diff --git a/Source/Core/Core/Src/PowerPC/Jit64/Jit.h b/Source/Core/Core/Src/PowerPC/Jit64/Jit.h
index 8676b1cb5a..71a556cb5f 100644
--- a/Source/Core/Core/Src/PowerPC/Jit64/Jit.h
+++ b/Source/Core/Core/Src/PowerPC/Jit64/Jit.h
@@ -119,7 +119,7 @@ public:
 	void tri_op(int d, int a, int b, bool reversible, void (XEmitter::*op)(Gen::X64Reg, Gen::OpArg));
 	typedef u32 (*Operation)(u32 a, u32 b);
 	void regimmop(int d, int a, bool binary, u32 value, Operation doop, void (XEmitter::*op)(int, const Gen::OpArg&, const Gen::OpArg&), bool Rc = false, bool carry = false);
-	void fp_tri_op(int d, int a, int b, bool reversible, bool dupe, void (XEmitter::*op)(Gen::X64Reg, Gen::OpArg));
+	void fp_tri_op(int d, int a, int b, bool reversible, bool dupe, void (XEmitter::*op_2)(Gen::X64Reg, Gen::OpArg), void (XEmitter::*op_3)(Gen::X64Reg, Gen::X64Reg, Gen::OpArg));
 
 	// OPCODES
 	void unknown_instruction(UGeckoInstruction _inst);
diff --git a/Source/Core/Core/Src/PowerPC/Jit64/Jit_FloatingPoint.cpp b/Source/Core/Core/Src/PowerPC/Jit64/Jit_FloatingPoint.cpp
index ec8ddc63d3..066ba69dfd 100644
--- a/Source/Core/Core/Src/PowerPC/Jit64/Jit_FloatingPoint.cpp
+++ b/Source/Core/Core/Src/PowerPC/Jit64/Jit_FloatingPoint.cpp
@@ -13,37 +13,62 @@ static const u64 GC_ALIGNED16(psAbsMask2[2])  = {0x7FFFFFFFFFFFFFFFULL, 0x7FFFFF
 static const double GC_ALIGNED16(psOneOne2[2]) = {1.0, 1.0};
 static const double one_const = 1.0f;
 
-void Jit64::fp_tri_op(int d, int a, int b, bool reversible, bool dupe, void (XEmitter::*op)(Gen::X64Reg, Gen::OpArg))
+void Jit64::fp_tri_op(int d, int a, int b, bool reversible, bool single,
+		      void (XEmitter::*op_2)(Gen::X64Reg, Gen::OpArg),
+		      void (XEmitter::*op_3)(Gen::X64Reg, Gen::X64Reg, Gen::OpArg))
 {
+	if (!cpu_info.bAVX)
+	{
+		op_3 = nullptr;
+	}
+
 	fpr.Lock(d, a, b);
 	if (d == a)
 	{
-		fpr.BindToRegister(d, true);
-		(this->*op)(fpr.RX(d), fpr.R(b));
+		fpr.BindToRegister(d);
+		(this->*op_2)(fpr.RX(d), fpr.R(b));
 	}
 	else if (d == b)
 	{
 		if (reversible)
 		{
-			fpr.BindToRegister(d, true);
-			(this->*op)(fpr.RX(d), fpr.R(a));
+			fpr.BindToRegister(d);
+			(this->*op_2)(fpr.RX(d), fpr.R(a));
 		}
 		else
 		{
-			MOVSD(XMM0, fpr.R(b));
-			fpr.BindToRegister(d, !dupe);
-			MOVSD(fpr.RX(d), fpr.R(a));
-			(this->*op)(fpr.RX(d), Gen::R(XMM0));
+			if (op_3)
+			{
+				fpr.BindToRegister(d);
+				fpr.BindToRegister(a, true, false);
+				(this->*op_3)(fpr.RX(d), fpr.RX(a), fpr.R(b));
+			}
+			else
+			{
+				MOVSD(XMM0, fpr.R(b));
+				fpr.BindToRegister(d, !single);
+				MOVSD(fpr.RX(d), fpr.R(a));
+				(this->*op_2)(fpr.RX(d), Gen::R(XMM0));
+			}
 		}
 	}
 	else
 	{
-		// Sources different from d, can use rather quick solution
-		fpr.BindToRegister(d, !dupe);
-		MOVSD(fpr.RX(d), fpr.R(a));
-		(this->*op)(fpr.RX(d), fpr.R(b));
+		if (op_3)
+		{
+			fpr.BindToRegister(d, !single);
+			fpr.BindToRegister(a);
+			(this->*op_3)(fpr.RX(d), fpr.RX(a), fpr.R(b));
+		}
+		else
+		{
+			fpr.BindToRegister(d, !single);
+			MOVSD(fpr.RX(d), fpr.R(a));
+			(this->*op_2)(fpr.RX(d), fpr.R(b));
+		}
 	}
-	if (dupe)
+
+	if (single)
 	{
 		ForceSinglePrecisionS(fpr.RX(d));
 		if (cpu_info.bSSE3)
@@ -73,13 +98,13 @@ void Jit64::fp_arith_s(UGeckoInstruction inst)
 		Default(inst); return;
 	}
 
