diff --git a/rpcs3/Emu/RSX/Common/BufferUtils.cpp b/rpcs3/Emu/RSX/Common/BufferUtils.cpp
index 5d391a7b46..463959e473 100644
--- a/rpcs3/Emu/RSX/Common/BufferUtils.cpp
+++ b/rpcs3/Emu/RSX/Common/BufferUtils.cpp
@@ -276,4 +276,10 @@ void write_index_array_data_to_buffer(char* dst, unsigned m_draw_mode, unsigned
 			return;
 		}
 	}
+}
+
+void stream_vector(void *dst, u32 x, u32 y, u32 z, u32 w) noexcept
+{
+	__m128i vector = _mm_set_epi32(w, z, y, x);
+	_mm_stream_si128((__m128i*)dst, vector);
 }
\ No newline at end of file
diff --git a/rpcs3/Emu/RSX/Common/BufferUtils.h b/rpcs3/Emu/RSX/Common/BufferUtils.h
index 106825b3ad..b98152ca27 100644
--- a/rpcs3/Emu/RSX/Common/BufferUtils.h
+++ b/rpcs3/Emu/RSX/Common/BufferUtils.h
@@ -42,4 +42,9 @@ void write_index_array_data_to_buffer(char* dst, unsigned m_draw_mode, unsigned
 /**
  * Write index data needed to emulate non indexed non native primitive mode.
  */
-void write_index_array_for_non_indexed_non_native_primitive_to_buffer(char* dst, unsigned m_draw_mode, unsigned first, unsigned count) noexcept;
\ No newline at end of file
+void write_index_array_for_non_indexed_non_native_primitive_to_buffer(char* dst, unsigned m_draw_mode, unsigned first, unsigned count) noexcept;
+
+/**
+ * Stream a 128 bits vector to dst.
+ */
+void stream_vector(void *dst, u32 x, u32 y, u32 z, u32 w) noexcept;
diff --git a/rpcs3/Emu/RSX/D3D12/D3D12Buffer.cpp b/rpcs3/Emu/RSX/D3D12/D3D12Buffer.cpp
index 1313df4f1f..6fbda2c511 100644
--- a/rpcs3/Emu/RSX/D3D12/D3D12Buffer.cpp
+++ b/rpcs3/Emu/RSX/D3D12/D3D12Buffer.cpp
@@ -142,30 +142,6 @@ void D3D12GSRender::load_vertex_index_data(u32 first, u32 count)
 
 void D3D12GSRender::upload_and_bind_scale_offset_matrix(size_t descriptorIndex)
 {
-	float scale_offset_matrix[16] =
-	{
-		1.0f, 0.0f, 0.0f, 0.0f,
-		0.0f, -1.0f, 0.0f, 0.0f,
-		0.0f, 0.0f, 1.0f, 0.0f,
-		0.0f, 0.0f, 0.0f, 1.0f
-	};
-
-	int clip_w = rsx::method_registers[NV4097_SET_SURFACE_CLIP_HORIZONTAL] >> 16;
-	int clip_h = rsx::method_registers[NV4097_SET_SURFACE_CLIP_VERTICAL] >> 16;
-
-	// Scale
-	scale_offset_matrix[0] *= (float&)rsx::method_registers[NV4097_SET_VIEWPORT_SCALE] / (clip_w / 2.f);
-	scale_offset_matrix[5] *= (float&)rsx::method_registers[NV4097_SET_VIEWPORT_SCALE + 1] / (clip_h / 2.f);
-	scale_offset_matrix[10] = (float&)rsx::method_registers[NV4097_SET_VIEWPORT_SCALE + 2];
-
-	// Offset
-	scale_offset_matrix[3] = (float&)rsx::method_registers[NV4097_SET_VIEWPORT_OFFSET] - (clip_w / 2.f);
-	scale_offset_matrix[7] = -((float&)rsx::method_registers[NV4097_SET_VIEWPORT_OFFSET + 1] - (clip_h / 2.f));
-	scale_offset_matrix[11] = (float&)rsx::method_registers[NV4097_SET_VIEWPORT_OFFSET + 2];
-
-	scale_offset_matrix[3] /= clip_w / 2.f;
-	scale_offset_matrix[7] /= clip_h / 2.f;
-
 	assert(m_constantsData.can_alloc(256));
 	size_t heap_offset = m_constantsData.alloc(256);
 
