diff --git a/GPU/Software/Clipper.cpp b/GPU/Software/Clipper.cpp
index 1ab91015e..eb181ce61 100644
--- a/GPU/Software/Clipper.cpp
+++ b/GPU/Software/Clipper.cpp
@@ -176,10 +176,10 @@ void ProcessRect(const VertexData& v0, const VertexData& v1)
 		}
 
 		// Four triangles to do backfaces as well. Two of them will get backface culled.
-		ProcessTriangle(*topleft, *topright, *bottomright);
-		ProcessTriangle(*bottomright, *topright, *topleft);
-		ProcessTriangle(*bottomright, *bottomleft, *topleft);
-		ProcessTriangle(*topleft, *bottomleft, *bottomright);
+		ProcessTriangle(*topleft, *topright, *bottomright, buf[3]);
+		ProcessTriangle(*bottomright, *topright, *topleft, buf[3]);
+		ProcessTriangle(*bottomright, *bottomleft, *topleft, buf[3]);
+		ProcessTriangle(*topleft, *bottomleft, *bottomright, buf[3]);
 	} else {
 		// through mode handling
 		VertexData buf[4];
@@ -271,10 +271,17 @@ void ProcessLine(VertexData& v0, VertexData& v1)
 	Rasterizer::DrawLine(data[0], data[1]);
 }
 
-void ProcessTriangle(VertexData& v0, VertexData& v1, VertexData& v2)
-{
+void ProcessTriangle(VertexData& v0, VertexData& v1, VertexData& v2, const VertexData &provoking) {
 	if (gstate.isModeThrough()) {
-		Rasterizer::DrawTriangle(v0, v1, v2);
+		// In case of cull reordering, make sure the right color is on the final vertex.
+		if (gstate.getShadeMode() == GE_SHADE_FLAT) {
+			VertexData corrected2 = v2;
+			corrected2.color0 = provoking.color0;
+			corrected2.color1 = provoking.color1;
+			Rasterizer::DrawTriangle(v0, v1, corrected2);
+		} else {
+			Rasterizer::DrawTriangle(v0, v1, v2);
+		}
 		return;
 	}
 
@@ -339,14 +346,19 @@ void ProcessTriangle(VertexData& v0, VertexData& v1, VertexData& v2)
 		return;
 	}
 
-	for (int i = 0; i+3 <= numIndices; i+=3)
-	{
-		if(indices[i] != SKIP_FLAG)
-		{
+	for (int i = 0; i + 3 <= numIndices; i += 3) {
+		if (indices[i] != SKIP_FLAG) {
 			VertexData data[3] = { *Vertices[indices[i]], *Vertices[indices[i+1]], *Vertices[indices[i+2]] };
 			data[0].screenpos = TransformUnit::ClipToScreen(data[0].clippos);
 			data[1].screenpos = TransformUnit::ClipToScreen(data[1].clippos);
 			data[2].screenpos = TransformUnit::ClipToScreen(data[2].clippos);
+
+			if (gstate.getShadeMode() == GE_SHADE_FLAT) {
+				// So that the order of clipping doesn't matter...
+				data[2].color0 = provoking.color0;
+				data[2].color1 = provoking.color1;
+			}
+
 			Rasterizer::DrawTriangle(data[0], data[1], data[2]);
 		}
 	}
diff --git a/GPU/Software/Clipper.h b/GPU/Software/Clipper.h
index e0926e624..d00262001 100644
--- a/GPU/Software/Clipper.h
+++ b/GPU/Software/Clipper.h
@@ -23,7 +23,7 @@ namespace Clipper {
 
 void ProcessPoint(VertexData& v0);
 void ProcessLine(VertexData& v0, VertexData& v1);
-void ProcessTriangle(VertexData& v0, VertexData& v1, VertexData& v2);
+void ProcessTriangle(VertexData& v0, VertexData& v1, VertexData& v2, const VertexData &provoking);
 void ProcessRect(const VertexData& v0, const VertexData& v1);
 
