diff --git a/Eigen/src/Core/arch/SSE/PacketMath.h b/Eigen/src/Core/arch/SSE/PacketMath.h
index f84e5b3ec..addb2fc0e 100644
--- a/Eigen/src/Core/arch/SSE/PacketMath.h
+++ b/Eigen/src/Core/arch/SSE/PacketMath.h
@@ -173,18 +173,26 @@ template<> EIGEN_STRONG_INLINE Packet4f pmin<Packet4f>(const Packet4f& a, const
 template<> EIGEN_STRONG_INLINE Packet2d pmin<Packet2d>(const Packet2d& a, const Packet2d& b) { return _mm_min_pd(a,b); }
 template<> EIGEN_STRONG_INLINE Packet4i pmin<Packet4i>(const Packet4i& a, const Packet4i& b)
 {
+#ifdef EIGEN_VECTORIZE_SSE4_1
+  return _mm_min_epi32(a,b);
+#else
   // after some bench, this version *is* faster than a scalar implementation
   Packet4i mask = _mm_cmplt_epi32(a,b);
   return _mm_or_si128(_mm_and_si128(mask,a),_mm_andnot_si128(mask,b));
+#endif
 }
 
 template<> EIGEN_STRONG_INLINE Packet4f pmax<Packet4f>(const Packet4f& a, const Packet4f& b) { return _mm_max_ps(a,b); }
 template<> EIGEN_STRONG_INLINE Packet2d pmax<Packet2d>(const Packet2d& a, const Packet2d& b) { return _mm_max_pd(a,b); }
 template<> EIGEN_STRONG_INLINE Packet4i pmax<Packet4i>(const Packet4i& a, const Packet4i& b)
 {
+#ifdef EIGEN_VECTORIZE_SSE4_1
+  return _mm_max_epi32(a,b);
+#else
   // after some bench, this version *is* faster than a scalar implementation
   Packet4i mask = _mm_cmpgt_epi32(a,b);
   return _mm_or_si128(_mm_and_si128(mask,a),_mm_andnot_si128(mask,b));
+#endif
 }
 
 template<> EIGEN_STRONG_INLINE Packet4f pand<Packet4f>(const Packet4f& a, const Packet4f& b) { return _mm_and_ps(a,b); }
diff --git a/test/packetmath.cpp b/test/packetmath.cpp
index cdc945813..804ae9063 100644
--- a/test/packetmath.cpp
+++ b/test/packetmath.cpp
@@ -250,6 +250,17 @@ template<typename Scalar> void packetmath_real()
     data1[internal::random<int>(0, PacketSize)] = 0;
   CHECK_CWISE1_IF(internal::packet_traits<Scalar>::HasLog, std::log, internal::plog);
   CHECK_CWISE1_IF(internal::packet_traits<Scalar>::HasSqrt, std::sqrt, internal::psqrt);
+}
+
+template<typename Scalar> void packetmath_notcomplex()
+{
+  using std::abs;
+  typedef typename internal::packet_traits<Scalar>::type Packet;
+  const int PacketSize = internal::packet_traits<Scalar>::size;
+
+  EIGEN_ALIGN16 Scalar data1[internal::packet_traits<Scalar>::size*4];
+  EIGEN_ALIGN16 Scalar data2[internal::packet_traits<Scalar>::size*4];
+  EIGEN_ALIGN16 Scalar ref[internal::packet_traits<Scalar>::size*4];
 
   ref[0] = data1[0];
   for (int i=0; i<PacketSize; ++i)
@@ -340,6 +351,10 @@ void test_packetmath()
     CALL_SUBTEST_1( packetmath<std::complex<float> >() );
     CALL_SUBTEST_2( packetmath<std::complex<double> >() );
 
+    CALL_SUBTEST_1( packetmath_notcomplex<float>() );
+    CALL_SUBTEST_2( packetmath_notcomplex<double>() );
+    CALL_SUBTEST_3( packetmath_notcomplex<int>() );
+    
     CALL_SUBTEST_1( packetmath_real<float>() );
     CALL_SUBTEST_2( packetmath_real<double>() );