@@ -173,7 +149,7 @@ void D3D12GSRender::upload_and_bind_scale_offset_matrix(size_t descriptorIndex)
 	// Separate constant buffer
 	void *mapped_buffer;
 	ThrowIfFailed(m_constantsData.m_heap->Map(0, &CD3DX12_RANGE(heap_offset, heap_offset + 256), &mapped_buffer));
-	streamToBuffer((char*)mapped_buffer + heap_offset, scale_offset_matrix, 16 * sizeof(float));
+	fill_scale_offset_data((char*)mapped_buffer + heap_offset);
 	int is_alpha_tested = !!(rsx::method_registers[NV4097_SET_ALPHA_TEST_ENABLE]);
 	float alpha_ref = (float&)rsx::method_registers[NV4097_SET_ALPHA_REF];
 	memcpy((char*)mapped_buffer + heap_offset + 16 * sizeof(float), &is_alpha_tested, sizeof(int));
diff --git a/rpcs3/Emu/RSX/RSXThread.cpp b/rpcs3/Emu/RSX/RSXThread.cpp
index e6f765d9d5..8bac746b79 100644
--- a/rpcs3/Emu/RSX/RSXThread.cpp
+++ b/rpcs3/Emu/RSX/RSXThread.cpp
@@ -1024,6 +1024,30 @@ namespace rsx
 		onexit_thread();
 	}
 
+	void thread::fill_scale_offset_data(void *buffer) const noexcept
+	{
+		int clip_w = rsx::method_registers[NV4097_SET_SURFACE_CLIP_HORIZONTAL] >> 16;
+		int clip_h = rsx::method_registers[NV4097_SET_SURFACE_CLIP_VERTICAL] >> 16;
+
+		float scale_x = (float&)rsx::method_registers[NV4097_SET_VIEWPORT_SCALE] / (clip_w / 2.f);
+		float offset_x = (float&)rsx::method_registers[NV4097_SET_VIEWPORT_OFFSET] - (clip_w / 2.f);
+		offset_x /= clip_w / 2.f;
+
+		float scale_y = -(float&)rsx::method_registers[NV4097_SET_VIEWPORT_SCALE + 1] / (clip_h / 2.f);
+		float offset_y = -((float&)rsx::method_registers[NV4097_SET_VIEWPORT_OFFSET + 1] - (clip_h / 2.f));
+		offset_y /= clip_h / 2.f;
+
+		float scale_z = (float&)rsx::method_registers[NV4097_SET_VIEWPORT_SCALE + 2];
+		float offset_z = (float&)rsx::method_registers[NV4097_SET_VIEWPORT_OFFSET + 2];
+
+		float one = 1.f;
+
+		stream_vector(buffer, (u32&)scale_x, 0, 0, (u32&)offset_x);
+		stream_vector((char*)buffer + 16, 0, (u32&)scale_y, 0, (u32&)offset_y);
+		stream_vector((char*)buffer + 32, 0, 0, (u32&)scale_z, (u32&)offset_z);
+		stream_vector((char*)buffer + 48, 0, 0, 0, (u32&)one);
+	}
+
 	u64 thread::timestamp() const
 	{
 		// Get timestamp, and convert it from microseconds to nanoseconds
diff --git a/rpcs3/Emu/RSX/RSXThread.h b/rpcs3/Emu/RSX/RSXThread.h
index 2131b26d82..8feaa9bd77 100644
--- a/rpcs3/Emu/RSX/RSXThread.h
+++ b/rpcs3/Emu/RSX/RSXThread.h
@@ -218,6 +218,12 @@ namespace rsx
 
 		void task();
 
+		/**
+		 * Fill buffer with 4x4 scale offset matrix.
+		 * Vertex shader's position is to be multiplied by this matrix.
+		 */
+		void fill_scale_offset_data(void *buffer) const noexcept;
+
 	public:
 		void reset();
 		void init(const u32 ioAddress, const u32 ioSize, const u32 ctrlAddress, const u32 localAddress);