diff --git a/headers/private/graphics/radeon/radeon_interface.h b/headers/private/graphics/radeon/radeon_interface.h
index e3edf7115b..3402fb77f7 100644
--- a/headers/private/graphics/radeon/radeon_interface.h
+++ b/headers/private/graphics/radeon/radeon_interface.h
@@ -20,7 +20,8 @@
 #include "ddc.h"
 
 // magic code for ioctls
-#define RADEON_PRIVATE_DATA_MAGIC	'TKRA'
+// changed from TKRA to TKR1 for RADEON_WAITFORFIFO ioctl
+#define RADEON_PRIVATE_DATA_MAGIC	'TKR1'
 
 #define MAX_RADEON_DEVICE_NAME_LENGTH MAXPATHLEN
 
@@ -35,6 +36,7 @@ enum {
 	RADEON_FREE_MEM,
 	
 	RADEON_WAITFORIDLE,
+	RADEON_WAITFORFIFO,
 	RADEON_RESETENGINE,
 	RADEON_VIPREAD,
 	RADEON_VIPWRITE,
@@ -396,6 +398,7 @@ typedef struct {
 typedef struct {
 	// filled out by kernel
 	CP_info	cp;				// info concerning command processor
+	bool	acc_dma;			// prevent use of dma engine
 	
 	// set by accelerant
 	struct {
@@ -499,6 +502,12 @@ typedef struct {
 	bool			keep_lock;	// keep lock after engine is idle
 } radeon_wait_for_idle;
 
+// wait for idle
+typedef struct {
+	uint32 			magic;
+	int				entries;	// keep lock after engine is idle
+} radeon_wait_for_fifo;
+
 // read VIP register
 typedef struct {
 	uint32 			magic;
diff --git a/headers/private/graphics/radeon/version.h b/headers/private/graphics/radeon/version.h
index 5abaa41fbe..70f526a862 100644
--- a/headers/private/graphics/radeon/version.h
+++ b/headers/private/graphics/radeon/version.h
@@ -8,4 +8,4 @@
 */
 
 // current version
-#define RADEON_DRIVER_VERSION "Version: 5.1.2.1"
+#define RADEON_DRIVER_VERSION "Version: 5.1.3.1"
diff --git a/src/add-ons/accelerants/radeon/Acceleration.c b/src/add-ons/accelerants/radeon/Acceleration.c
index 1d01288769..8c118a1999 100644
--- a/src/add-ons/accelerants/radeon/Acceleration.c
+++ b/src/add-ons/accelerants/radeon/Acceleration.c
@@ -13,6 +13,7 @@
 #include "generic.h"
 #include "3d_regs.h"
 #include "2d_regs.h"
+#include "radeon_regs.h"
 #include "mmio.h"
 #include "CP.h"
 
@@ -21,7 +22,7 @@
 //	et - ignored
 //	list - list of rectangles
 //	count - number of rectangles
-void SCREEN_TO_SCREEN_BLIT(engine_token *et, blit_params *list, uint32 count) 
+void SCREEN_TO_SCREEN_BLIT_DMA(engine_token *et, blit_params *list, uint32 count) 
 {
 	virtual_card *vc = ai->vc;
 		
@@ -55,13 +56,63 @@ void SCREEN_TO_SCREEN_BLIT(engine_token *et, blit_params *list, uint32 count)
 	++ai->si->engine.count;
 }
 
+#define SRC_DSTCOLOR			0x00030000
+#define DP_SRC_RECT				0x00000200
+
+void SCREEN_TO_SCREEN_BLIT_PIO(engine_token *et, blit_params *list, uint32 count)
+{
+	int xdir;
+	int ydir;
+	virtual_card *vc = ai->vc;
+	
+	SHOW_FLOW0( 4, "" );
+	
+	Radeon_WaitForFifo ( ai , 1 );
+
+	// Setup for Screen to screen blit
+	OUTREG(ai->regs, RADEON_DP_GUI_MASTER_CNTL, (vc->datatype << RADEON_GMC_DST_DATATYPE_SHIFT
+								 | RADEON_GMC_BRUSH_NONE
+								 | RADEON_GMC_SRC_DATATYPE_COLOR
+								 | RADEON_ROP3_S
+								 | RADEON_DP_SRC_SOURCE_MEMORY
+								 | RADEON_GMC_SRC_PITCH_OFFSET_CNTL));
+	
+   
+	for( ; count > 0; --count, ++list ) {
+		
+		// make sure there is space in the FIFO for 4 register writes
+		Radeon_WaitForFifo ( ai , 4 );
+	
+		xdir = ((list->src_left < list->dest_left) && (list->src_top == list->dest_top)) ? -1 : 1;
+		ydir = (list->src_top < list->dest_top) ? -1 : 1;
+
+		if (xdir < 0) list->src_left += list->width , list->dest_left += list->width ;
+    	if (ydir < 0) list->src_top += list->height , list->dest_top += list->height ;
+    
+		OUTREG(ai->regs, RADEON_DP_CNTL, ((xdir >= 0 ? RADEON_DST_X_LEFT_TO_RIGHT : 0) 
+										| (ydir >= 0 ? RADEON_DST_Y_TOP_TO_BOTTOM : 0)));
+ 		
+		// Tell the engine where the source data resides.
+		OUTREG( ai->regs, RADEON_SRC_Y_X, (list->src_top << 16 ) | list->src_left);
+		OUTREG( ai->regs, RADEON_DST_Y_X, (list->dest_top << 16 ) | list->dest_left);
+
+		// this is the blt initiator.
+		OUTREG( ai->regs, RADEON_DST_HEIGHT_WIDTH, ((list->height + 1) << 16 ) | (list->width + 1));
+    
+	}
+		
+	++ai->si->engine.count;
+	et = et;
+
+}
+
 
 // fill rectangles on screen
 //	et - ignored
 //	colorIndex - fill colour
 //	list - list of rectangles
 //	count - number of rectangles
-void FILL_RECTANGLE(engine_token *et, uint32 colorIndex, 
+void FILL_RECTANGLE_DMA(engine_token *et, uint32 colorIndex, 
 	fill_rect_params *list, uint32 count) 
 {
 	virtual_card *vc = ai->vc;
@@ -98,11 +149,47 @@ void FILL_RECTANGLE(engine_token *et, uint32 colorIndex,
 }
 
 
+// fill rectangles on screen
+//	et - ignored
+//	colorIndex - fill colour
+//	list - list of rectangles
+//	count - number of rectangles
+#define BRUSH_SOLIDCOLOR		0x00000d00
+
+void FILL_RECTANGLE_PIO(engine_token *et, uint32 colorIndex, fill_rect_params *list, uint32 count) 
+{
+	virtual_card *vc = ai->vc;
+			
+	SHOW_FLOW( 4, "colorIndex", colorIndex);
+	
+	Radeon_WaitForFifo(ai, 3);
+    OUTREG(ai->regs, RADEON_DP_GUI_MASTER_CNTL, ((vc->datatype << RADEON_GMC_DST_DATATYPE_SHIFT)
+									 | RADEON_GMC_BRUSH_SOLID_COLOR
+									 | RADEON_GMC_SRC_DATATYPE_COLOR
+									 | RADEON_ROP3_P));
+	// Set brush colour								 
+    OUTREG(ai->regs, RADEON_DP_BRUSH_FRGD_CLR,  colorIndex);
+    OUTREG(ai->regs, RADEON_DP_CNTL, (RADEON_DST_X_LEFT_TO_RIGHT | RADEON_DST_Y_TOP_TO_BOTTOM));
+
+	for( ; count > 0; --count, ++list ) 
+	{
+		
+		Radeon_WaitForFifo(ai, 2);
+		OUTREG(ai->regs, RADEON_DST_Y_X,          (list->top << 16) | list->left);
+		OUTREG(ai->regs, RADEON_DST_WIDTH_HEIGHT, ((list->right - list->left + 1) << 16) | (list->bottom - list->top + 1));
+
+	}
+	++ai->si->engine.count;
+	et = et;
+}
+
+
+
 // invert rectangle on screen
 //	et - ignored
 //	list - list of rectangles
 //	count - number of rectangles
-void INVERT_RECTANGLE(engine_token *et, fill_rect_params *list, uint32 count) 
+void INVERT_RECTANGLE_DMA(engine_token *et, fill_rect_params *list, uint32 count) 
 {
 	virtual_card *vc = ai->vc;
 	
