diff --git a/Source/Core/Core/PowerPC/Jit64/JitAsm.cpp b/Source/Core/Core/PowerPC/Jit64/JitAsm.cpp
index 68434f1789..9c62303f50 100644
--- a/Source/Core/Core/PowerPC/Jit64/JitAsm.cpp
+++ b/Source/Core/Core/PowerPC/Jit64/JitAsm.cpp
@@ -222,14 +222,6 @@ void Jit64AsmRoutineManager::ResetStack(X64CodeBlock& emitter)
 
 void Jit64AsmRoutineManager::GenerateCommon()
 {
-  fifoDirectWrite8 = AlignCode4();
-  GenFifoWrite(8);
-  fifoDirectWrite16 = AlignCode4();
-  GenFifoWrite(16);
-  fifoDirectWrite32 = AlignCode4();
-  GenFifoWrite(32);
-  fifoDirectWrite64 = AlignCode4();
-  GenFifoWrite(64);
   frsqrte = AlignCode4();
   GenFrsqrte();
   fres = AlignCode4();
diff --git a/Source/Core/Core/PowerPC/Jit64Common/EmuCodeBlock.cpp b/Source/Core/Core/PowerPC/Jit64Common/EmuCodeBlock.cpp
index a3054f7509..cdc91294a9 100644
--- a/Source/Core/Core/PowerPC/Jit64Common/EmuCodeBlock.cpp
+++ b/Source/Core/Core/PowerPC/Jit64Common/EmuCodeBlock.cpp
@@ -203,28 +203,6 @@ bool EmuCodeBlock::UnsafeLoadToReg(X64Reg reg_value, OpArg opAddress, int access
   return offsetAddedToAddress;
 }
 
-void EmuCodeBlock::UnsafeWriteGatherPipe(int accessSize)
-{
-  // No need to protect these, they don't touch any state
-  // question - should we inline them instead? Pro: Lose a CALL   Con: Code bloat
-  switch (accessSize)
-  {
-  case 8:
-    CALL(g_jit->GetAsmRoutines()->fifoDirectWrite8);
-    break;
-  case 16:
-    CALL(g_jit->GetAsmRoutines()->fifoDirectWrite16);
-    break;
-  case 32:
-    CALL(g_jit->GetAsmRoutines()->fifoDirectWrite32);
-    break;
-  case 64:
-    CALL(g_jit->GetAsmRoutines()->fifoDirectWrite64);
-    break;
-  }
-  g_jit->js.fifoBytesSinceCheck += accessSize >> 3;
-}
-
 // Visitor that generates code to read a MMIO value.
 template <typename T>
 class MMIOReadCodeGenerator : public MMIO::ReadHandlingMethodVisitor<T>
@@ -622,10 +600,22 @@ bool EmuCodeBlock::WriteToConstAddress(int accessSize, OpArg arg, u32 address,
   // fun tricks...
   if (g_jit->jo.optimizeGatherPipe && PowerPC::IsOptimizableGatherPipeWrite(address))
   {
-    if (!arg.IsSimpleReg(RSCRATCH))
-      MOV(accessSize, R(RSCRATCH), arg);
+    X64Reg arg_reg = RSCRATCH;
 
-    UnsafeWriteGatherPipe(accessSize);
+    // With movbe, we can store inplace without temporary register
+    if (arg.IsSimpleReg() && cpu_info.bMOVBE)
+      arg_reg = arg.GetSimpleReg();
+
+    if (!arg.IsSimpleReg(arg_reg))
+      MOV(accessSize, R(arg_reg), arg);
+
+    // And store it in the gather pipe
+    MOV(64, R(RSCRATCH2), PPCSTATE(gather_pipe_ptr));
+    SwapAndStore(accessSize, MatR(RSCRATCH2), arg_reg);
+    ADD(64, R(RSCRATCH2), Imm8(accessSize >> 3));
+    MOV(64, PPCSTATE(gather_pipe_ptr), R(RSCRATCH2));
+
+    g_jit->js.fifoBytesSinceCheck += accessSize >> 3;
     return false;
   }
   else if (PowerPC::IsOptimizableRAMAddress(address))
diff --git a/Source/Core/Core/PowerPC/Jit64Common/EmuCodeBlock.h b/Source/Core/Core/PowerPC/Jit64Common/EmuCodeBlock.h
index 20e44d0bc0..acf10f80a5 100644
--- a/Source/Core/Core/PowerPC/Jit64Common/EmuCodeBlock.h
+++ b/Source/Core/Core/PowerPC/Jit64Common/EmuCodeBlock.h
@@ -61,7 +61,6 @@ public:
 
