diff --git a/src/core/cpu_recompiler_code_generator.cpp b/src/core/cpu_recompiler_code_generator.cpp
index 3346906f7..a4918eccd 100644
--- a/src/core/cpu_recompiler_code_generator.cpp
+++ b/src/core/cpu_recompiler_code_generator.cpp
@@ -138,6 +138,11 @@ bool CodeGenerator::CompileInstruction(const CodeBlockInstruction& cbi)
           result = Compile_Add(cbi);
           break;
 
+        case InstructionFunct::sub:
+        case InstructionFunct::subu:
+          result = Compile_Subtract(cbi);
+          break;
+
         case InstructionFunct::mult:
         case InstructionFunct::multu:
           result = Compile_Multiply(cbi);
@@ -312,6 +317,46 @@ Value CodeGenerator::AddValues(const Value& lhs, const Value& rhs, bool set_flag
   }
 }
 
+Value CodeGenerator::SubValues(const Value& lhs, const Value& rhs, bool set_flags)
+{
+  DebugAssert(lhs.size == rhs.size);
+  if (lhs.IsConstant() && rhs.IsConstant() && !set_flags)
+  {
+    // compile-time
+    u64 new_cv = lhs.constant_value - rhs.constant_value;
+    switch (lhs.size)
+    {
+      case RegSize_8:
+        return Value::FromConstantU8(Truncate8(new_cv));
+
+      case RegSize_16:
+        return Value::FromConstantU16(Truncate16(new_cv));
+
+      case RegSize_32:
+        return Value::FromConstantU32(Truncate32(new_cv));
+
+      case RegSize_64:
+        return Value::FromConstantU64(new_cv);
+
+      default:
+        return Value();
+    }
+  }
+
+  Value res = m_register_cache.AllocateScratch(lhs.size);
+  if (rhs.HasConstantValue(0) && !set_flags)
+  {
+    EmitCopyValue(res.host_reg, lhs);
+    return res;
+  }
+  else
+  {
+    EmitCopyValue(res.host_reg, lhs);
+    EmitSub(res.host_reg, rhs, set_flags);
+    return res;
+  }
+}
+
 std::pair<Value, Value> CodeGenerator::MulValues(const Value& lhs, const Value& rhs, bool signed_multiply)
 {
   DebugAssert(lhs.size == rhs.size);
@@ -1051,6 +1096,26 @@ bool CodeGenerator::Compile_Add(const CodeBlockInstruction& cbi)
   return true;
 }
 
+bool CodeGenerator::Compile_Subtract(const CodeBlockInstruction& cbi)
+{
+  InstructionPrologue(cbi, 1);
+
+  Assert(cbi.instruction.op == InstructionOp::funct);
+  const bool check_overflow = (cbi.instruction.r.funct == InstructionFunct::sub);
+
+  Value lhs = m_register_cache.ReadGuestRegister(cbi.instruction.r.rs);
+  Value rhs = m_register_cache.ReadGuestRegister(cbi.instruction.r.rt);
+
+  Value result = SubValues(lhs, rhs, check_overflow);
+  if (check_overflow)
+    EmitRaiseException(Exception::Ov, Condition::Overflow);
+
+  m_register_cache.WriteGuestRegister(cbi.instruction.r.rd, std::move(result));
+
+  InstructionEpilogue(cbi);
+  return true;
+}
+
 bool CodeGenerator::Compile_Multiply(const CodeBlockInstruction& cbi)
 {
   InstructionPrologue(cbi, 1);
diff --git a/src/core/cpu_recompiler_code_generator.h b/src/core/cpu_recompiler_code_generator.h
index 162a973eb..d1ab8af87 100644
--- a/src/core/cpu_recompiler_code_generator.h
+++ b/src/core/cpu_recompiler_code_generator.h
@@ -48,7 +48,7 @@ public:
   void EmitZeroExtend(HostReg to_reg, RegSize to_size, HostReg from_reg, RegSize from_size);
   void EmitCopyValue(HostReg to_reg, const Value& value);
   void EmitAdd(HostReg to_reg, const Value& value, bool set_flags);
-  void EmitSub(HostReg to_reg, const Value& value);
+  void EmitSub(HostReg to_reg, const Value& value, bool set_flags);
   void EmitCmp(HostReg to_reg, const Value& value);
   void EmitMul(HostReg to_reg_hi, HostReg to_reg_lo, const Value& lhs, const Value& rhs, bool signed_multiply);
   void EmitInc(HostReg to_reg, RegSize size);
@@ -134,6 +134,7 @@ public:
 
   // Value ops
   Value AddValues(const Value& lhs, const Value& rhs, bool set_flags);
+  Value SubValues(const Value& lhs, const Value& rhs, bool set_flags);
   std::pair<Value, Value> MulValues(const Value& lhs, const Value& rhs, bool signed_multiply);
   Value ShlValues(const Value& lhs, const Value& rhs);
   Value ShrValues(const Value& lhs, const Value& rhs);
@@ -179,6 +180,7 @@ private:
   bool Compile_Store(const CodeBlockInstruction& cbi);
   bool Compile_MoveHiLo(const CodeBlockInstruction& cbi);
   bool Compile_Add(const CodeBlockInstruction& cbi);
+  bool Compile_Subtract(const CodeBlockInstruction& cbi);
   bool Compile_Multiply(const CodeBlockInstruction& cbi);
   bool Compile_Branch(const CodeBlockInstruction& cbi);
   bool Compile_lui(const CodeBlockInstruction& cbi);
diff --git a/src/core/cpu_recompiler_code_generator_x64.cpp b/src/core/cpu_recompiler_code_generator_x64.cpp
index bc8d0f4b0..ac23c8dfb 100644
--- a/src/core/cpu_recompiler_code_generator_x64.cpp
+++ b/src/core/cpu_recompiler_code_generator_x64.cpp
@@ -396,7 +396,7 @@ void CodeGenerator::EmitAdd(HostReg to_reg, const Value& value, bool set_flags)
   }
 }
 
-void CodeGenerator::EmitSub(HostReg to_reg, const Value& value)
+void CodeGenerator::EmitSub(HostReg to_reg, const Value& value, bool set_flags)
 {
   DebugAssert(value.IsConstant() || value.IsInHostRegister());