@@ -151,12 +238,41 @@ void INVERT_RECTANGLE(engine_token *et, fill_rect_params *list, uint32 count)
 	++ai->si->engine.count;
 }
 
+
+void INVERT_RECTANGLE_PIO(engine_token *et, fill_rect_params *list, uint32 count) 
+{
+	virtual_card *vc = ai->vc;
+	
+	SHOW_FLOW0( 4, "" );
+	
+	Radeon_WaitForFifo(ai, 3);
+    OUTREG(ai->regs, RADEON_DP_GUI_MASTER_CNTL, ((vc->datatype << RADEON_GMC_DST_DATATYPE_SHIFT)
+									 | RADEON_GMC_BRUSH_NONE
+									 | RADEON_GMC_SRC_DATATYPE_COLOR
+									 | RADEON_ROP3_Dn
+									 | RADEON_DP_SRC_SOURCE_MEMORY));
+
+    OUTREG(ai->regs, RADEON_DP_CNTL, (RADEON_DST_X_LEFT_TO_RIGHT | RADEON_DST_Y_TOP_TO_BOTTOM));
+
+	for( ; count > 0; --count, ++list ) 
+	{
+		
+		Radeon_WaitForFifo(ai, 2);
+		OUTREG(ai->regs, RADEON_DST_Y_X,          (list->top << 16) | list->left);
+		OUTREG(ai->regs, RADEON_DST_WIDTH_HEIGHT, ((list->right - list->left + 1) << 16) | (list->bottom - list->top + 1));
+	}
+		
+	++ai->si->engine.count;
+	et = et;
+	
+}
+
 // fill horizontal spans on screen
 //	et - ignored
 //	colorIndex - fill colour
 //	list - list of spans
 //	count - number of spans
-void FILL_SPAN(engine_token *et, uint32 colorIndex, uint16 *list, uint32 count) 
+void FILL_SPAN_DMA(engine_token *et, uint32 colorIndex, uint16 *list, uint32 count) 
 {
 	virtual_card *vc = ai->vc;
 	
@@ -196,72 +312,181 @@ void FILL_SPAN(engine_token *et, uint32 colorIndex, uint16 *list, uint32 count)
 }
 
 
+// fill horizontal spans on screen
+//	et - ignored
+//	colorIndex - fill colour
+//	list - list of spans
+//	count - number of spans
+void FILL_SPAN_PIO(engine_token *et, uint32 colorIndex, uint16 *list, uint32 count) 
+{
+
+	virtual_card *vc = ai->vc;
+	//int offset = 0;
+	uint16 y, x, width;
+	
+	SHOW_FLOW0( 4, "" );
+	
+	Radeon_WaitForFifo( ai , 1);
+    OUTREG( ai->regs, RADEON_DP_GUI_MASTER_CNTL, 0
+									 | RADEON_GMC_BRUSH_SOLID_COLOR
+									 | (vc->datatype << RADEON_GMC_DST_DATATYPE_SHIFT)
+									 | RADEON_GMC_SRC_DATATYPE_COLOR
+									 | RADEON_ROP3_P);
+	
+	if ( ai->si->asic >= rt_rv200 ) {
+		Radeon_WaitForFifo( ai , 1);
+		OUTREG( ai->regs, RADEON_DST_LINE_PATCOUNT, 0x55 << RADEON_BRES_CNTL_SHIFT);
+	}
+
+	Radeon_WaitForFifo( ai , 1);
+	OUTREG( ai->regs, RADEON_DP_BRUSH_FRGD_CLR,  colorIndex);	
+								   
+	for( ; count > 0; --count ) {
+		
+		Radeon_WaitForFifo( ai , 2);
+		  			
+		y = *list++;
+		x = *list++;
+		width = *list++ - x + 1;
+
+	    OUTREG( ai->regs, RADEON_DST_LINE_START,           (y << 16) | x);
+    	OUTREG( ai->regs, RADEON_DST_LINE_END,             ((y) << 16) | (x + width));
+		
+	}
+		
+	++ai->si->engine.count;
+	et = et;
+}
+
+
+void SCREEN_TO_SCREEN_BLIT(engine_token *et, blit_params *list, uint32 count)
+{
+	if ( ai->si->acc_dma )
+		SCREEN_TO_SCREEN_BLIT_DMA(et,list,count);
+	else
+		SCREEN_TO_SCREEN_BLIT_PIO(et,list,count);
+}
+
+void FILL_RECTANGLE(engine_token *et, uint32 color, fill_rect_params *list, uint32 count)
+{
+	if ( ai->si->acc_dma )
+		FILL_RECTANGLE_DMA(et, color, list, count);
+	else
+		FILL_RECTANGLE_PIO(et, color, list, count);
+}
+
+void INVERT_RECTANGLE(engine_token *et, fill_rect_params *list, uint32 count)
+{
+	if ( ai->si->acc_dma )
+		INVERT_RECTANGLE_DMA(et, list, count);
+	else
+		INVERT_RECTANGLE_PIO(et, list, count);
+}
+
+void FILL_SPAN(engine_token *et, uint32 color, uint16 *list, uint32 count)
+{
+	if ( ai->si->acc_dma )
+		FILL_SPAN_DMA(et, color, list, count);
+	else
+		FILL_SPAN_PIO(et, color, list, count);
+
+}
+
+
 // prepare 2D acceleration
 void Radeon_Init2D( accelerator_info *ai )
 {
 	SHOW_FLOW0( 3, "" );
 
-	START_IB();
-	
 	// forget about 3D	
-	WRITE_IB_REG( RADEON_RB3D_CNTL, 0 );
-		
-	SUBMIT_IB();
+	if ( ai->si->acc_dma ) {
+		START_IB();
+		WRITE_IB_REG( RADEON_RB3D_CNTL, 0 );
+		SUBMIT_IB();
+	}
+	else {
+		OUTREG( ai->regs, RADEON_RB3D_CNTL, 0 );
+	}
 }
 
-
 // fill state buffer that sets 2D registers up for accelerated operations
 void Radeon_FillStateBuffer( accelerator_info *ai, uint32 datatype )
 {
 	virtual_card *vc = ai->vc;
 	uint32 pitch_offset;
-	uint32 *buffer, *buffer_start;
+	uint32 *buffer = NULL, *buffer_start = NULL;
 	
 	SHOW_FLOW0( 4, "" );
 
-	// make sure buffer is not used
-	Radeon_InvalidateStateBuffer( ai, vc->state_buffer_idx );
-	
-	buffer = buffer_start = Radeon_GetIndirectBufferPtr( ai, vc->state_buffer_idx );
-
 	// set offset of frame buffer and pitch
 	pitch_offset = 
 		((ai->si->memory[mt_local].virtual_addr_start + vc->fb_offset) >> 10) | 
 		((vc->pitch >> 6) << 22);
-	WRITE_IB_REG( RADEON_DEFAULT_OFFSET, pitch_offset );
-	WRITE_IB_REG( RADEON_DST_PITCH_OFFSET, pitch_offset );
-	WRITE_IB_REG( RADEON_SRC_PITCH_OFFSET, pitch_offset );
-	
-	// no siccors	
-	WRITE_IB_REG( RADEON_DEFAULT_SC_BOTTOM_RIGHT, 
-		(RADEON_DEFAULT_SC_RIGHT_MAX | RADEON_DEFAULT_SC_BOTTOM_MAX));
-
-	// setup general flags - perhaps this is not needed as all
-	// 2D commands contain this register
-	WRITE_IB_REG( RADEON_DP_GUI_MASTER_CNTL, 
-		(datatype << RADEON_GMC_DST_DATATYPE_SHIFT)
-		| RADEON_GMC_CLR_CMP_CNTL_DIS
 		
