diff --git a/rpcs3/Emu/RSX/D3D12/D3D12.h b/rpcs3/Emu/RSX/D3D12/D3D12.h
index 5582cc0e41..bf93a190cc 100644
--- a/rpcs3/Emu/RSX/D3D12/D3D12.h
+++ b/rpcs3/Emu/RSX/D3D12/D3D12.h
@@ -2,6 +2,7 @@
 #if defined(DX12_SUPPORT)
 
 #include <d3d12.h>
+#include <cassert>
 
 inline
 void check(HRESULT hr)
@@ -10,6 +11,17 @@ void check(HRESULT hr)
 		abort();
 }
 
+/**
+ * Get next value that is aligned by the corresponding power of 2
+ */
+inline
+size_t powerOf2Align(size_t unalignedVal, size_t powerOf2)
+{
+	// check that powerOf2 is power of 2
+	assert(!(powerOf2 & (powerOf2 - 1)));
+	return (unalignedVal + powerOf2 - 1) & ~powerOf2;
+}
+
 /**
  * Send data to dst pointer without polluting cache.
  * Usefull to write to mapped memory from upload heap.
@@ -24,4 +36,28 @@ void streamToBuffer(void* dst, void* src, size_t sizeInBytes)
 	}
 }
 
+/**
+* copy src to dst pointer without polluting cache.
+* Usefull to write to mapped memory from upload heap.
+*/
+inline
+void streamBuffer(void* dst, void* src, size_t sizeInBytes)
+{
+	// Assume 64 bytes cache line
+	assert(powerOf2Align(sizeInBytes, 64));
+	for (unsigned i = 0; i < sizeInBytes / 64; i++)
+	{
+		char *line = (char*)src + i * 64;
+		_mm_prefetch(line, _MM_HINT_NTA);
+		__m128i *srcPtr = (__m128i*) (line);
+		_mm_stream_si128((__m128i*)((char*)dst + i * 64), *srcPtr);
+		srcPtr = (__m128i*) (line + 16);
+		_mm_stream_si128((__m128i*)((char*)dst + i * 64 + 16), *srcPtr);
+		srcPtr = (__m128i*) (line + 32);
+		_mm_stream_si128((__m128i*)((char*)dst + i * 64 + 32), *srcPtr);
+		srcPtr = (__m128i*) (line + 48);
+		_mm_stream_si128((__m128i*)((char*)dst + i * 64 + 48), *srcPtr);
+	}
+}
+
 #endif
\ No newline at end of file
diff --git a/rpcs3/Emu/RSX/D3D12/D3D12Buffer.cpp b/rpcs3/Emu/RSX/D3D12/D3D12Buffer.cpp
index 3a11858e96..9b6a2f2d3f 100644
--- a/rpcs3/Emu/RSX/D3D12/D3D12Buffer.cpp
+++ b/rpcs3/Emu/RSX/D3D12/D3D12Buffer.cpp
@@ -486,7 +486,7 @@ void D3D12GSRender::FillVertexShaderConstantsBuffer()
 
 	void *constantsBufferMap;
 	check(constantsBuffer->Map(0, nullptr, &constantsBufferMap));
-	streamToBuffer(constantsBufferMap, vertexConstantShadowCopy, 512 * 4 * sizeof(float));
+	streamBuffer(constantsBufferMap, vertexConstantShadowCopy, 512 * 4 * sizeof(float));
 	constantsBuffer->Unmap(0, nullptr);
 
 	D3D12_CONSTANT_BUFFER_VIEW_DESC constantBufferViewDesc = {};