diff --git a/driver/ControlListUtil.c b/driver/ControlListUtil.c
index bbc0427..400df6c 100644
--- a/driver/ControlListUtil.c
+++ b/driver/ControlListUtil.c
@@ -530,18 +530,25 @@ void clInsertClipWindow(ControlList* cl,
 	*(uint32_t*)cl->nextFreeByte = moveBits(width, 16, 0) | moveBits(height, 16, 16); cl->nextFreeByte += 4;
 }
 
+uint16_t get16bitSignedFixedNumber(float x)
+{
+	int32_t integerPart = roundf(x * 16.0f);
+	return integerPart & 0xffff;
+}
+
 //viewport centre x/y coordinate
 void clInsertViewPortOffset(ControlList* cl,
-						int16_t x, //sint16
-						int16_t y //sint16
+						float x,
+						float y
 						)
 {
 	assert(cl);
 	assert(cl->buffer);
 	assert(cl->nextFreeByte);
 	*cl->nextFreeByte = V3D21_VIEWPORT_OFFSET_opcode; cl->nextFreeByte++;
-	*(int16_t*)cl->nextFreeByte = x * 16; cl->nextFreeByte += 2;
-	*(int16_t*)cl->nextFreeByte = y * 16; cl->nextFreeByte += 2;
+	//expects 16 bit signed fixed point number with 4 fractional bits
+	*(uint16_t*)cl->nextFreeByte = get16bitSignedFixedNumber(x); cl->nextFreeByte += 2;
+	*(uint16_t*)cl->nextFreeByte = get16bitSignedFixedNumber(y); cl->nextFreeByte += 2;
 }
 
 void clInsertZMinMaxClippingPlanes(ControlList* cl,
diff --git a/driver/ControlListUtil.h b/driver/ControlListUtil.h
index 0b57f25..db3c97c 100644
--- a/driver/ControlListUtil.h
+++ b/driver/ControlListUtil.h
@@ -37,6 +37,7 @@ typedef struct CLMarker
 	uint32_t clearColor[2];
 	uint32_t clearDepth, clearStencil;
 	uint32_t width, height; //render w/h
+	uint32_t renderToMip;
 
 	//pointers that point to where all the other CL data is
 	//plus sizes
@@ -144,8 +145,8 @@ void clInsertClipWindow(ControlList* cl,
 						uint32_t bottomPixelCoord, //uint16
 						uint32_t leftPixelCoord);  //uint16
 void clInsertViewPortOffset(ControlList* cl,
-						int16_t x, //sint16
-						int16_t y //sint16
+						float x,
+						float y
 						);
 void clInsertZMinMaxClippingPlanes(ControlList* cl,
 						float minZw,
diff --git a/driver/command.c b/driver/command.c
index 798705e..f1398c0 100644
--- a/driver/command.c
+++ b/driver/command.c
@@ -357,12 +357,21 @@ VKAPI_ATTR VkResult VKAPI_CALL rpi_vkQueueSubmit(
 			//fill out submit cl fields
 			if(writeImage)
 			{
+				uint32_t nonPaddedSize = (marker->width * marker->height * getFormatBpp(writeImage->format)) >> 3;
+
+				uint32_t tiling = writeImage->tiling;
+
+				if(writeImage->tiling == VC4_TILING_FORMAT_T && nonPaddedSize <= 4096)
+				{
+					tiling = VC4_TILING_FORMAT_LT;
+				}
+
 				submitCl.color_write.hindex = writeImageIdx;
 				submitCl.color_write.offset = marker->writeImageOffset;
 				submitCl.color_write.flags = 0;
 				submitCl.color_write.bits =
 						VC4_SET_FIELD(getRenderTargetFormatVC4(writeImage->format), VC4_RENDER_CONFIG_FORMAT) |
-						VC4_SET_FIELD(writeImage->tiling, VC4_RENDER_CONFIG_MEMORY_FORMAT);
+						VC4_SET_FIELD(tiling, VC4_RENDER_CONFIG_MEMORY_FORMAT);
 
 				if(performResolve)
 				{
@@ -380,20 +389,38 @@ VKAPI_ATTR VkResult VKAPI_CALL rpi_vkQueueSubmit(
 