-		| RADEON_GMC_BRUSH_SOLID_COLOR 
-		| RADEON_GMC_SRC_DATATYPE_COLOR
-		
-		| RADEON_ROP3_P
-		| RADEON_DP_SRC_SOURCE_MEMORY
-		| RADEON_GMC_WR_MSK_DIS );
-	    
+	if ( ai->si->acc_dma ) {
+		// make sure buffer is not used
+		Radeon_InvalidateStateBuffer( ai, vc->state_buffer_idx );
+		buffer = buffer_start = Radeon_GetIndirectBufferPtr( ai, vc->state_buffer_idx );
 	
-	// most of this init is probably not nessacary
-	// as we neither draw lines nor use brushes
-	WRITE_IB_REG( RADEON_DST_LINE_START,    0);
-	WRITE_IB_REG( RADEON_DST_LINE_END,      0);
-	WRITE_IB_REG( RADEON_DP_BRUSH_FRGD_CLR, 0xffffffff);
-	WRITE_IB_REG( RADEON_DP_BRUSH_BKGD_CLR, 0x00000000);
-	WRITE_IB_REG( RADEON_DP_SRC_FRGD_CLR,   0xffffffff);
-	WRITE_IB_REG( RADEON_DP_SRC_BKGD_CLR,   0x00000000);
-	WRITE_IB_REG( RADEON_DP_WRITE_MASK,     0xffffffff);
+		WRITE_IB_REG( RADEON_DEFAULT_OFFSET, pitch_offset );
+		WRITE_IB_REG( RADEON_DST_PITCH_OFFSET, pitch_offset );
+		WRITE_IB_REG( RADEON_SRC_PITCH_OFFSET, pitch_offset );
+		// no sissors	
+		WRITE_IB_REG( RADEON_DEFAULT_SC_BOTTOM_RIGHT, 
+			(RADEON_DEFAULT_SC_RIGHT_MAX | RADEON_DEFAULT_SC_BOTTOM_MAX));
 
+		// general fluff
+		WRITE_IB_REG( RADEON_DP_GUI_MASTER_CNTL, 
+			(datatype << RADEON_GMC_DST_DATATYPE_SHIFT)
+			| RADEON_GMC_CLR_CMP_CNTL_DIS
+			
+			| RADEON_GMC_BRUSH_SOLID_COLOR 
+			| RADEON_GMC_SRC_DATATYPE_COLOR
+			
+			| RADEON_ROP3_P
+			| RADEON_DP_SRC_SOURCE_MEMORY
+			| RADEON_GMC_WR_MSK_DIS );
+		
+		// most of this init is probably not necessary
+		// as we neither draw lines nor use brushes
+		WRITE_IB_REG( RADEON_DP_BRUSH_FRGD_CLR, 0xffffffff);
+		WRITE_IB_REG( RADEON_DP_BRUSH_BKGD_CLR, 0x00000000);
+		WRITE_IB_REG( RADEON_DP_SRC_FRGD_CLR,   0xffffffff);
+		WRITE_IB_REG( RADEON_DP_SRC_BKGD_CLR,   0x00000000);
+		WRITE_IB_REG( RADEON_DP_WRITE_MASK,     0xffffffff);
+	
+		// this is required
+		vc->state_buffer_size = buffer - buffer_start;
 
-	vc->state_buffer_size = buffer - buffer_start;
+	} else {
+		Radeon_WaitForFifo( ai, 10 );
+		OUTREG( ai->regs, RADEON_DEFAULT_OFFSET, pitch_offset );
+		OUTREG( ai->regs, RADEON_DST_PITCH_OFFSET, pitch_offset );
+		OUTREG( ai->regs, RADEON_SRC_PITCH_OFFSET, pitch_offset );
+		// no sissors	
+		OUTREG( ai->regs, RADEON_DEFAULT_SC_BOTTOM_RIGHT, (RADEON_DEFAULT_SC_RIGHT_MAX
+		    | RADEON_DEFAULT_SC_BOTTOM_MAX));
+		// general fluff
+		OUTREG( ai->regs, RADEON_DP_GUI_MASTER_CNTL, 
+			(datatype << RADEON_GMC_DST_DATATYPE_SHIFT)
+			| RADEON_GMC_CLR_CMP_CNTL_DIS
+			
+			| RADEON_GMC_BRUSH_SOLID_COLOR 
+			| RADEON_GMC_SRC_DATATYPE_COLOR
+			
+			| RADEON_ROP3_P
+			| RADEON_DP_SRC_SOURCE_MEMORY
+			| RADEON_GMC_WR_MSK_DIS );
+		
+		// most of this init is probably not necessary
+		// as we neither draw lines nor use brushes
+		OUTREG( ai->regs, RADEON_DP_BRUSH_FRGD_CLR, 0xffffffff);
+		OUTREG( ai->regs, RADEON_DP_BRUSH_BKGD_CLR, 0x00000000);
+		OUTREG( ai->regs, RADEON_DP_SRC_FRGD_CLR,   0xffffffff);
+		OUTREG( ai->regs, RADEON_DP_SRC_BKGD_CLR,   0x00000000);
+		OUTREG( ai->regs, RADEON_DP_WRITE_MASK,     0xffffffff);
+
+	}
 	
 	ai->si->active_vc = vc->id;
 }
diff --git a/src/add-ons/accelerants/radeon/EngineManagment.c b/src/add-ons/accelerants/radeon/EngineManagment.c
index c572ad4d85..03811dcf52 100644
--- a/src/add-ons/accelerants/radeon/EngineManagment.c
+++ b/src/add-ons/accelerants/radeon/EngineManagment.c
@@ -65,21 +65,30 @@ static void writeSyncToken( accelerator_info *ai )
 	if( ai->si->engine.count == ai->si->engine.written )
 		return;
 
-	START_IB();
-
-	// flush pending data
-	WRITE_IB_REG( RADEON_RB2D_DSTCACHE_CTLSTAT, RADEON_RB2D_DC_FLUSH_ALL );
+	if( ai->si->acc_dma ) {
+		START_IB();
 	
-	// make sure commands are finished
-	WRITE_IB_REG( RADEON_WAIT_UNTIL, RADEON_WAIT_2D_IDLECLEAN |
-		RADEON_WAIT_3D_IDLECLEAN | RADEON_WAIT_HOST_IDLECLEAN );
+		// flush pending data
+		WRITE_IB_REG( RADEON_RB2D_DSTCACHE_CTLSTAT, RADEON_RB2D_DC_FLUSH_ALL );
 		
-	// write scratch register
-	WRITE_IB_REG( RADEON_SCRATCH_REG0, ai->si->engine.count );
-	
-	ai->si->engine.written = ai->si->engine.count;
-	
-	SUBMIT_IB();
+		// make sure commands are finished
+		WRITE_IB_REG( RADEON_WAIT_UNTIL, RADEON_WAIT_2D_IDLECLEAN |
+			RADEON_WAIT_3D_IDLECLEAN | RADEON_WAIT_HOST_IDLECLEAN );
+			
+		// write scratch register
+		WRITE_IB_REG( RADEON_SCRATCH_REG0, ai->si->engine.count );
+		
+		ai->si->engine.written = ai->si->engine.count;
+		
+		SUBMIT_IB();
+	} else {
+		Radeon_WaitForFifo( ai, 2 );
+		OUTREG( ai->regs, RADEON_RB2D_DSTCACHE_CTLSTAT, RADEON_RB2D_DC_FLUSH_ALL);
+		OUTREG( ai->regs, RADEON_WAIT_UNTIL, RADEON_WAIT_2D_IDLECLEAN |
+		   RADEON_WAIT_3D_IDLECLEAN |
+		   RADEON_WAIT_HOST_IDLECLEAN);
+		ai->si->engine.written = ai->si->engine.count;
+	}
 }
 
 // public function: acquire engine for future use
@@ -178,6 +187,13 @@ status_t SYNC_TO_TOKEN( sync_token *st )
 	
 	SHOW_FLOW0( 4, "" );
 	
