diff --git a/Common/x64Emitter.cpp b/Common/x64Emitter.cpp
index 1d98544e68..73e84a3739 100644
--- a/Common/x64Emitter.cpp
+++ b/Common/x64Emitter.cpp
@@ -140,6 +140,37 @@ const u8 *XEmitter::AlignCodePage()
 	return code;
 }
 
+const u8 *XEmitter::NopAlignCode16() {
+	int nops = 16 - ((u64)code & 15);
+	if (nops == 16)
+		return code;
+
+	// note: the string lengths are obviously not computable with strlen, but are equal to the index.
+	// Nop strings from https://stackoverflow.com/questions/25545470/long-multi-byte-nops-commonly-understood-macros-or-other-notation
+	static const char * const nopStrings[16] = {
+		"",
+		"\x90",
+		"\x66\x90",
+		"\x0f\x1f\00",
+		"\x0f\x1f\x40\x00",
+		"\x0f\x1f\x44\x00\x00",
+		"\x66\x0f\x1f\x44\x00\x00",
+		"\x0f\x1f\x80\x00\x00\x00\x00",
+		"\x0f\x1f\x84\x00\x00\x00\x00\x00",
+		"\x66\x0f\x1f\x84\x00\x00\x00\x00\x00",
+		"\x66\x66\x0f\x1f\x84\x00\x00\x00\x00\x00",
+		"\x66\x66\x66\x0f\x1f\x84\x00\x00\x00\x00\x00",
+		"\x66\x66\x66\x0f\x1f\x84\x00\x00\x00\x00\x00\x90",
+		"\x66\x66\x66\x0f\x1f\x84\x00\x00\x00\x00\x00\x66\x90",
+		"\x66\x66\x66\x0f\x1f\x84\x00\x00\x00\x00\x00\x0f\x1f\00",
+		"\x66\x66\x66\x0f\x1f\x84\x00\x00\x00\x00\x00\x0f\x1f\x40\x00",
+	};
+
+	memcpy(code, nopStrings[nops], nops);
+	code += nops;
+	return code;
+}
+
 // This operation modifies flags; check to see the flags are locked.
 // If the flags are locked, we should immediately and loudly fail before
 // causing a subtle JIT bug.
diff --git a/Common/x64Emitter.h b/Common/x64Emitter.h
index b52a81f35b..4dcfda6d70 100644
--- a/Common/x64Emitter.h
+++ b/Common/x64Emitter.h
@@ -406,6 +406,10 @@ public:
 	const u8 *AlignCode4();
 	const u8 *AlignCode16();
 	const u8 *AlignCodePage();
+
+	// Nops until the code pointer is 16-byte aligned. Good for loops.
+	const u8 *NopAlignCode16();
+
 	u8 *GetWritableCodePtr();
 
 	void LockFlags() { flags_locked = true; }
diff --git a/GPU/Common/VertexDecoderX86.cpp b/GPU/Common/VertexDecoderX86.cpp
index 9a4dc23449..d92bc964bb 100644
--- a/GPU/Common/VertexDecoderX86.cpp
+++ b/GPU/Common/VertexDecoderX86.cpp
@@ -272,7 +272,7 @@ JittedVertexDecoder VertexDecoderJitCache::Compile(const VertexDecoder &dec, int
 	}
 
 	// Let's not bother with a proper stack frame. We just grab the arguments and go.
-	JumpTarget loopStart = GetCodePtr();
+	JumpTarget loopStart = NopAlignCode16();
 	for (int i = 0; i < dec.numSteps_; i++) {
 		if (!CompileStep(dec, i)) {
 			EndWrite();