 			if(readImage)
 			{
+				uint32_t nonPaddedSize = (marker->width * marker->height * getFormatBpp(readImage->format)) >> 3;
+
+				uint32_t tiling = readImage->tiling;
+
+				if(readImage->tiling == VC4_TILING_FORMAT_T && nonPaddedSize <= 4096)
+				{
+					tiling = VC4_TILING_FORMAT_LT;
+				}
+
 				submitCl.color_read.hindex = readImageIdx;
 				submitCl.color_read.offset = marker->readImageOffset;
 				submitCl.color_read.flags = readMSAAimage ? VC4_SUBMIT_RCL_SURFACE_READ_IS_FULL_RES : 0;
 				submitCl.color_read.bits = VC4_SET_FIELD(getRenderTargetFormatVC4(readImage->format), VC4_RENDER_CONFIG_FORMAT) |
-						VC4_SET_FIELD(readImage->tiling, VC4_RENDER_CONFIG_MEMORY_FORMAT);
+						VC4_SET_FIELD(tiling, VC4_RENDER_CONFIG_MEMORY_FORMAT);
 			}
 
 			if(writeDepthStencilImage)
 			{
+				uint32_t nonPaddedSize = (marker->width * marker->height * getFormatBpp(writeDepthStencilImage->format)) >> 3;
+
+				uint32_t tiling = writeDepthStencilImage->tiling;
+
+				if(writeDepthStencilImage->tiling == VC4_TILING_FORMAT_T && nonPaddedSize <= 4096)
+				{
+					tiling = VC4_TILING_FORMAT_LT;
+				}
+
 				submitCl.zs_write.hindex = writeDepthStencilImageIdx;
 				submitCl.zs_write.offset = marker->writeDepthStencilImageOffset;
 				submitCl.zs_write.flags = 0;
 				submitCl.zs_write.bits = VC4_SET_FIELD(VC4_LOADSTORE_TILE_BUFFER_ZS, VC4_LOADSTORE_TILE_BUFFER_BUFFER) |
-										 VC4_SET_FIELD(writeDepthStencilImage->tiling, VC4_LOADSTORE_TILE_BUFFER_TILING);	
+										 VC4_SET_FIELD(tiling, VC4_LOADSTORE_TILE_BUFFER_TILING);
 			}
 
 			if(writeMSAAdepthStencilImage)
@@ -406,11 +433,20 @@ VKAPI_ATTR VkResult VKAPI_CALL rpi_vkQueueSubmit(
 
 			if(readDepthStencilImage)
 			{
+				uint32_t nonPaddedSize = (marker->width * marker->height * getFormatBpp(readDepthStencilImage->format)) >> 3;
+
+				uint32_t tiling = readDepthStencilImage->tiling;
+
+				if(readDepthStencilImage->tiling == VC4_TILING_FORMAT_T && nonPaddedSize <= 4096)
+				{
+					tiling = VC4_TILING_FORMAT_LT;
+				}
+
 				submitCl.zs_read.hindex = readDepthStencilImageIdx;
 				submitCl.zs_read.offset = marker->readDepthStencilImageOffset;
 				submitCl.zs_read.flags = readMSAAdepthStencilImage ? VC4_SUBMIT_RCL_SURFACE_READ_IS_FULL_RES : 0; //TODO is this valid?
 				submitCl.zs_read.bits = VC4_SET_FIELD(getRenderTargetFormatVC4(readDepthStencilImage->format), VC4_RENDER_CONFIG_FORMAT) |
-						VC4_SET_FIELD(readDepthStencilImage->tiling, VC4_RENDER_CONFIG_MEMORY_FORMAT);
+						VC4_SET_FIELD(tiling, VC4_RENDER_CONFIG_MEMORY_FORMAT);
 			}
 
 			submitCl.clear_color[0] = marker->clearColor[0];
@@ -460,6 +496,13 @@ VKAPI_ATTR VkResult VKAPI_CALL rpi_vkQueueSubmit(
 			widthInTiles = divRoundUp(width, tileSizeW);
 			heightInTiles = divRoundUp(height, tileSizeH);
 