+	if ( !ai->si->acc_dma )
+	{
+		Radeon_WaitForFifo( ai, 64 );
+		Radeon_WaitForIdle( ai, false );
+		return B_OK;
+	}
+	
 	start_time = system_time();
 
 	while( 1 ) {
diff --git a/src/add-ons/accelerants/radeon/InitAccelerant.c b/src/add-ons/accelerants/radeon/InitAccelerant.c
index a8d242646a..a8f8059d1d 100644
--- a/src/add-ons/accelerants/radeon/InitAccelerant.c
+++ b/src/add-ons/accelerants/radeon/InitAccelerant.c
@@ -111,7 +111,8 @@ err:
 static void uninit_common( void )
 {
 	if( !ai->accelerant_is_clone ) {
-		Radeon_FreeVirtualCardStateBuffer( ai );
+		if ( ai->si->acc_dma ) 
+			Radeon_FreeVirtualCardStateBuffer( ai );
 
 		// the last accelerant should must wait for the card to become quite,
 		// else some nasty command could lunger in some FIFO
@@ -200,7 +201,8 @@ status_t INIT_ACCELERANT( int the_fd )
 	// mark engine as having no state
 	//si->cp.active_state_buffer = -1;
 	
-	Radeon_AllocateVirtualCardStateBuffer( ai );
+	if ( ai->si->acc_dma )
+		Radeon_AllocateVirtualCardStateBuffer( ai );
 
 	// everything else is initialized upon set_display_mode
 	return B_OK;
diff --git a/src/add-ons/accelerants/radeon/driver_wrapper.c b/src/add-ons/accelerants/radeon/driver_wrapper.c
index c89eebc560..6c793f68de 100644
--- a/src/add-ons/accelerants/radeon/driver_wrapper.c
+++ b/src/add-ons/accelerants/radeon/driver_wrapper.c
@@ -23,6 +23,27 @@ status_t Radeon_WaitForIdle( accelerator_info *ai, bool keep_lock )
 	return ioctl( ai->fd, RADEON_WAITFORIDLE, &wfi, sizeof( wfi ));
 }
 
+// wait until "entries" FIFO entries are empty
+// lock must be hold
+status_t Radeon_WaitForFifo( accelerator_info *ai, int entries )
+{
+	while( 1 ) {
+		bigtime_t start_time = system_time();
+	
+		do {
+			int slots = INREG( ai->regs, RADEON_RBBM_STATUS ) & RADEON_RBBM_FIFOCNT_MASK;
+			
+			if ( slots >= entries ) 
+				return B_OK;
+
+			snooze( 1 );
+		} while( system_time() - start_time < 1000000 );
+			
+		Radeon_ResetEngine( ai );
+	}
+	
+	return B_ERROR;
+}
 
 void Radeon_ResetEngine( accelerator_info *ai )
 {
diff --git a/src/add-ons/accelerants/radeon/generic.h b/src/add-ons/accelerants/radeon/generic.h
index 0509c956b7..afc31a4196 100644
--- a/src/add-ons/accelerants/radeon/generic.h
+++ b/src/add-ons/accelerants/radeon/generic.h
@@ -49,6 +49,18 @@ void INVERT_RECTANGLE(engine_token *et, fill_rect_params *list, uint32 count);
 
 void FILL_SPAN(engine_token *et, uint32 color, uint16 *list, uint32 count);
 
+void SCREEN_TO_SCREEN_BLIT_DMA(engine_token *et, blit_params *list, uint32 count);
+void FILL_RECTANGLE_DMA(engine_token *et, uint32 color, fill_rect_params *list, uint32 count);
+void INVERT_RECTANGLE_DMA(engine_token *et, fill_rect_params *list, uint32 count);
+
+void FILL_SPAN_DMA(engine_token *et, uint32 color, uint16 *list, uint32 count);
+
+void SCREEN_TO_SCREEN_BLIT_PIO(engine_token *et, blit_params *list, uint32 count);
+void FILL_RECTANGLE_PIO(engine_token *et, uint32 color, fill_rect_params *list, uint32 count);
+void INVERT_RECTANGLE_PIO(engine_token *et, fill_rect_params *list, uint32 count);
+
+void FILL_SPAN_PIO(engine_token *et, uint32 color, uint16 *list, uint32 count);
+
 uint32 OVERLAY_COUNT(const display_mode *dm);
 const uint32 *OVERLAY_SUPPORTED_SPACES(const display_mode *dm);
 uint32 OVERLAY_SUPPORTED_FEATURES(uint32 a_color_space);
diff --git a/src/add-ons/accelerants/radeon/overlay.c b/src/add-ons/accelerants/radeon/overlay.c
index 6134671583..ba1758bd11 100644
--- a/src/add-ons/accelerants/radeon/overlay.c
+++ b/src/add-ons/accelerants/radeon/overlay.c
@@ -973,17 +973,27 @@ static void Radeon_ReplaceOverlayBuffer(
 	shared_info *si = ai->si;
 	uint32 offset;
 	
-	START_IB();
-
-	offset = si->pending_overlay.on->mem_offset + si->active_overlay.rel_offset;
-	
-	WRITE_IB_REG( RADEON_OV0_VID_BUF0_BASE_ADRS, offset);
-	
-	si->overlay_mgr.auto_flip_reg ^= RADEON_OV0_SOFT_EOF_TOGGLE;
-	WRITE_IB_REG( RADEON_OV0_AUTO_FLIP_CNTRL, si->overlay_mgr.auto_flip_reg );
-	
-	SUBMIT_IB();
+	if ( ai->si->acc_dma )
+	{
+		START_IB();
 	
+		offset = si->pending_overlay.on->mem_offset + si->active_overlay.rel_offset;
+		
+		WRITE_IB_REG( RADEON_OV0_VID_BUF0_BASE_ADRS, offset);
+		
+		si->overlay_mgr.auto_flip_reg ^= RADEON_OV0_SOFT_EOF_TOGGLE;
+		WRITE_IB_REG( RADEON_OV0_AUTO_FLIP_CNTRL, si->overlay_mgr.auto_flip_reg );
+		
+		SUBMIT_IB();
+	} else {
+		Radeon_WaitForFifo( ai, 2 );
+		offset = si->pending_overlay.on->mem_offset + si->active_overlay.rel_offset;
+		
+		OUTREG( ai->regs, RADEON_OV0_VID_BUF0_BASE_ADRS, offset);
+		
+		si->overlay_mgr.auto_flip_reg ^= RADEON_OV0_SOFT_EOF_TOGGLE;
+		OUTREG( ai->regs, RADEON_OV0_AUTO_FLIP_CNTRL, si->overlay_mgr.auto_flip_reg );
+	}	
 	ai->si->active_overlay.on = ai->si->pending_overlay.on;
 #endif
 }
diff --git a/src/add-ons/accelerants/radeon/palette.c b/src/add-ons/accelerants/radeon/palette.c
index ed8e99a590..5aab89affe 100644
--- a/src/add-ons/accelerants/radeon/palette.c
+++ b/src/add-ons/accelerants/radeon/palette.c
@@ -13,33 +13,41 @@
 
 #include "GlobalData.h"
 #include "dac_regs.h"
+#include "mmio.h"
 #include "CP.h"
 #include "generic.h"
 
-// Radeon's DACs share same public registers, this function
-// selects the DAC you'll talk to
-#define selectPalette( crtc_idx ) \
-	WRITE_IB_REG( RADEON_DAC_CNTL2, \
-		(crtc_idx == 0 ? 0 : RADEON_DAC2_PALETTE_ACC_CTL) | \
-		(ai->si->dac_cntl2 & ~RADEON_DAC2_PALETTE_ACC_CTL) );
-
-
 // set standard colour palette (needed for non-palette modes)
 void Radeon_InitPalette( 
 	accelerator_info *ai, int crtc_idx )
 {
 	int i;
 	
-	START_IB();
-	
-	selectPalette( crtc_idx );
-	
-	WRITE_IB_REG( RADEON_PALETTE_INDEX, 0 );
-	
-	for( i = 0; i < 256; ++i )
-		WRITE_IB_REG( RADEON_PALETTE_DATA, (i << 16) | (i << 8) | i );
+	if ( ai->si->acc_dma ) {
+		START_IB();
 		
