diff --git a/GPU/Common/DepalettizeShaderCommon.cpp b/GPU/Common/DepalettizeShaderCommon.cpp
index 871fea767c..3e9cd383ab 100644
--- a/GPU/Common/DepalettizeShaderCommon.cpp
+++ b/GPU/Common/DepalettizeShaderCommon.cpp
@@ -21,6 +21,7 @@
 
 #include "GPU/Common/ShaderId.h"
 #include "GPU/Common/ShaderCommon.h"
+#include "Common/StringUtils.h"
 #include "Common/Log.h"
 #include "Core/Reporting.h"
 #include "GPU/GPUState.h"
@@ -242,8 +243,16 @@ void GenerateDepalShaderFloat(char *buffer, GEBufferFormat pixelFormat, ShaderLa
 		}
 		break;
 	case GE_FORMAT_DEPTH16:
-		sprintf(lookupMethod, "index.r * (1.0 / 256.0)");
+	{
+		// TODO: I think we can handle most scenarios here, but texturing from depth buffers requires an extension on ES 2.0 anyway.
+		if ((mask & (mask + 1)) == 0 && shift < 16) {
+			index_multiplier = 1.0f / (float)(1 << shift);
+			truncate_cpy(lookupMethod, "index.r");
+		} else {
+			formatOK = false;
+		}
 		break;
+	}
 	default:
 		break;
 	}
diff --git a/GPU/Common/FramebufferManagerCommon.cpp b/GPU/Common/FramebufferManagerCommon.cpp
index 06d6b08328..3a8e443ba5 100644
--- a/GPU/Common/FramebufferManagerCommon.cpp
+++ b/GPU/Common/FramebufferManagerCommon.cpp
@@ -1170,9 +1170,9 @@ bool FramebufferManagerCommon::NotifyFramebufferCopy(u32 src, u32 dst, int size,
 
 	if (dstBuffer && srcBuffer && !isMemset) {
 		if (srcBuffer == dstBuffer) {
-			WARN_LOG_REPORT_ONCE(dstsrccpy, G3D, "Intra-buffer memcpy (not supported) %08x -> %08x (size: %x)", src, dst, size);
+			WARN_LOG_ONCE(dstsrccpy, G3D, "Intra-buffer memcpy (not supported) %08x -> %08x (size: %x)", src, dst, size);
 		} else {
-			WARN_LOG_REPORT_ONCE(dstnotsrccpy, G3D, "Inter-buffer memcpy %08x -> %08x (size: %x)", src, dst, size);
+			WARN_LOG_ONCE(dstnotsrccpy, G3D, "Inter-buffer memcpy %08x -> %08x (size: %x)", src, dst, size);
 			// Just do the blit!
 			BlitFramebuffer(dstBuffer, 0, dstY, srcBuffer, 0, srcY, srcBuffer->width, srcH, 0);
 			SetColorUpdated(dstBuffer, skipDrawReason);
@@ -1195,7 +1195,7 @@ bool FramebufferManagerCommon::NotifyFramebufferCopy(u32 src, u32 dst, int size,
 		WARN_LOG_ONCE(btdcpy, G3D, "Memcpy fbo download %08x -> %08x", src, dst);
 		FlushBeforeCopy();
 		if (srcH == 0 || srcY + srcH > srcBuffer->bufferHeight) {
-			WARN_LOG_REPORT_ONCE(btdcpyheight, G3D, "Memcpy fbo download %08x -> %08x skipped, %d+%d is taller than %d", src, dst, srcY, srcH, srcBuffer->bufferHeight);
+			WARN_LOG_ONCE(btdcpyheight, G3D, "Memcpy fbo download %08x -> %08x skipped, %d+%d is taller than %d", src, dst, srcY, srcH, srcBuffer->bufferHeight);
 		} else if (g_Config.bBlockTransferGPU && !srcBuffer->memoryUpdated && !PSP_CoreParameter().compat.flags().DisableReadbacks) {
 			ReadFramebufferToMemory(srcBuffer, 0, srcY, srcBuffer->width, srcH);
 			srcBuffer->usageFlags = (srcBuffer->usageFlags | FB_USAGE_DOWNLOAD) & ~FB_USAGE_DOWNLOAD_CLEAR;
diff --git a/GPU/Common/TextureCacheCommon.cpp b/GPU/Common/TextureCacheCommon.cpp
index 2789d1ff29..14f49ffa90 100644
--- a/GPU/Common/TextureCacheCommon.cpp
+++ b/GPU/Common/TextureCacheCommon.cpp
@@ -764,7 +764,7 @@ bool TextureCacheCommon::AttachFramebuffer(TexCacheEntry *entry, u32 address, Vi
 
 	const u32 mirrorMask = 0x00600000;
 	u32 addr = address & 0x3FFFFFFF;
-	u32 texaddr = (entry->addr + texaddrOffset) & ~mirrorMask;
+	u32 texaddr = entry->addr + texaddrOffset;
 	if (entry->addr & 0x04000000) {
 		addr &= ~mirrorMask;
 		texaddr &= ~mirrorMask;
@@ -800,8 +800,9 @@ bool TextureCacheCommon::AttachFramebuffer(TexCacheEntry *entry, u32 address, Vi
 
