diff --git a/plugins/zzogl-pg/opengl/targets.cpp b/plugins/zzogl-pg/opengl/targets.cpp
index 2e8c265d78..cc5b840a0c 100644
--- a/plugins/zzogl-pg/opengl/targets.cpp
+++ b/plugins/zzogl-pg/opengl/targets.cpp
@@ -26,7 +26,9 @@
 #include "zerogs.h"
 #include "targets.h"
 #include "ZZoglShaders.h"
-#include <xmmintrin.h>
+#ifdef ZEROGS_SSE2
+#include <emmintrin.h>
+#endif
 
 #define RHA
 //#define RW
@@ -2301,6 +2303,9 @@ ZeroGS::CMemoryTarget* ZeroGS::CMemoryTargetMngr::GetMemoryTarget(const tex0Info
                 src += 32;
                 dst += 64;
             }
+            // It is advise to use a fence instruction after non temporal move (mm_stream) instruction...
+            // store fence insures that previous store are finish before execute new one.
+            _mm_sfence();
 #else
 			SSE2_UnswizzleZ16Target(dst, src, targ->height * GPU_TEXWIDTH / 16);
 #endif
diff --git a/plugins/zzogl-pg/opengl/x86.cpp b/plugins/zzogl-pg/opengl/x86.cpp
index d29442d6f3..a315083e6c 100644
--- a/plugins/zzogl-pg/opengl/x86.cpp
+++ b/plugins/zzogl-pg/opengl/x86.cpp
@@ -22,7 +22,6 @@
 #include "x86.h"
 
 #if defined(ZEROGS_SSE2)
-#include <xmmintrin.h>
 #include <emmintrin.h>
 #endif
 
@@ -679,7 +678,7 @@ extern "C" void __fastcall WriteCLUT_T16_I4_CSM1_sse2(u32* vm, u32* clut)
     // Create a zero register.
     __m128i zero_128 = _mm_setzero_si128();
 
-    if ((u32)vm & 0x0F) {
+    if ((u32)clut & 0x0F) {
         // Unaligned write.
 
         u16* clut_word_ptr = (u16*)clut;