diff --git a/src/core/cpu_patches.cpp b/src/core/cpu_patches.cpp
index 29c9fa88..a961caec 100644
--- a/src/core/cpu_patches.cpp
+++ b/src/core/cpu_patches.cpp
@@ -424,6 +424,9 @@ static void GenerateINSERTQ(const ZydisDecodedOperand* operands, Xbyak::CodeGene
 
         SaveRegisters(c, {scratch1, scratch2, index, mask});
 
+        // Backup rcx temporarily since we need it to shift
+        c.mov(scratch2, rcx);
+
         // Get upper 64 bits of src
         c.pextrq(index, src, 1);
         c.mov(mask, index);
@@ -437,23 +440,28 @@ static void GenerateINSERTQ(const ZydisDecodedOperand* operands, Xbyak::CodeGene
         c.and_(index, 0x3F);
 
         // Create a mask out of the length
-        c.mov(scratch1, 1);
-        c.shlx(mask, scratch1, mask);
-        c.sub(mask, 1);
+        c.mov(cl, mask.cvt8());
+        c.mov(mask, 1);
+        c.shl(mask, cl);
 
         c.movq(scratch1, src);
-        c.movq(scratch2, dst);
 
         // src &= mask
         c.and_(scratch1, mask);
 
         // dst &= ~(mask << index)
-        c.shlx(mask, mask, index);
+        c.mov(cl, index.cvt8());
+        c.shl(mask, cl);
         c.not_(mask);
-        c.and_(scratch2, mask);
 
-        // dst |= (src << index)
-        c.shlx(scratch1, scratch1, index);
+        // src <<= index
+        c.shl(scratch1, cl);
+
+        // Restore rcx
+        c.mov(rcx, scratch2);
+
+        c.movq(scratch2, dst);
+        c.and_(scratch2, mask);
         c.or_(scratch2, scratch1);
 
         // Insert scratch2 into low 64 bits of dst, upper 64 bits are unaffected