   bool UnsafeLoadToReg(Gen::X64Reg reg_value, Gen::OpArg opAddress, int accessSize, s32 offset,
                        bool signExtend, Gen::MovInfo* info = nullptr);
-  void UnsafeWriteGatherPipe(int accessSize);
 
   // Generate a load/write from the MMIO handler for a given address. Only
   // call for known addresses in MMIO range (MMIO::IsMMIOAddress).
diff --git a/Source/Core/Core/PowerPC/Jit64Common/Jit64AsmCommon.cpp b/Source/Core/Core/PowerPC/Jit64Common/Jit64AsmCommon.cpp
index fb9e5f3433..afdd2b1a83 100644
--- a/Source/Core/Core/PowerPC/Jit64Common/Jit64AsmCommon.cpp
+++ b/Source/Core/Core/PowerPC/Jit64Common/Jit64AsmCommon.cpp
@@ -24,22 +24,6 @@
 
 using namespace Gen;
 
-void CommonAsmRoutines::GenFifoWrite(int size)
-{
-  const void* start = GetCodePtr();
-
-  // Assume value in RSCRATCH
-  MOV(64, R(RSCRATCH2), ImmPtr(&PowerPC::ppcState.gather_pipe_ptr));
-  MOV(64, R(RSCRATCH2), MatR(RSCRATCH2));
-  SwapAndStore(size, MatR(RSCRATCH2), RSCRATCH);
-  MOV(64, R(RSCRATCH), ImmPtr(&PowerPC::ppcState.gather_pipe_ptr));
-  ADD(64, R(RSCRATCH2), Imm8(size >> 3));
-  MOV(64, MatR(RSCRATCH), R(RSCRATCH2));
-  RET();
-
-  JitRegister::Register(start, GetCodePtr(), "JIT_FifoWrite_%i", size);
-}
-
 void CommonAsmRoutines::GenFrsqrte()
 {
   const void* start = GetCodePtr();
diff --git a/Source/Core/Core/PowerPC/Jit64Common/Jit64AsmCommon.h b/Source/Core/Core/PowerPC/Jit64Common/Jit64AsmCommon.h
index 09087f94b1..533330264a 100644
--- a/Source/Core/Core/PowerPC/Jit64Common/Jit64AsmCommon.h
+++ b/Source/Core/Core/PowerPC/Jit64Common/Jit64AsmCommon.h
@@ -24,7 +24,6 @@ private:
 class CommonAsmRoutines : public CommonAsmRoutinesBase, public QuantizedMemoryRoutines
 {
 public:
-  void GenFifoWrite(int size);
   void GenFrsqrte();
   void GenFres();
   void GenMfcr();
diff --git a/Source/Core/Core/PowerPC/JitCommon/JitAsmCommon.h b/Source/Core/Core/PowerPC/JitCommon/JitAsmCommon.h
index 3322dccade..fbcdee675f 100644
--- a/Source/Core/Core/PowerPC/JitCommon/JitAsmCommon.h
+++ b/Source/Core/Core/PowerPC/JitCommon/JitAsmCommon.h
@@ -15,11 +15,6 @@ alignas(16) extern const float m_dequantizeTableS[128];
 class CommonAsmRoutinesBase
 {
 public:
-  const u8* fifoDirectWrite8;
-  const u8* fifoDirectWrite16;
-  const u8* fifoDirectWrite32;
-  const u8* fifoDirectWrite64;
-
   const u8* enterCode;
 
   const u8* dispatcherMispredictedBLR;