-	SUBMIT_IB();
+		WRITE_IB_REG( RADEON_DAC_CNTL2,
+			(crtc_idx == 0 ? 0 : RADEON_DAC2_PALETTE_ACC_CTL) |
+			(ai->si->dac_cntl2 & ~RADEON_DAC2_PALETTE_ACC_CTL) );
+		
+		WRITE_IB_REG( RADEON_PALETTE_INDEX, 0 );
+		
+		for( i = 0; i < 256; ++i )
+			WRITE_IB_REG( RADEON_PALETTE_DATA, (i << 16) | (i << 8) | i );
+			
+		SUBMIT_IB();
+	} else {
+		Radeon_WaitForFifo ( ai , 1 );
+		OUTREG( ai->regs, RADEON_DAC_CNTL2,
+			(crtc_idx == 0 ? 0 : RADEON_DAC2_PALETTE_ACC_CTL) |
+			(ai->si->dac_cntl2 & ~RADEON_DAC2_PALETTE_ACC_CTL) );
+
+		OUTREG( ai->regs, RADEON_PALETTE_INDEX, 0 );
+		for( i = 0; i < 256; ++i ) {
+			Radeon_WaitForFifo ( ai , 1 );  // TODO FIXME good god this is gonna be slow...
+			OUTREG( ai->regs, RADEON_PALETTE_DATA, (i << 16) | (i << 8) | i );
+		}
+	}
 }
 
 static void setPalette( 
@@ -75,17 +83,35 @@ static void setPalette(
 {
 	uint i;
 	
-	START_IB();
-	
-	selectPalette( crtc_idx );
-	
-	WRITE_IB_REG( RADEON_PALETTE_INDEX, first );
-	
-	for( i = 0; i < count; ++i, color_data += 3 )
-		WRITE_IB_REG( RADEON_PALETTE_DATA, 
-			((uint32)color_data[0] << 16) | 
-			((uint32)color_data[1] << 8) | 
-			 color_data[2] );
-			 
-	SUBMIT_IB();
+	if ( ai->si->acc_dma ) {
+		START_IB();
+		
+		WRITE_IB_REG( RADEON_DAC_CNTL2,
+			(crtc_idx == 0 ? 0 : RADEON_DAC2_PALETTE_ACC_CTL) |
+			(ai->si->dac_cntl2 & ~RADEON_DAC2_PALETTE_ACC_CTL) );
+		
+		WRITE_IB_REG( RADEON_PALETTE_INDEX, first );
+		
+		for( i = 0; i < count; ++i, color_data += 3 )
+			WRITE_IB_REG( RADEON_PALETTE_DATA, 
+				((uint32)color_data[0] << 16) | 
+				((uint32)color_data[1] << 8) | 
+				 color_data[2] );
+				 
+		SUBMIT_IB();
+	} else {
+		Radeon_WaitForFifo ( ai , 1 );
+		OUTREG( ai->regs, RADEON_DAC_CNTL2,
+			(crtc_idx == 0 ? 0 : RADEON_DAC2_PALETTE_ACC_CTL) |
+			(ai->si->dac_cntl2 & ~RADEON_DAC2_PALETTE_ACC_CTL) );
+
+		OUTREG( ai->regs, RADEON_PALETTE_INDEX, first );
+		for( i = 0; i < count; ++i, color_data += 3 ) {
+			Radeon_WaitForFifo ( ai , 1 );  // TODO FIXME good god this is gonna be slow...
+			OUTREG( ai->regs, RADEON_PALETTE_DATA, 
+				((uint32)color_data[0] << 16) | 
+				((uint32)color_data[1] << 8) | 
+				 color_data[2] );
+		}
+	}
 }
diff --git a/src/add-ons/accelerants/radeon/radeon_accelerant.h b/src/add-ons/accelerants/radeon/radeon_accelerant.h
index de54181708..c69035bc24 100644
--- a/src/add-ons/accelerants/radeon/radeon_accelerant.h
+++ b/src/add-ons/accelerants/radeon/radeon_accelerant.h
@@ -115,6 +115,7 @@ void Radeon_FillStateBuffer( accelerator_info *ai, uint32 datatype );
 
 // driver_wrapper.c
 status_t Radeon_WaitForIdle( accelerator_info *ai, bool keep_lock );
+status_t Radeon_WaitForFifo( accelerator_info *ai, int entries );
 void Radeon_ResetEngine( accelerator_info *ai );
 
 status_t Radeon_VIPRead( accelerator_info *ai, uint channel, uint address, uint32 *data );
diff --git a/src/add-ons/kernel/drivers/graphics/radeon/CP_setup.c b/src/add-ons/kernel/drivers/graphics/radeon/CP_setup.c
index 6f673a0b63..760fa07f4a 100644
--- a/src/add-ons/kernel/drivers/graphics/radeon/CP_setup.c
+++ b/src/add-ons/kernel/drivers/graphics/radeon/CP_setup.c
@@ -227,71 +227,54 @@ void Radeon_ResetEngine( device_info *di )
     host_path_cntl = INREG( regs, RADEON_HOST_PATH_CNTL );
 	rbbm_soft_reset = INREG( regs, RADEON_RBBM_SOFT_RESET );
 
-	switch( di->asic ) {
-	case rt_r300:
-		OUTREG( regs, RADEON_RBBM_SOFT_RESET, (rbbm_soft_reset |
-			RADEON_SOFT_RESET_CP |
-			RADEON_SOFT_RESET_HI |
-			RADEON_SOFT_RESET_E2 |
-			RADEON_SOFT_RESET_AIC ));
-		INREG( regs, RADEON_RBBM_SOFT_RESET);
-		OUTREG( regs, RADEON_RBBM_SOFT_RESET, 0);
-		// this bit has no description
-		OUTREGP( regs, RADEON_RB2D_DSTCACHE_MODE, (1 << 17), ~0 );
-
-		break;
-	default:
-		OUTREG( regs, RADEON_RBBM_SOFT_RESET, rbbm_soft_reset |
+	OUTREG( regs, RADEON_RBBM_SOFT_RESET, (rbbm_soft_reset |
 			RADEON_SOFT_RESET_CP |
 			RADEON_SOFT_RESET_HI |
 			RADEON_SOFT_RESET_SE |
 			RADEON_SOFT_RESET_RE |
 			RADEON_SOFT_RESET_PP |
 			RADEON_SOFT_RESET_E2 |
-			RADEON_SOFT_RESET_RB |
-			RADEON_SOFT_RESET_AIC );
-		INREG( regs, RADEON_RBBM_SOFT_RESET );
-		OUTREG( regs, RADEON_RBBM_SOFT_RESET, rbbm_soft_reset &
-			~( RADEON_SOFT_RESET_CP |
-			RADEON_SOFT_RESET_HI |
-			RADEON_SOFT_RESET_SE |
-			RADEON_SOFT_RESET_RE |
-			RADEON_SOFT_RESET_PP |
-			RADEON_SOFT_RESET_E2 |
-			RADEON_SOFT_RESET_RB |
-			RADEON_SOFT_RESET_AIC ) );
-		INREG( regs, RADEON_RBBM_SOFT_RESET );
-	}
-
+			RADEON_SOFT_RESET_RB ) );
+	INREG( regs, RADEON_RBBM_SOFT_RESET);
+	OUTREG( regs, RADEON_RBBM_SOFT_RESET, rbbm_soft_reset &
+		~( RADEON_SOFT_RESET_CP |
+		   RADEON_SOFT_RESET_HI |
+		   RADEON_SOFT_RESET_SE |
+		   RADEON_SOFT_RESET_RE |
+		   RADEON_SOFT_RESET_PP |
+		   RADEON_SOFT_RESET_E2 |
+		   RADEON_SOFT_RESET_RB ) );
+	INREG( regs, RADEON_RBBM_SOFT_RESET);
+	
     OUTREG( regs, RADEON_HOST_PATH_CNTL, host_path_cntl | RADEON_HDP_SOFT_RESET );
     INREG( regs, RADEON_HOST_PATH_CNTL );
     OUTREG( regs, RADEON_HOST_PATH_CNTL, host_path_cntl );
 