+			//pad width if rendering to miplevel
+			if(marker->renderToMip)
+			{
+				width = getPow2Pad(width);
+				width = width < 4 ? 4 : width;
+			}
+
 			submitCl.max_x_tile = widthInTiles - 1;
 			submitCl.max_y_tile = heightInTiles - 1;
 			submitCl.width = width;
diff --git a/driver/draw.c b/driver/draw.c
index 64609df..a65ff98 100644
--- a/driver/draw.c
+++ b/driver/draw.c
@@ -71,7 +71,7 @@ static uint32_t drawCommon(VkCommandBuffer commandBuffer)
 
 		//Viewport Offset
 		clFit(commandBuffer, &commandBuffer->binCl, V3D21_VIEWPORT_OFFSET_length);
-		clInsertViewPortOffset(&commandBuffer->binCl, ((int16_t)vp.width) >> 1, ((int16_t)vp.height) >> 1);
+		clInsertViewPortOffset(&commandBuffer->binCl, vp.width * 0.5f, vp.height * 0.5f);
 
 		cb->viewportDirty = 0;
 	}
diff --git a/driver/renderpass.c b/driver/renderpass.c
index b38e79d..2ef5742 100644
--- a/driver/renderpass.c
+++ b/driver/renderpass.c
@@ -229,6 +229,23 @@ void rpi_vkCmdBeginRenderPass(VkCommandBuffer commandBuffer, const VkRenderPassB
 		bpp = getFormatBpp(writeMSAAimage->format);
 	}
 
+	uint32_t biggestMip = 0;
+	for(uint32_t c = 0; c < fb->numAttachmentViews; ++c)
+	{
+		biggestMip = max(biggestMip, fb->attachmentViews[c].subresourceRange.baseMipLevel);
+	}
+
+	//pad render size if we are rendering to a mip level
+	cb->binCl.currMarker->renderToMip = biggestMip > 0;
+
+	uint32_t width = cb->binCl.currMarker->width;
+
+	if(cb->binCl.currMarker->renderToMip)
+	{
+		width = getPow2Pad(width);
+		width = width < 4 ? 4 : width;
+	}
+
 	clFit(commandBuffer, &commandBuffer->binCl, V3D21_TILE_BINNING_MODE_CONFIGURATION_length);
 	clInsertTileBinningModeConfiguration(&commandBuffer->binCl,
 										 0, //double buffer in non ms mode
@@ -237,7 +254,7 @@ void rpi_vkCmdBeginRenderPass(VkCommandBuffer commandBuffer, const VkRenderPassB
 										 0, //auto initialize tile state data array
 										 bpp == 64, //64 bit color mode
 										 writeMSAAimage || writeMSAAdepthStencilImage || performResolve ? 1 : 0, //msaa
-										 cb->binCl.currMarker->width, cb->binCl.currMarker->height,
+										 width, cb->binCl.currMarker->height,
 										 0, //tile state data array address
 										 0, //tile allocation memory size
 										 0); //tile allocation memory address
diff --git a/driver/stateChange.c b/driver/stateChange.c
index defed55..ea79e49 100644
--- a/driver/stateChange.c
+++ b/driver/stateChange.c
@@ -135,6 +135,9 @@ VKAPI_ATTR void VKAPI_CALL rpi_vkCmdClearColorImage(
 
 		commandBuffer->binCl.currMarker->clearColor[0] = commandBuffer->binCl.currMarker->clearColor[1] = packVec4IntoABGR8(pColor->float32);
 		commandBuffer->binCl.currMarker->flags |= VC4_SUBMIT_CL_USE_CLEAR_COLOR;
+
+		commandBuffer->binCl.currMarker->width = i->width;
+		commandBuffer->binCl.currMarker->height = i->height;
 	}
 }
 
diff --git a/test/depthTex/depthTex.cpp b/test/depthTex/depthTex.cpp
index 2d4744c..a91ad2e 100644
--- a/test/depthTex/depthTex.cpp
+++ b/test/depthTex/depthTex.cpp
@@ -1368,7 +1368,7 @@ void CreateTexture()
 
 		VkBufferImageCopy bufferCopyRegion = {};
 		bufferCopyRegion.imageSubresource.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
-		bufferCopyRegion.imageSubresource.mipLevel = 1;
+		bufferCopyRegion.imageSubresource.mipLevel = 0;
 		bufferCopyRegion.imageSubresource.baseArrayLayer = 0;
 		bufferCopyRegion.imageSubresource.layerCount = 1;
 		bufferCopyRegion.imageExtent.width = width;