = 15; c >= 0; c--) {
+                    if ((val & mask) & (1 << c))
+                        svga->vram[addr] = fg[(c & 1) ^ 1];
+                    addr += (c & 3) ? 1 : 13;
+                    addr &= svga->vram_mask;
+                }
+            } else {
+                for (c = 15; c >= 0; c--) {
+                    if ((val & mask) & (1 << c))
+                        svga->vram[addr] = tgui->ext_gdc_regs[4];
 
-        /*16-bit mono->colour expansion, unmasked*/
-        case 3:
-            for (c = 15; c >= 0; c--) {
-                if (mask & (1 << c))
-                    *(uint8_t *) &svga->vram[addr] = (val & (1 << c)) ? fg[(c & 1) ^ 1] : bg[(c & 1) ^ 1];
-                addr += (c & 3) ? 1 : 13;
+                    addr += (c & 3) ? 1 : 13;
+                    addr &= svga->vram_mask;
+                }
             }
-            break;
+        } else {
+            if (bpp) {
+                for (c = 15; c >= 0; c--) {
+                    if (mask & (1 << c))
+                       svga->vram[addr] = (val & (1 << c)) ? fg[(c & 1) ^ 1] : bg[(c & 1) ^ 1];
+                    addr += (c & 3) ? 1 : 13;
+                    addr &= svga->vram_mask;
+                }
+            } else {
+                for (c = 15; c >= 0; c--) {
+                    if (mask & (1 << c))
+                        svga->vram[addr] = (val & (1 << c)) ? tgui->ext_gdc_regs[4] : tgui->ext_gdc_regs[1];
 
-        /*8-bit mono->colour expansion, masked*/
-        case 6:
-            for (c = 15; c >= 0; c--) {
-                if ((val & mask) & (1 << c))
-                    *(uint8_t *) &svga->vram[addr] = fg[0];
-                addr += (c & 3) ? 1 : 13;
+                    addr += (c & 3) ? 1 : 13;
+                    addr &= svga->vram_mask;
+                }
             }
-            break;
-
-        /*16-bit mono->colour expansion, masked*/
-        case 7:
-            for (c = 15; c >= 0; c--) {
-                if ((val & mask) & (1 << c))
-                    *(uint8_t *) &svga->vram[addr] = fg[(c & 1) ^ 1];
-                addr += (c & 3) ? 1 : 13;
-            }
-            break;
-
-        case 0x8:
-        case 0x9:
-        case 0xa:
-        case 0xb:
-        case 0xc:
-        case 0xd:
-        case 0xe:
-        case 0xf:
-            for (c = 0; c < 16; c++) {
-                *(uint8_t *) &svga->vram[addr + c] = tgui->copy_latch[c];
-                addr += ((c & 3) == 3) ? 13 : 1;
-            }
-            break;
-    }
+        }
+    } else
+        svga_writew_linear(addr, val, svga);
 }
 
 static void
 tgui_ext_linear_writel(uint32_t addr, uint32_t val, void *priv)
 {
-    tgui_ext_linear_writew(addr, val, priv);
+    svga_t  *svga = (svga_t *) priv;
+    tgui_t  *tgui = (tgui_t *) svga->priv;
+    cycles -= video_timing_write_l;
+
+    addr &= svga->decode_mask;
+    if (addr >= svga->vram_max)
+        return;
+    addr &= svga->vram_mask;
+    addr &= (tgui->ext_gdc_regs[0] & EXT_CTRL_LATCH_COPY) ? ~0x0f : ~0x07;
+
+    addr                          = dword_remap(svga, addr);
+    svga->changedvram[addr >> 12] = svga->monitor->mon_changeframecount;
+
+    if (tgui->ext_gdc_regs[0] & (EXT_CTRL_MONO_EXPANSION | EXT_CTRL_MONO_TRANSPARENT | EXT_CTRL_LATCH_COPY)) {
+        tgui_ext_linear_writew(addr, val & 0xffff, priv);
+        tgui_ext_linear_writew(addr + 2, val >> 16, priv);
+    } else {
+        svga_writel_linear(addr, val, svga);
+    }
 }
 
 static void
@@ -1261,7 +1262,7 @@ tgui_ext_write(uint32_t addr, uint8_t val, void *priv)
 {
     svga_t *svga = (svga_t *) priv;
 
-    addr = (addr & svga->banked_mask) + svga->read_bank;
+    addr = (addr & svga->banked_mask) + svga->write_bank;
 
     tgui_ext_linear_write(addr, val, svga);
 }
@@ -1270,7 +1271,7 @@ tgui_ext_writew(uint32_t addr, uint16_t val, void *priv)
 {
     svga_t *svga = (svga_t *) priv;
 
