diff --git a/pcsx2/IPU/IPU.cpp b/pcsx2/IPU/IPU.cpp
index c981cbdae0..10593fbd2c 100644
--- a/pcsx2/IPU/IPU.cpp
+++ b/pcsx2/IPU/IPU.cpp
@@ -1158,7 +1158,7 @@ void __fastcall ipu_csc(macroblock_8 *mb8, macroblock_rgb32 *rgb32, int sgn)
 	int i;
 	u8* p = (u8*)rgb32;
 
-	yuv2rgb_sse2();
+	yuv2rgb();
 
 	if (s_thresh[0] > 0)
 	{
diff --git a/pcsx2/IPU/yuv2rgb.cpp b/pcsx2/IPU/yuv2rgb.cpp
index 6ba8a27751..423cacc71c 100644
--- a/pcsx2/IPU/yuv2rgb.cpp
+++ b/pcsx2/IPU/yuv2rgb.cpp
@@ -23,10 +23,38 @@
 #include "IPU.h"
 #include "yuv2rgb.h"
 
+#define IPU_Y_BIAS 16
+#define IPU_C_BIAS 128
+#define IPU_Y_COEFF 0x95	//  1.1640625
+#define IPU_GCR_COEFF -0x68	// -0.8125
+#define IPU_GCB_COEFF -0x32	// -0.390625
+#define IPU_RCR_COEFF 0xcc	//  1.59375
+#define IPU_BCB_COEFF 0x102	//  2.015625
+
+// conforming implementation for reference, do not optimise
+void yuv2rgb_reference(void)
+{
+	for (int y = 0; y < 16; y++)
+		for (int x = 0; x < 16; x++)
+		{
+			s32 lum = (IPU_Y_COEFF * (max(0, (s32)mb8.Y[y][x] - IPU_Y_BIAS))) >> 6;
+			s32 rcr = (IPU_RCR_COEFF * ((s32)mb8.Cr[y>>1][x>>1] - 128)) >> 6;
+			s32 gcr = (IPU_GCR_COEFF * ((s32)mb8.Cr[y>>1][x>>1] - 128)) >> 6;
+			s32 gcb = (IPU_GCB_COEFF * ((s32)mb8.Cb[y>>1][x>>1] - 128)) >> 6;
+			s32 bcb = (IPU_BCB_COEFF * ((s32)mb8.Cb[y>>1][x>>1] - 128)) >> 6;
+
+			rgb32.c[y][x].r = max(0, min(255, (lum + rcr + 1) >> 1));
+			rgb32.c[y][x].g = max(0, min(255, (lum + gcr + gcb + 1) >> 1));
+			rgb32.c[y][x].b = max(0, min(255, (lum + bcb + 1) >> 1));
+			rgb32.c[y][x].a = 0x80; // the norm to save doing this on the alpha pass
+		}
+}
+
 // Everything below is bit accurate to the IPU specification (except maybe rounding).
 // Know the specification before you touch it.
-#define SSE_COEFFICIENTS(x) \
-	{(x)<<2,(x)<<2,(x)<<2,(x)<<2,(x)<<2,(x)<<2,(x)<<2,(x)<<2}
+#define SSE_BYTES(x) {x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x}
+#define SSE_WORDS(x) {x, x, x, x, x, x, x, x}
+#define SSE_COEFFICIENTS(x) SSE_WORDS((x)<<2)
 
 struct SSE2_Tables
 {
@@ -58,19 +86,19 @@ enum
 
 static const __aligned16 SSE2_Tables sse2_tables =
 {
-	{0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000,0x8000},	// c_bias
-	{16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16},			// y_bias
-	{0xff00,0xff00,0xff00,0xff00,0xff00,0xff00,0xff00,0xff00},	// y_mask
+	SSE_WORDS(0x8000),		// c_bias
+	SSE_BYTES(IPU_Y_BIAS),	// y_bias
+	SSE_WORDS(0xff00),		// y_mask
 
 	// Specifying round off instead of round down as everywhere else
 	// implies that this is right
-	{1,1,1,1,1,1,1,1},		// round_1bit
+	SSE_WORDS(1),		// round_1bit
 
-	SSE_COEFFICIENTS(0x95),   // 1.1640625 [Y_coefficients]
-	SSE_COEFFICIENTS(-0x68),  // -0.8125 [GCr_coefficients]
-	SSE_COEFFICIENTS(-0x32),  // -0.390625 [GCb_coefficients]
-	SSE_COEFFICIENTS(0xcc),   // 1.59375 [RCr_coefficients]
-	SSE_COEFFICIENTS(0x102),  // 2.015625 [BCb_coefficients]
+	SSE_COEFFICIENTS(IPU_Y_COEFF),
+	SSE_COEFFICIENTS(IPU_GCR_COEFF),
+	SSE_COEFFICIENTS(IPU_GCB_COEFF),
+	SSE_COEFFICIENTS(IPU_RCR_COEFF),
+	SSE_COEFFICIENTS(IPU_BCB_COEFF),
 };
 
 static __aligned16 u16 yuv2rgb_temp[3][8];
diff --git a/pcsx2/IPU/yuv2rgb.h b/pcsx2/IPU/yuv2rgb.h
index 7b46f969d9..37d1049028 100644
--- a/pcsx2/IPU/yuv2rgb.h
+++ b/pcsx2/IPU/yuv2rgb.h
@@ -15,5 +15,8 @@
 
 #pragma once
 
+#define yuv2rgb yuv2rgb_sse2
+
+extern void yuv2rgb_reference(void);
 extern void yuv2rgb_sse2(void);
 extern void yuv2rgb_init(void);