-	// restore regs
-	OUTREG( regs, RADEON_RBBM_SOFT_RESET, rbbm_soft_reset);
-
-	OUTREG( regs, RADEON_CLOCK_CNTL_INDEX, clock_cntl_index );
-	//R300_PLLFix( regs, di->asic );
 	Radeon_OUTPLL( regs, di->asic, RADEON_MCLK_CNTL, mclk_cntl );
+   	OUTREG( regs, RADEON_CLOCK_CNTL_INDEX, clock_cntl_index );
+   	//RADEONPllErrataAfterIndex( regs, di->asic ); // drm doesn't do this here!
+   	OUTREG( regs, RADEON_RBBM_SOFT_RESET, rbbm_soft_reset);
+   	
+	if ( di->acc_dma )
+	{
+		// reset ring buffer
+		cur_read_ptr = INREG( regs, RADEON_CP_RB_RPTR );
+		OUTREG( regs, RADEON_CP_RB_WPTR, cur_read_ptr );
+		
+		//if( si->cp.ring.head ) {
+		// during init, there are no feedback data
+		if( si->cp.feedback.mem_handle != 0 ) {
+			*(uint32 *)MEM2CPU( si->cp.feedback.mem_type, si->cp.feedback.head_mem_offset) = 
+				cur_read_ptr;
+			//	*si->cp.ring.head = cur_read_ptr;
+			si->cp.ring.tail = cur_read_ptr;
+		}
 	
-	// reset ring buffer
-	cur_read_ptr = INREG( regs, RADEON_CP_RB_RPTR );
-	OUTREG( regs, RADEON_CP_RB_WPTR, cur_read_ptr );
-	
-	//if( si->cp.ring.head ) {
-	// during init, there are no feedback data
-	if( si->cp.feedback.mem_handle != 0 ) {
-		*(uint32 *)MEM2CPU( si->cp.feedback.mem_type, si->cp.feedback.head_mem_offset) = 
-			cur_read_ptr;
-		//	*si->cp.ring.head = cur_read_ptr;
-		si->cp.ring.tail = cur_read_ptr;
+		// mark all buffers as being finished
+		Radeon_DiscardAllIndirectBuffers( di );
 	}
 
 	++si->engine.count;
-	
-	// mark all buffers as being finished
-	Radeon_DiscardAllIndirectBuffers( di );
-
 	return;
 }
 
@@ -522,7 +505,7 @@ status_t Radeon_InitCP( device_info *di )
     set_sem_owner( di->si->cp.lock.sem, thinfo.team );
 	
 	// init raw CP
-	loadMicroEngineRAMData( di );
+	if ( di->acc_dma ) loadMicroEngineRAMData( di );
 
 	// do soft-reset
 	Radeon_ResetEngine( di );
@@ -536,26 +519,30 @@ status_t Radeon_InitCP( device_info *di )
 	// reset CP to make disabling active
 	Radeon_ResetEngine( di );
 
-	res = initRingBuffer( di, CP_RING_SIZE );
-	if( res < 0 )
-		goto err4;
-	
-	res = initCPFeedback( di );
-	if( res < 0 )
-		goto err3;
+	if ( di->acc_dma )
+	{
+		res = initRingBuffer( di, CP_RING_SIZE );
+		if( res < 0 )
+			goto err4;
 		
-	res = initIndirectBuffers( di );
-	if( res < 0 )
-		goto err2;
+		res = initCPFeedback( di );
+		if( res < 0 )
+			goto err3;
+			
+		res = initIndirectBuffers( di );
+		if( res < 0 )
+			goto err2;
+			
+		// tell CP to use BM
+		Radeon_WaitForIdle( di, false, false );
 		
-	// tell CP to use BM
-	Radeon_WaitForIdle( di, false, false );
-	
-	// enable direct and indirect CP bus mastering
-	OUTREG( di->regs, RADEON_CP_CSQ_CNTL, RADEON_CSQ_PRIBM_INDBM );
-	
-	// allow bus mastering in general
-	OUTREGP( di->regs, RADEON_BUS_CNTL, 0, ~RADEON_BUS_MASTER_DIS );
+		// enable direct and indirect CP bus mastering
+		OUTREG( di->regs, RADEON_CP_CSQ_CNTL, RADEON_CSQ_PRIBM_INDBM );
+		
+		// allow bus mastering in general
+		OUTREGP( di->regs, RADEON_BUS_CNTL, 0, ~RADEON_BUS_MASTER_DIS );
+	}
+
 
 	// don't allow mixing of 2D/3D/scratch/wait_until commands
 	// (in fact, this doesn't seem to make any difference as we do a
@@ -596,11 +583,14 @@ void Radeon_UninitCP( device_info *di )
 	OUTREG( regs, RADEON_CP_CSQ_CNTL, RADEON_CSQ_PRIDIS_INDDIS );	
 	// read-back for flushing
 	INREG( regs, RADEON_CP_CSQ_CNTL );
-
-	uninitRingBuffer( di );
-	uninitCPFeedback( di );
-	uninitIndirectBuffers( di );
 	
+	if ( di->acc_dma )
+	{
+		uninitRingBuffer( di );
+		uninitCPFeedback( di );
+		uninitIndirectBuffers( di );
+	}
+		
 	DELETE_BEN( di->si->cp.lock );
 }
 
diff --git a/src/add-ons/kernel/drivers/graphics/radeon/bios.c b/src/add-ons/kernel/drivers/graphics/radeon/bios.c
index 5394ab7746..03d24d59a1 100644
--- a/src/add-ons/kernel/drivers/graphics/radeon/bios.c
+++ b/src/add-ons/kernel/drivers/graphics/radeon/bios.c
@@ -536,58 +536,142 @@ static void Radeon_GetFPData( device_info *di )
 	SHOW_INFO( 2, "pixel_clock=%d", di->fp_info.dot_clock );
 }
 
+
+// Depending on card genertation, chipset bugs, etc... the amount of vram
+// accessible to the CPU can vary. This function is our best shot at figuring
+// it out. Returns a value in KB.
+static uint32 RADEON_GetAccessibleVRAM( device_info *di )
+{
+	vuint8 *regs = di->regs;
+	pci_info *pcii = &(di->pcii);
+
+    uint32 aper_size = INREG( regs, RADEON_CONFIG_APER_SIZE );
+
+    // Set HDP_APER_CNTL only on cards that are known not to be broken,
+    // that is has the 2nd generation multifunction PCI interface
+    if (di->asic == rt_rv280 ||
+		di->asic == rt_rv350 ||
+		di->asic == rt_rv380 ||
+		di->asic == rt_r420  ) {
+			OUTREGP( regs, RADEON_HOST_PATH_CNTL, RADEON_HDP_APER_CNTL,
+		     ~RADEON_HDP_APER_CNTL);
+	    SHOW_INFO0( 0, "Generation 2 PCI interface, using max accessible memory");
+	    return aper_size * 2;
+    }
+
+    // Older cards have all sorts of funny issues to deal with. First
+    // check if it's a multifunction card by reading the PCI config
+    // header type... Limit those to one aperture size
+    if (get_pci(PCI_header_type, 1) & 0x80) {
+		SHOW_INFO0( 0, "Generation 1 PCI interface in multifunction mode"
+				", accessible memory limited to one aperture\n");
+		return aper_size;
+    }
+
+    // Single function older card. We read HDP_APER_CNTL to see how the BIOS
+    // have set it up. We don't write this as it's broken on some ASICs but
+    // we expect the BIOS to have done the right thing (might be too optimistic...)
+    if (INREG( regs, RADEON_HOST_PATH_CNTL ) & RADEON_HDP_APER_CNTL )
+        return aper_size * 2;
+    
+    return aper_size;
+}
+
+
 // detect amount of graphics memory
 static void Radeon_DetectRAM( device_info *di )
 {
 	vuint8 *regs = di->regs;
-
-	if( !di->is_igp	)
-		di->local_mem_size = INREG( regs, RADEON_CONFIG_MEMSIZE ) & RADEON_CONFIG_MEMSIZE_MASK;
-	else {
+	uint32 accessible, bar_size, tmp = 0;
+	
+	if( di->is_igp	) {
 		uint32 tom;
 		
 		tom = INREG( regs, RADEON_NB_TOM );
 		di->local_mem_size = ((tom >> 16) + 1 - (tom & 0xffff)) << 16;
+		OUTREG( regs, RADEON_CONFIG_MEMSIZE, di->local_mem_size * 1024);
+	} else {
+		di->local_mem_size = INREG( regs, RADEON_CONFIG_MEMSIZE ) & RADEON_CONFIG_MEMSIZE_MASK;
 	}
 