-    addr = (addr & svga->banked_mask) + svga->read_bank;
+    addr = (addr & svga->banked_mask) + svga->write_bank;
 
     tgui_ext_linear_writew(addr, val, svga);
 }
@@ -1279,7 +1280,7 @@ tgui_ext_writel(uint32_t addr, uint32_t val, void *priv)
 {
     svga_t *svga = (svga_t *) priv;
 
-    addr = (addr & svga->banked_mask) + svga->read_bank;
+    addr = (addr & svga->banked_mask) + svga->write_bank;
 
     tgui_ext_linear_writel(addr, val, svga);
 }
@@ -1329,13 +1330,13 @@ enum {
 #define WRITE(addr, dat)                                                               \
     if (tgui->accel.bpp == 0) {                                                        \
         svga->vram[(addr) &tgui->vram_mask]                   = dat;                   \
-        svga->changedvram[((addr) & (tgui->vram_mask)) >> 12] = changeframecount;      \
+        svga->changedvram[((addr) & (tgui->vram_mask)) >> 12] = svga->monitor->mon_changeframecount;      \
     } else if (tgui->accel.bpp == 1) {                                                 \
         vram_w[(addr) & (tgui->vram_mask >> 1)]                    = dat;              \
-        svga->changedvram[((addr) & (tgui->vram_mask >> 1)) >> 11] = changeframecount; \
+        svga->changedvram[((addr) & (tgui->vram_mask >> 1)) >> 11] = svga->monitor->mon_changeframecount; \
     } else {                                                                           \
         vram_l[(addr) & (tgui->vram_mask >> 2)]                    = dat;              \
-        svga->changedvram[((addr) & (tgui->vram_mask >> 2)) >> 10] = changeframecount; \
+        svga->changedvram[((addr) & (tgui->vram_mask >> 2)) >> 10] = svga->monitor->mon_changeframecount; \
     }
 
 static void
@@ -1365,7 +1366,7 @@ tgui_accel_command(int count, uint32_t cpu_dat, tgui_t *tgui)
         trans_col &= 0xffff;
     }
 
-    if (count != -1 && !tgui->accel.x && (tgui->accel.flags & TGUI_SRCMONO)) {
+    if ((count != -1) && !tgui->accel.x && (tgui->accel.flags & TGUI_SRCMONO)) {
         count -= (tgui->accel.flags >> 24) & 7;
         cpu_dat <<= (tgui->accel.flags >> 24) & 7;
     }
@@ -1413,38 +1414,6 @@ tgui_accel_command(int count, uint32_t cpu_dat, tgui_t *tgui)
         }
     }
 
-    switch (svga->hdisp) {
-        case 640:
-        case 1024:
-        case 1280:
-            tgui->accel.pitch = svga->hdisp;
-            break;
-        case 800: /*Disassembly of the TGUI9440/96x0 drivers shows that 800x600 is treated as 832 in the acceleration pitch (0x340 as horizontal display)*/
-            tgui->accel.pitch = svga->hdisp + 32;
-            break;
-        case 1600:
-            tgui->accel.pitch = 2048;
-            break;
-    }
-
-    switch (ger22lower) {
-        case 4: /*8-bit mode for modes up to 1024x768.*/
-        case 9: /*15-bit and 16-bit modes.*/
-            if (!(ger22upper & 0x01)) {
-                if (ger22upper == 0x00)
-                    tgui->accel.pitch = 1024;
-            }
-            break;
-        case 8: /*8-bit mode for modes greater than 1024x768 and 24-bit mode for 640x480 (latter is TGUI9440AGi only).*/
-            if (!(ger22upper & 0x01)) {
-                if (ger22upper == 0x00) {
-                    if (svga->bpp == 24)
-                        tgui->accel.pitch = 2048;
-                }
-            }
-            break;
-    }
-
     // pclog("TGUI accel command = %x, ger22 = %04x, hdisp = %d, dispend = %d, vtotal = %d, rowoffset = %d, svgabpp = %d, interlace = %d, accelbpp = %d, pitch = %d.\n", tgui->accel.command, tgui->accel.ger22, svga->hdisp, svga->dispend, svga->vtotal, svga->rowoffset, svga->bpp, svga->interlace, tgui->accel.bpp, tgui->accel.pitch);
 
     switch (tgui->accel.command) {
@@ -2002,49 +1971,40 @@ static void
 tgui_accel_out(uint16_t addr, uint8_t val, void *priv)
 {
     tgui_t *tgui = (tgui_t *) priv;
+    svga_t *svga = &tgui->svga;
 