 }
diff --git a/GPU/Software/TransformUnit.cpp b/GPU/Software/TransformUnit.cpp
index e72e9a85c..f8081ec8d 100644
--- a/GPU/Software/TransformUnit.cpp
+++ b/GPU/Software/TransformUnit.cpp
@@ -342,12 +342,12 @@ void TransformUnit::SubmitPrimitive(void* vertices, void* indices, GEPrimitiveTy
 				case GE_PRIM_TRIANGLES:
 				{
 					if (!gstate.isCullEnabled() || gstate.isModeClear()) {
-						Clipper::ProcessTriangle(data[0], data[1], data[2]);
-						Clipper::ProcessTriangle(data[2], data[1], data[0]);
+						Clipper::ProcessTriangle(data[0], data[1], data[2], data[2]);
+						Clipper::ProcessTriangle(data[2], data[1], data[0], data[2]);
 					} else if (!gstate.getCullMode()) {
-						Clipper::ProcessTriangle(data[2], data[1], data[0]);
+						Clipper::ProcessTriangle(data[2], data[1], data[0], data[2]);
 					} else {
-						Clipper::ProcessTriangle(data[0], data[1], data[2]);
+						Clipper::ProcessTriangle(data[0], data[1], data[2], data[2]);
 					}
 					break;
 				}
@@ -413,7 +413,8 @@ void TransformUnit::SubmitPrimitive(void* vertices, void* indices, GEPrimitiveTy
 					vreader.Goto(vtx);
 				}
 
-				data[(data_index++) % 3] = ReadVertex(vreader);
+				int provoking_index = (data_index++) % 3;
+				data[provoking_index] = ReadVertex(vreader);
 				if (outside_range_flag) {
 					// Drop all primitives containing the current vertex
 					skip_count = 2;
@@ -427,14 +428,14 @@ void TransformUnit::SubmitPrimitive(void* vertices, void* indices, GEPrimitiveTy
 				}
 
 				if (!gstate.isCullEnabled() || gstate.isModeClear()) {
-					Clipper::ProcessTriangle(data[0], data[1], data[2]);
-					Clipper::ProcessTriangle(data[2], data[1], data[0]);
+					Clipper::ProcessTriangle(data[0], data[1], data[2], data[provoking_index]);
+					Clipper::ProcessTriangle(data[2], data[1], data[0], data[provoking_index]);
 				} else if ((!gstate.getCullMode()) ^ ((data_index - 1) % 2)) {
 					// We need to reverse the vertex order for each second primitive,
 					// but we additionally need to do that for every primitive if CCW cullmode is used.
-					Clipper::ProcessTriangle(data[2], data[1], data[0]);
+					Clipper::ProcessTriangle(data[2], data[1], data[0], data[provoking_index]);
 				} else {
-					Clipper::ProcessTriangle(data[0], data[1], data[2]);
+					Clipper::ProcessTriangle(data[0], data[1], data[2], data[provoking_index]);
 				}
 			}
 			break;
@@ -466,7 +467,8 @@ void TransformUnit::SubmitPrimitive(void* vertices, void* indices, GEPrimitiveTy
 					vreader.Goto(vtx);
 				}
 
-				data[2 - ((data_index++) % 2)] = ReadVertex(vreader);
+				int provoking_index = 2 - ((data_index++) % 2);
+				data[provoking_index] = ReadVertex(vreader);
 				if (outside_range_flag) {
 					// Drop all primitives containing the current vertex
 					skip_count = 2;
@@ -480,14 +482,14 @@ void TransformUnit::SubmitPrimitive(void* vertices, void* indices, GEPrimitiveTy
 				}
 
 				if (!gstate.isCullEnabled() || gstate.isModeClear()) {
-					Clipper::ProcessTriangle(data[0], data[1], data[2]);
-					Clipper::ProcessTriangle(data[2], data[1], data[0]);
+					Clipper::ProcessTriangle(data[0], data[1], data[2], data[provoking_index]);
+					Clipper::ProcessTriangle(data[2], data[1], data[0], data[provoking_index]);
 				} else if ((!gstate.getCullMode()) ^ ((data_index - 1) % 2)) {
 					// We need to reverse the vertex order for each second primitive,
 					// but we additionally need to do that for every primitive if CCW cullmode is used.
-					Clipper::ProcessTriangle(data[2], data[1], data[0]);
+					Clipper::ProcessTriangle(data[2], data[1], data[0], data[provoking_index]);
 				} else {
-					Clipper::ProcessTriangle(data[0], data[1], data[2]);
+					Clipper::ProcessTriangle(data[0], data[1], data[2], data[provoking_index]);
 				}
 			}
 			break;