 	// some production boards of m6 will return 0 if it's 8 MB
-	if( di->local_mem_size == 0 )
+	if( di->local_mem_size == 0 ) {
 		di->local_mem_size = 8 * 1024 *1024;
+		OUTREG( regs, RADEON_CONFIG_MEMSIZE, di->local_mem_size);
+	}
 
-	switch( INREG( regs, RADEON_MEM_SDRAM_MODE_REG ) & RADEON_MEM_CFG_TYPE_MASK ) {
-		case RADEON_MEM_CFG_SDR:
-			// SDR SGRAM (2:1)
-			strcpy(di->ram_type, "SDR SGRAM");
-			di->ram.ml = 4;
-			di->ram.MB = 4;
-			di->ram.Trcd = 1;
-			di->ram.Trp = 2;
-			di->ram.Twr = 1;
-			di->ram.CL = 2;
-			di->ram.loop_latency = 16;
-			di->ram.Rloop = 16;
-			di->ram.Tr2w = 0;			
-			break;
-			
-		case RADEON_MEM_CFG_DDR:
-			// DDR SGRAM
-			strcpy(di->ram_type, "DDR SGRAM");
-			di->ram.ml = 4;
-			di->ram.MB = 4;
-			di->ram.Trcd = 3;
-			di->ram.Trp = 3;
-			di->ram.Twr = 2;
-			di->ram.CL = 3;
-			di->ram.Tr2w = 1;
-			di->ram.loop_latency = 16;
-			di->ram.Rloop = 16;
-			break;
-			
-		// only one bit, so there's no default
+
+	// Get usable Vram, after asic bugs, configuration screw ups etc
+	accessible = RADEON_GetAccessibleVRAM( di );
+
+	// Crop it to the size of the PCI BAR 
+	bar_size = di->pcii.u.h0.base_register_sizes[0];
+	if (bar_size == 0)
+	    bar_size = 0x200000;
+	if (accessible > bar_size)
+	    accessible = bar_size;
+
+	SHOW_INFO( 0, "Detected total video RAM=%ldK, accessible=%ldK (PCI BAR=%ldK)"
+		, di->local_mem_size/1024, accessible/1024, bar_size/1024);
+	if (di->local_mem_size > accessible) 
+	    di->local_mem_size = accessible;
+
+	// detect ram bus width only used by dynamic clocks for now.
+	tmp = INREG( regs, RADEON_MEM_CNTL );
+	if (IS_DI_R300_VARIANT) {
+		tmp &=  R300_MEM_NUM_CHANNELS_MASK;
+		switch (tmp) {
+			case 0: di->ram.width = 64; break;
+			case 1: di->ram.width = 128; break;
+			case 2: di->ram.width = 256; break;
+			default: di->ram.width = 128; break;
+		}
+	} else if ( (di->asic >= rt_rv100) ||
+				(di->asic >= rt_rs100) ||
+				(di->asic >= rt_rs200)) {
+		if (tmp & RV100_HALF_MODE) 
+			di->ram.width = 32;
+		else 
+			di->ram.width = 64;
+	} else {
+		if (tmp & RADEON_MEM_NUM_CHANNELS_MASK) 
+			di->ram.width = 128;
+		else 
+			di->ram.width = 64;
+	}
+
+	if (di->is_igp || (di->asic >= rt_r300))
+	{
+		uint32 mem_type = INREG( regs, RADEON_MEM_SDRAM_MODE_REG ) & RADEON_MEM_CFG_TYPE_MASK;
+		if ( mem_type == RADEON_MEM_CFG_SDR) {
+				// SDR SGRAM (2:1)
+				strcpy(di->ram_type, "SDR SGRAM");
+				di->ram.ml = 4;
+				di->ram.MB = 4;
+				di->ram.Trcd = 1;
+				di->ram.Trp = 2;
+				di->ram.Twr = 1;
+				di->ram.CL = 2;
+				di->ram.loop_latency = 16;
+				di->ram.Rloop = 16;
+				di->ram.Tr2w = 0;			
+		} else {  // RADEON_MEM_CFG_DDR		
+				// DDR SGRAM
+				strcpy(di->ram_type, "DDR SGRAM");
+				di->ram.ml = 4;
+				di->ram.MB = 4;
+				di->ram.Trcd = 3;
+				di->ram.Trp = 3;
+				di->ram.Twr = 2;
+				di->ram.CL = 3;
+				di->ram.Tr2w = 1;
+				di->ram.loop_latency = 16;
+				di->ram.Rloop = 16;
+		}
 	}
 	
-	SHOW_INFO( 1, "%ld MB %s found", di->local_mem_size / 1024 / 1024, 
-		di->ram_type );
+	SHOW_INFO( 1, "%ld MB %s found on %d wide bus", 
+		di->local_mem_size / 1024 / 1024, di->ram_type, di->ram.width);
 		
 /*	if( di->local_mem_size > 64 * 1024 * 1024 ) {
 		di->local_mem_size = 64 * 1024 * 1024;
diff --git a/src/add-ons/kernel/drivers/graphics/radeon/detect.c b/src/add-ons/kernel/drivers/graphics/radeon/detect.c
index 9209a14e70..11379f3168 100644
--- a/src/add-ons/kernel/drivers/graphics/radeon/detect.c
+++ b/src/add-ons/kernel/drivers/graphics/radeon/detect.c
@@ -592,6 +592,60 @@ static bool probeDevice( device_info *di )
 				
 			}
 
+			// disable 2d DMA engine for chips that don't work with our 
+			// dma code (yet).  I would have used asic type, but R410s are
+			// treated same asic as r420s in code, and the AGP x800 works fine. 
+			switch ( device->device_id )
+			{
+				// rv370                   
+				case DEVICE_ID_RADEON_5b60:
+				case DEVICE_ID_RADEON_5b62:
+				case DEVICE_ID_RADEON_5b64:
+				case DEVICE_ID_RADEON_5b65:
+				case DEVICE_ID_RADEON_5460:
+				case DEVICE_ID_RADEON_5464:
+				                           
+				// rv380                   
+				case DEVICE_ID_RADEON_3e50:
+				case DEVICE_ID_RADEON_3e54:
+				case DEVICE_ID_RADEON_3150:
+				case DEVICE_ID_RADEON_3154:
+				case DEVICE_ID_RADEON_5462:
+				                           
+				// rv410                   
+				case DEVICE_ID_RADEON_5e48:
+				case DEVICE_ID_RADEON_564a:
+				case DEVICE_ID_RADEON_564b:
+				case DEVICE_ID_RADEON_564f:
+				case DEVICE_ID_RADEON_5652:
+				case DEVICE_ID_RADEON_5653:
+				case DEVICE_ID_RADEON_5e4b:
+				case DEVICE_ID_RADEON_5e4a:
+				case DEVICE_ID_RADEON_5e4d:
+				case DEVICE_ID_RADEON_5e4c:
+				case DEVICE_ID_RADEON_5e4f:
+
+				// rs400                  
+				case DEVICE_ID_RS400_5a41:
+				case DEVICE_ID_RS400_5a42:
+				                          
+				// rs410                  
+				case DEVICE_ID_RS410_5a61:
+				case DEVICE_ID_RS410_5a62:
+				                          
+				// rs480                  
+				case DEVICE_ID_RS480_5954:
+				case DEVICE_ID_RS480_5955:
+				case DEVICE_ID_RS482_5974:
+				case DEVICE_ID_RS482_5975:
+
+					di->acc_dma = false;
+					break;
+				default:
+					di->acc_dma = true;
+					break;
+			}
+			
 			if( Radeon_MapBIOS( &di->pcii, &di->rom ) != B_OK )
 				// give up checking this device - no BIOS, no fun
 				return false;
diff --git a/src/add-ons/kernel/drivers/graphics/radeon/driver.c b/src/add-ons/kernel/drivers/graphics/radeon/driver.c
index 1287bec8ef..01f628ca76 100644
--- a/src/add-ons/kernel/drivers/graphics/radeon/driver.c
+++ b/src/add-ons/kernel/drivers/graphics/radeon/driver.c
@@ -273,6 +273,16 @@ static status_t control_hook( void *dev, uint32 msg, void *buf, size_t len )
 			result = B_OK;
 		} break;
 		