     switch (addr) {
         case 0x2122:
             tgui->accel.ger22 = (tgui->accel.ger22 & 0xff00) | val;
-            switch (val & 0xff) {
-                case 4:
+            tgui->accel.pitch = 0x200 << ((val >> 2) & 3);
+            switch (svga->bpp) {
                 case 8:
+                case 24:
                     tgui->accel.bpp = 0;
                     break;
-
-                case 9:
-                    switch (tgui->svga.bpp) {
-                        case 32:
-                            tgui->accel.bpp = 3;
-                            break;
-                        default:
-                            tgui->accel.bpp = 1;
-                            break;
-                    }
+                case 15:
+                case 16:
+                    tgui->accel.bpp = 1;
                     break;
-
-                case 13:
-                case 14:
-                    switch (tgui->svga.bpp) {
-                        case 15:
-                        case 16:
-                            tgui->accel.bpp = 1;
-                            break;
-
-                        case 24:
-                            tgui->accel.bpp = 0;
-                            break;
-
-                        case 32:
-                            tgui->accel.bpp = 3;
-                            break;
-                    }
+                case 32:
+                    tgui->accel.bpp = 3;
                     break;
             }
             break;
 
         case 0x2123:
             tgui->accel.ger22 = (tgui->accel.ger22 & 0xff) | (val << 8);
+            if ((val & 0x80) || (((val & 0xc0) == 0x40)))
+                tgui->accel.pitch = svga->rowoffset << 3;
+            else if (tgui->accel.pitch <= 1024)
+                tgui->accel.pitch = svga->rowoffset << 3;
+
+            if (tgui->accel.bpp == 1)
+                tgui->accel.pitch >>= 1;
+            else if (tgui->accel.bpp == 3)
+                tgui->accel.pitch >>= 2;
+
+            svga_recalctimings(svga);
             break;
 
         case 0x2124: /*Command*/
@@ -2656,45 +2616,35 @@ tgui_accel_write(uint32_t addr, uint8_t val, void *priv)
     switch (addr & 0xff) {
         case 0x22:
             tgui->accel.ger22 = (tgui->accel.ger22 & 0xff00) | val;
-            switch (val & 0xff) {
-                case 4:
+            tgui->accel.pitch = 0x200 << ((val >> 2) & 3);
+            switch (svga->bpp) {
                 case 8:
+                case 24:
                     tgui->accel.bpp = 0;
                     break;
-
-                case 9:
-                    switch (tgui->svga.bpp) {
-                        case 32:
-                            tgui->accel.bpp = 3;
-                            break;
-                        default:
-                            tgui->accel.bpp = 1;
-                            break;
-                    }
+                case 15:
+                case 16:
+                    tgui->accel.bpp = 1;
                     break;
-
-                case 13:
-                case 14:
-                    switch (tgui->svga.bpp) {
-                        case 15:
-                        case 16:
-                            tgui->accel.bpp = 1;
-                            break;
-
-                        case 24:
-                            tgui->accel.bpp = 0;
-                            break;
-
-                        case 32:
-                            tgui->accel.bpp = 3;
-                            break;
-                    }
+                case 32:
+                    tgui->accel.bpp = 3;
                     break;
             }
             break;
 
         case 0x23:
             tgui->accel.ger22 = (tgui->accel.ger22 & 0xff) | (val << 8);
+            if ((val & 0x80) || (((val & 0xc0) == 0x40)))
+                tgui->accel.pitch = svga->rowoffset << 3;
+            else if (tgui->accel.pitch <= 1024)
+                tgui->accel.pitch = svga->rowoffset << 3;
+
+            if (tgui->accel.bpp == 1)
+                tgui->accel.pitch >>= 1;
+            else if (tgui->accel.bpp == 3)
+                tgui->accel.pitch >>= 2;
+
+            svga_recalctimings(svga);
             break;
 
         case 0x24: /*Command*/
@@ -3516,6 +3466,9 @@ tgui_init(const device_t *info)
     mem_mapping_disable(&tgui->accel_mapping);
     mem_mapping_disable(&tgui->mmio_mapping);
 
+    if (tgui->vram_size == (2 << 20))
+        svga->crtc[0x21] |= 0x10;
+
     tgui_set_io(tgui);
 
     if (tgui->pci && (tgui->type >= TGUI_9440)) {
@@ -3589,7 +3542,7 @@ tgui_force_redraw(void *priv)
 {
     tgui_t *tgui = (tgui_t *) priv;
 
-    tgui->svga.fullchange = changeframecount;
+    tgui->svga.fullchange = tgui->svga.monitor->mon_changeframecount;
 }
 
 // clang-format off
diff --git a/src/win/Makefile.mingw b/src/win/Makefile.mingw
index b000a7dd3..1b06ff8f5 100644
--- a/src/win/Makefile.mingw
+++ b/src/win/Makefile.mingw
@@ -738,6 +738,7 @@ VIDOBJ := agpgart.o video.o \
           vid_ati_eeprom.o \
           vid_ati18800.o vid_ati28800.o \
 	  vid_ati_mach8.o \
+          vid_ati68875_ramdac.o \
           vid_ati_mach64.o vid_ati68860_ramdac.o \
           vid_bt48x_ramdac.o \
           vid_av9194.o vid_icd2061.o vid_ics2494.o vid_ics2595.o \