-	bool dupe = inst.OPCD == 59;
+	bool single = inst.OPCD == 59;
 	switch (inst.SUBOP5)
 	{
-	case 18: fp_tri_op(inst.FD, inst.FA, inst.FB, false, dupe, &XEmitter::DIVSD); break; //div
-	case 20: fp_tri_op(inst.FD, inst.FA, inst.FB, false, dupe, &XEmitter::SUBSD); break; //sub
-	case 21: fp_tri_op(inst.FD, inst.FA, inst.FB, true,  dupe, &XEmitter::ADDSD); break; //add
-	case 25: fp_tri_op(inst.FD, inst.FA, inst.FC, true, dupe, &XEmitter::MULSD); break; //mul
+	case 18: fp_tri_op(inst.FD, inst.FA, inst.FB, false, single, &XEmitter::DIVSD, &XEmitter::VDIVSD); break; //div
+	case 20: fp_tri_op(inst.FD, inst.FA, inst.FB, false, single, &XEmitter::SUBSD, &XEmitter::VSUBSD); break; //sub
+	case 21: fp_tri_op(inst.FD, inst.FA, inst.FB, true,  single, &XEmitter::ADDSD, &XEmitter::VADDSD); break; //add
+	case 25: fp_tri_op(inst.FD, inst.FA, inst.FC, true,  single, &XEmitter::MULSD, &XEmitter::VMULSD); break; //mul
 	default:
 		_assert_msg_(DYNA_REC, 0, "fp_arith_s WTF!!!");
 	}
@@ -87,17 +112,24 @@ void Jit64::fp_arith_s(UGeckoInstruction inst)
 
 void Jit64::frsqrtex(UGeckoInstruction inst)
 {
-       INSTRUCTION_START
-       JITDISABLE(bJITFloatingPointOff)
-       int d = inst.FD;
-       int b = inst.FB;
-       fpr.Lock(b, d);
-       fpr.BindToRegister(d, true, true);
-       MOVSD(XMM0, M((void *)&one_const));
-       SQRTSD(XMM1, fpr.R(b));
-       DIVSD(XMM0, R(XMM1));
-       MOVSD(fpr.R(d), XMM0);
-       fpr.UnlockAll();
+	INSTRUCTION_START
+	JITDISABLE(bJITFloatingPointOff)
+	int d = inst.FD;
+	int b = inst.FB;
+	fpr.Lock(b, d);
+	fpr.BindToRegister(d, d == b, true);
+	MOVSD(XMM0, M((void *)&one_const));
+	SQRTSD(XMM1, fpr.R(b));
+	if (cpu_info.bAVX)
+	{
+		VDIVSD(fpr.RX(d), XMM0, R(XMM1));
+	}
+	else
+	{
+		DIVSD(XMM0, R(XMM1));
+		MOVSD(fpr.R(d), XMM0);
+	}
+	fpr.UnlockAll();
 }
 
 void Jit64::fmaddXX(UGeckoInstruction inst)
@@ -198,7 +230,7 @@ void Jit64::fmrx(UGeckoInstruction inst)
 	int d = inst.FD;
 	int b = inst.FB;
 	fpr.Lock(b, d);
-	fpr.BindToRegister(d, true, true);
+	fpr.BindToRegister(d, d == b, true);
 	MOVSD(XMM0, fpr.R(b));
 	MOVSD(fpr.R(d), XMM0);
 	fpr.UnlockAll();