+		case RADEON_WAITFORFIFO: {
+			radeon_wait_for_fifo *wff = (radeon_wait_for_fifo *)buf;
+			
+			if( wff->magic != RADEON_PRIVATE_DATA_MAGIC )
+				break;
+				
+			Radeon_WaitForFifo( di, wff->entries );
+			result = B_OK;
+		} break;
+		
 		case RADEON_RESETENGINE: {
 			radeon_no_arg *na = (radeon_no_arg *)buf;
 			
diff --git a/src/add-ons/kernel/drivers/graphics/radeon/init.c b/src/add-ons/kernel/drivers/graphics/radeon/init.c
index 7aac99b57f..de1a7b5a9b 100644
--- a/src/add-ons/kernel/drivers/graphics/radeon/init.c
+++ b/src/add-ons/kernel/drivers/graphics/radeon/init.c
@@ -250,6 +250,7 @@ status_t Radeon_FirstOpen( device_info *di )
 	si->is_atombios = di->is_atombios;
 	si->is_mobility = di->is_mobility;
 	si->panel_pwr_delay = di->si->panel_pwr_delay;
+	si->acc_dma = di->acc_dma;
 
 	
 	// detecting theatre channel in kernel would lead to code duplication,
@@ -359,15 +360,22 @@ status_t Radeon_FirstOpen( device_info *di )
 	if( di->asic == rt_rs100 || di->asic == rt_rs200 || di->asic == rt_rs300)
 		Radeon_Fix_AGP();
 
-	// time to init Command Processor
-	result = Radeon_InitCP( di );
-	if( result != B_OK )
-		goto err;
+	if ( di->acc_dma )
+	{
+		// time to init Command Processor
+		result = Radeon_InitCP( di );
+		if( result != B_OK )
+			goto err;
+			
+		result = Radeon_InitDMA( di );
+		if( result != B_OK )
+			goto err0;
+	}
+	else
+	{
+		SHOW_INFO0( 0, "DMA is diabled using PIO mode");
+	}
 		
-	result = Radeon_InitDMA( di );
-	if( result != B_OK )
-		goto err0;
-	
 //	mem_alloc( di->local_memmgr, 0x100000, (void *)-1, &dma_block, &dma_offset );
 /*	dma_offset = 15 * 1024 * 1024;
 	
@@ -402,7 +410,8 @@ err8:
 // during tests)
 void Radeon_LastClose( device_info *di )
 {
-	Radeon_UninitCP( di );
+	if ( di->acc_dma )
+		Radeon_UninitCP( di );
 
 	// M6 fix - unfortunately, the device is never closed by app_server,
 	// not even before reboot
diff --git a/src/add-ons/kernel/drivers/graphics/radeon/mem_controller.c b/src/add-ons/kernel/drivers/graphics/radeon/mem_controller.c
index 6f5ffc7d11..d073d7c3d3 100644
--- a/src/add-ons/kernel/drivers/graphics/radeon/mem_controller.c
+++ b/src/add-ons/kernel/drivers/graphics/radeon/mem_controller.c
@@ -39,10 +39,21 @@ static uint32 getTopOfRam()
 static void Radeon_SetupMCAddresses_Direct( device_info *di )
 {
 	shared_info *si = di->si;
+	uint32 mc_fb_location;
+	uint32 aper0 = INREG( di->regs, RADEON_CONFIG_APER_0_BASE );
+	
+	// bug in asics mean memory must be aligned to memory size... 
+	if ( IS_DI_R300_VARIANT || di->asic == rt_rv280 ) {
+		aper0 &= ~( di->local_mem_size - 1 );
+	}
+	
+	mc_fb_location = ( aper0 >> 16 ) || 
+		(aper0 + di->local_mem_size - 1 ) & 0xffff0000;
+	
 	
 	// set address range of video memory;
 	// use the same addresses the CPU sees
-	si->memory[mt_local].virtual_addr_start = (uint32)si->framebuffer_pci;
+	si->memory[mt_local].virtual_addr_start = mc_fb_location;
 	si->memory[mt_local].virtual_size = di->local_mem_size;
 
 	// PCI GART has no corresponding CPU address space, so we must find an unused
@@ -148,6 +159,10 @@ void Radeon_InitMemController( device_info *di )
 		si->memory[mt_AGP].virtual_size, si->memory[mt_AGP].virtual_addr_start );
 	
 	//si->nonlocal_mem = di->DMABuffer.ptr;
+	
+	// Turn on PCI GART
+	OUTREGP( regs, RADEON_AIC_CNTL, RADEON_PCIGART_TRANSLATE_EN, 
+		~RADEON_PCIGART_TRANSLATE_EN );
 		
 	// set PCI GART page-table base address
 	OUTREG( regs, RADEON_AIC_PT_BASE, di->pci_gart.GATT.phys );
@@ -159,9 +174,12 @@ void Radeon_InitMemController( device_info *di )
 		si->memory[mt_PCI].virtual_size/*di->pci_gart.buffer.size*/ - 1 );
 
 	// set AGP address range
-	OUTREG( regs, RADEON_MC_AGP_LOCATION, 
+	OUTREG( regs, RADEON_MC_AGP_LOCATION, 0xffffffc0 /* EK magic numbers from X.org
 		(si->memory[mt_AGP].virtual_addr_start >> 16) |
-		((si->memory[mt_AGP].virtual_addr_start + si->memory[mt_AGP].virtual_size - 1) & 0xffff0000 ));
+		((si->memory[mt_AGP].virtual_addr_start + si->memory[mt_AGP].virtual_size - 1) & 0xffff0000 )*/);
+
+	// disable AGP
+	OUTREG( regs, RADEON_AGP_COMMAND, 0 );
 	
 	// set address range of video memory
 	// (lower word = begin >> 16
@@ -175,19 +193,11 @@ void Radeon_InitMemController( device_info *di )
 	OUTREG( regs, RADEON_DISPLAY_BASE_ADDRESS, si->memory[mt_local].virtual_addr_start );
 	OUTREG( regs, RADEON_CRTC2_DISPLAY_BASE_ADDRESS, si->memory[mt_local].virtual_addr_start );
 	OUTREG( regs, RADEON_OV0_BASE_ADDRESS, si->memory[mt_local].virtual_addr_start );
-
-	// disable AGP
-	OUTREG( regs, RADEON_AGP_COMMAND, 0 );
 		 
-	// Turn on PCI GART
-	OUTREGP( regs, RADEON_AIC_CNTL, RADEON_PCIGART_TRANSLATE_EN, 
-		~RADEON_PCIGART_TRANSLATE_EN );
-
-
 	// fix some bus controller setting
 	// I reckon this takes care of discarding data unnecessarily read 
 	// during a burst; let's hope this will fix the nasty CP crashing problem
-	OUTREGP( regs, RADEON_BUS_CNTL, RADEON_BUS_RD_DISCARD_EN, ~RADEON_BUS_RD_DISCARD_EN );
+	// EK this seems unecessary. OUTREGP( regs, RADEON_BUS_CNTL, RADEON_BUS_RD_DISCARD_EN, ~RADEON_BUS_RD_DISCARD_EN );
 	
 //	SHOW_FLOW0( 3, "done" );
 }
diff --git a/src/add-ons/kernel/drivers/graphics/radeon/radeon_driver.h b/src/add-ons/kernel/drivers/graphics/radeon/radeon_driver.h
index 3a0c39cac7..625943d9ae 100644
--- a/src/add-ons/kernel/drivers/graphics/radeon/radeon_driver.h
+++ b/src/add-ons/kernel/drivers/graphics/radeon/radeon_driver.h
@@ -67,6 +67,7 @@ typedef struct {
 	int Tr2w;
 	int loop_latency;
 	int Rloop;
+	int width;
 } ram_info;
 
 // ROM information
@@ -109,7 +110,8 @@ typedef struct device_info {
 	bool		has_vip;
 	bool		is_igp;
 	bool		has_no_i2c;
-	//display_type_e disp_type[2];
+	bool		acc_dma;
+
 	fp_info		fp_info;
 	
 	general_pll_info pll;