 		// Check works for D16 too (???)
 		const bool matchingClutFormat =
-			(framebuffer->format == GE_FORMAT_8888 && entry->format == GE_TFMT_CLUT32) ||
-			(framebuffer->format != GE_FORMAT_8888 && entry->format == GE_TFMT_CLUT16);
+			(channel != NOTIFY_FB_COLOR && entry->format == GE_TFMT_CLUT16) ||
+			(channel == NOTIFY_FB_COLOR && framebuffer->format == GE_FORMAT_8888 && entry->format == GE_TFMT_CLUT32) ||
+			(channel == NOTIFY_FB_COLOR && framebuffer->format != GE_FORMAT_8888 && entry->format == GE_TFMT_CLUT16);
 
 		const bool clutFormat = IsClutFormat((GETextureFormat)(entry->format));
 
diff --git a/ext/native/thin3d/VulkanQueueRunner.cpp b/ext/native/thin3d/VulkanQueueRunner.cpp
index 0ede7acccc..0130f901db 100644
--- a/ext/native/thin3d/VulkanQueueRunner.cpp
+++ b/ext/native/thin3d/VulkanQueueRunner.cpp
@@ -1047,22 +1047,44 @@ void VulkanQueueRunner::PerformRenderPass(const VKRStep &step, VkCommandBuffer c
 		return;
 	}
 
-	// Write-after-write hazards. Fixed flicker in God of War on ARM (before we added another fix).
-	// TODO: depth too
-	if (step.render.framebuffer && step.render.framebuffer->color.layout == VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL) {
-		VkImageMemoryBarrier barrier{};
-		barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
-		barrier.oldLayout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL;
-		barrier.newLayout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL;
-		barrier.subresourceRange.layerCount = 1;
-		barrier.subresourceRange.levelCount = 1;
-		barrier.image = step.render.framebuffer->color.image;
-		barrier.srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
-		barrier.dstAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT | VK_ACCESS_COLOR_ATTACHMENT_READ_BIT;
-		barrier.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
-		barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
-		barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
-		vkCmdPipelineBarrier(cmd, VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT, VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT, 0, 0, nullptr, 0, nullptr, 1, &barrier);
+	// Write-after-write hazards. Fixed flicker in God of War on ARM (before we added another fix that removed these).
+	if (step.render.framebuffer) {
+		int n = 0;
+		int stage = 0;
+		VkImageMemoryBarrier barriers[2]{};
+		if (step.render.framebuffer->color.layout == VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL) {
+			barriers[n].sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
+			barriers[n].oldLayout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL;
+			barriers[n].newLayout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL;
+			barriers[n].subresourceRange.layerCount = 1;
+			barriers[n].subresourceRange.levelCount = 1;
+			barriers[n].image = step.render.framebuffer->color.image;
+			barriers[n].srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
+			barriers[n].dstAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT | VK_ACCESS_COLOR_ATTACHMENT_READ_BIT;
+			barriers[n].subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
+			barriers[n].srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
+			barriers[n].dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
+			stage |= VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT;
+			n++;
+		}
+		if (step.render.framebuffer->depth.layout == VK_IMAGE_LAYOUT_DEPTH_ATTACHMENT_OPTIMAL) {
+			barriers[n].sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
+			barriers[n].oldLayout = VK_IMAGE_LAYOUT_DEPTH_ATTACHMENT_OPTIMAL;
+			barriers[n].newLayout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL;
+			barriers[n].subresourceRange.layerCount = 1;
+			barriers[n].subresourceRange.levelCount = 1;
+			barriers[n].image = step.render.framebuffer->depth.image;
+			barriers[n].srcAccessMask = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT;
+			barriers[n].dstAccessMask = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT | VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT;
+			barriers[n].subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
+			barriers[n].srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
+			barriers[n].dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
+			stage |= VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT;
+			n++;
+		}
+		if (stage) {
+			vkCmdPipelineBarrier(cmd, stage, stage, 0, 0, nullptr, 0, nullptr, n, barriers);
+		}
 	}
 
 	// This is supposed to bind a vulkan render pass to the command buffer.