diff --git a/GPU/Directx9/ShaderManagerDX9.cpp b/GPU/Directx9/ShaderManagerDX9.cpp
index 86589218c2..25a235a3e5 100644
--- a/GPU/Directx9/ShaderManagerDX9.cpp
+++ b/GPU/Directx9/ShaderManagerDX9.cpp
@@ -308,7 +308,7 @@ void ShaderManagerDX9::VSUpdateUniforms(int dirtyUniforms) {
 #else
 	for (int i = 0; i < 8; i++) {
 		if (dirtyUniforms & (DIRTY_BONEMATRIX0 << i)) {
-			VSSetMatrix4x3(CONST_VS_BONE0 + 4 * i, gstate.boneMatrix + 12 * i);
+			VSSetMatrix4x3_3(CONST_VS_BONE0 + 4 * i, gstate.boneMatrix + 12 * i);
 		}
 	}
 #endif
diff --git a/GPU/Directx9/VertexShaderGeneratorDX9.cpp b/GPU/Directx9/VertexShaderGeneratorDX9.cpp
index dad8ae6b04..38df0f8191 100644
--- a/GPU/Directx9/VertexShaderGeneratorDX9.cpp
+++ b/GPU/Directx9/VertexShaderGeneratorDX9.cpp
@@ -182,10 +182,10 @@ void GenerateVertexShaderDX9(int prim, char *buffer, bool useHWTransform) {
 		if (vertTypeIsSkinningEnabled(vertType)) {
 			int numBones = TranslateNumBonesDX9(vertTypeGetNumBoneWeights(vertType));
 #ifdef USE_BONE_ARRAY
-			WRITE(p, "float4x4 u_bone[%i] : register(c%i);\n", numBones, CONST_VS_BONE0);
+			WRITE(p, "float4x3 u_bone[%i] : register(c%i);\n", numBones, CONST_VS_BONE0);
 #else
 			for (int i = 0; i < numBones; i++) {
-				WRITE(p, "float4x4 u_bone%i : register(c%i);\n", i, CONST_VS_BONE0 + i * 4);
+				WRITE(p, "float4x3 u_bone%i : register(c%i);\n", i, CONST_VS_BONE0 + i * 4);
 			}
 #endif
 		}
@@ -343,9 +343,9 @@ void GenerateVertexShaderDX9(int prim, char *buffer, bool useHWTransform) {
 
 #ifdef USE_BONE_ARRAY
 			if (numWeights == 1)
-				WRITE(p, "  mat4 skinMatrix = a_w1 * u_bone[0]");
+				WRITE(p, "  float4x3 skinMatrix = a_w1 * u_bone[0]");
 			else
-				WRITE(p, "  mat4 skinMatrix = a_w1.x * u_bone[0]");
+				WRITE(p, "  float4x3 skinMatrix = a_w1.x * u_bone[0]");
 			for (int i = 1; i < numWeights; i++) {
 				const char *weightAttr = boneWeightAttr[i];
 				// workaround for "cant do .x of scalar" issue
@@ -357,9 +357,9 @@ void GenerateVertexShaderDX9(int prim, char *buffer, bool useHWTransform) {
 			// Uncomment this to screw up bone shaders to check the vertex shader software fallback
 			// WRITE(p, "THIS SHOULD ERROR! #error");
 			if (numWeights == 1)
-				WRITE(p, "  float4x4 skinMatrix = mul(In.a_w1, u_bone0)");
+				WRITE(p, "  float4x3 skinMatrix = mul(In.a_w1, u_bone0)");
 			else
-				WRITE(p, "  float4x4 skinMatrix = mul(In.a_w1.x, u_bone0)");
+				WRITE(p, "  float4x3 skinMatrix = mul(In.a_w1.x, u_bone0)");
 			for (int i = 1; i < numWeights; i++) {
 				const char *weightAttr = boneWeightAttr[i];
 				// workaround for "cant do .x of scalar" issue
@@ -374,11 +374,11 @@ void GenerateVertexShaderDX9(int prim, char *buffer, bool useHWTransform) {
 			WRITE(p, ";\n");
 
