diff --git a/Core/MIPS/MIPS.cpp b/Core/MIPS/MIPS.cpp
index b68aebddf5..91223bd637 100644
--- a/Core/MIPS/MIPS.cpp
+++ b/Core/MIPS/MIPS.cpp
@@ -122,7 +122,7 @@ MIPSState::MIPSState() {
 	// * 4x4 Matrices are contiguous in RAM, making them, too, fast-loadable in NEON
 
 	// Disadvantages:
-	// * Extra indirection, can be confusing and slower (interpreter only)
+	// * Extra indirection, can be confusing and slower (interpreter only, however we can often skip the table by rerranging formulas)
 	// * Flushing and reloading row registers is now slower
 
 	int i = 0;
diff --git a/Core/MIPS/MIPSVFPUUtils.cpp b/Core/MIPS/MIPSVFPUUtils.cpp
index 87d9875399..2ba4efaa09 100644
--- a/Core/MIPS/MIPSVFPUUtils.cpp
+++ b/Core/MIPS/MIPSVFPUUtils.cpp
@@ -164,6 +164,7 @@ void GetMatrixRows(int matrixReg, MatrixSize msize, u8 vecs[4]) {
 	}
 }
 
+
 void ReadVector(float *rd, VectorSize size, int reg) {
 	int row;
 	int length;
@@ -175,16 +176,17 @@ void ReadVector(float *rd, VectorSize size, int reg) {
 	default: length = 0; break;
 	}
 	int transpose = (reg >> 5) & 1;
-	const int mtx = reg & (7 << 2);
+	const int mtx = ((reg << 2) & 0x70);
 	const int col = reg & 3;
+	// NOTE: We now skip the voffset lookups.
 	if (transpose) {
-		const int base = mtx + col * 32;
-		for (int i = 0; i < length; i++)
-			rd[i] = V(base + ((row+i)&3));
-	} else {
 		const int base = mtx + col;
 		for (int i = 0; i < length; i++)
-			rd[i] = V(base + ((row+i)&3)*32);
+			rd[i] = currentMIPS->v[base + ((row + i) & 3) * 4];
+	} else {
+		const int base = mtx + col * 4;
+		for (int i = 0; i < length; i++)
+			rd[i] = currentMIPS->v[base + ((row + i) & 3)];
 	}
 }
 
@@ -200,28 +202,32 @@ void WriteVector(const float *rd, VectorSize size, int reg) {
 	default: length = 0; break;
 	}
 
-	const int mtx = reg & (7 << 2);
+	const int mtx = ((reg << 2) & 0x70);
 	const int col = reg & 3;
 	bool transpose = (reg >> 5) & 1;
-	if (currentMIPS->VfpuWriteMask() == 0) {
-		if (transpose) {
-			const int base = mtx + col * 32;
+	// NOTE: We now skip the voffset lookups.
+	if (transpose) {
+		const int base = mtx + col;
+		if (currentMIPS->VfpuWriteMask() == 0) {
 			for (int i = 0; i < length; i++)
-				V(base + ((row+i)&3)) = rd[i];
+				currentMIPS->v[base + ((row+i) & 3) * 4] = rd[i];
 		} else {
-			const int base = mtx + col;
-			for (int i = 0; i < length; i++)
-				V(base + ((row+i)&3)*32) = rd[i];
+			for (int i = 0; i < length; i++) {
+				if (!currentMIPS->VfpuWriteMask(i)) {
+					currentMIPS->v[base + ((row+i) & 3) * 4] = rd[i];
+				}
+			}
 		}
 	} else {
-		for (int i = 0; i < length; i++) {
-			if (!currentMIPS->VfpuWriteMask(i)) {
-				int index = mtx;
-				if (transpose)
-					index += ((row+i)&3) + col*32;
-				else
-					index += col + ((row+i)&3)*32;
-				V(index) = rd[i];
+		const int base = mtx + col * 4;
+		if (currentMIPS->VfpuWriteMask() == 0) {
+			for (int i = 0; i < length; i++)
+				currentMIPS->v[base + ((row + i) & 3)] = rd[i];
+		} else {
+			for (int i = 0; i < length; i++) {
+				if (!currentMIPS->VfpuWriteMask(i)) {
+					currentMIPS->v[base + ((row + i) & 3)] = rd[i];
+				}
 			}
 		}
 	}