 			// Trying to simplify this results in bugs in LBP...
-			WRITE(p, "  float3 skinnedpos = mul(float4(In.position.xyz, 1.0), skinMatrix).xyz;\n");
+			WRITE(p, "  float3 skinnedpos = mul(float4(In.position.xyz, 1.0), skinMatrix);\n");
 			WRITE(p, "  float3 worldpos = mul(float4(skinnedpos, 1.0), u_world);\n");
 
 			if (hasNormal) {
-				WRITE(p, "  float3 skinnednormal = mul(float4(%sIn.normal, 0.0), skinMatrix).xyz;\n", flipNormal ? "-" : "");
+				WRITE(p, "  float3 skinnednormal = mul(float4(%sIn.normal, 0.0), skinMatrix);\n", flipNormal ? "-" : "");
 				WRITE(p, "  float3 worldnormal = normalize(mul(float4(skinnednormal, 0.0), u_world));\n");
 			} else {
 				WRITE(p, "  float3 worldnormal = mul( mul( float4(0.0, 0.0, 1.0, 0.0), skinMatrix), u_world);\n");
diff --git a/GPU/Directx9/VertexShaderGeneratorDX9.h b/GPU/Directx9/VertexShaderGeneratorDX9.h
index 0623d81261..ee889167d4 100644
--- a/GPU/Directx9/VertexShaderGeneratorDX9.h
+++ b/GPU/Directx9/VertexShaderGeneratorDX9.h
@@ -67,25 +67,25 @@ int TranslateNumBonesDX9(int bones);
 #define CONST_VS_FOGCOEF 18
 #define CONST_VS_AMBIENT 19
 #define CONST_VS_BONE0 20
-#define CONST_VS_BONE1 24
-#define CONST_VS_BONE2 28
-#define CONST_VS_BONE3 32
-#define CONST_VS_BONE4 36
-#define CONST_VS_BONE5 40
-#define CONST_VS_BONE6 44
-#define CONST_VS_BONE7 48
-#define CONST_VS_BONE8 52
-#define CONST_VS_MATAMBIENTALPHA 59
-#define CONST_VS_MATDIFFUSE 60
-#define CONST_VS_MATSPECULAR 61
-#define CONST_VS_MATEMISSIVE 62
-#define CONST_VS_LIGHTPOS 64
-#define CONST_VS_LIGHTDIR 68
-#define CONST_VS_LIGHTATT 72
-#define CONST_VS_LIGHTANGLE 76
-#define CONST_VS_LIGHTSPOTCOEF 80
-#define CONST_VS_LIGHTDIFFUSE 84
-#define CONST_VS_LIGHTSPECULAR 88
-#define CONST_VS_LIGHTAMBIENT 92
+#define CONST_VS_BONE1 23
+#define CONST_VS_BONE2 26
+#define CONST_VS_BONE3 29
+#define CONST_VS_BONE4 32
+#define CONST_VS_BONE5 35
+#define CONST_VS_BONE6 38
+#define CONST_VS_BONE7 41
+#define CONST_VS_BONE8 44
+#define CONST_VS_MATAMBIENTALPHA 47
+#define CONST_VS_MATDIFFUSE 48
+#define CONST_VS_MATSPECULAR 49
+#define CONST_VS_MATEMISSIVE 50
+#define CONST_VS_LIGHTPOS 51
+#define CONST_VS_LIGHTDIR 55
+#define CONST_VS_LIGHTATT 59
+#define CONST_VS_LIGHTANGLE 63
+#define CONST_VS_LIGHTSPOTCOEF 67
+#define CONST_VS_LIGHTDIFFUSE 71
+#define CONST_VS_LIGHTSPECULAR 75
+#define CONST_VS_LIGHTAMBIENT 79
 
 };