diff --git a/common/mc.c b/common/mc.c
index ac740cf..d062af3 100644
--- a/common/mc.c
+++ b/common/mc.c
@@ -317,7 +317,7 @@ MC_COPY( 16 )
 MC_COPY( 8 )
 MC_COPY( 4 )
 
-static void plane_copy( uint8_t *dst, int i_dst,
+void x264_plane_copy_c( uint8_t *dst, int i_dst,
                         uint8_t *src, int i_src, int w, int h)
 {
     while( h-- )
@@ -483,7 +483,7 @@ void x264_mc_init( int cpu, x264_mc_functions_t *pf )
     pf->copy[PIXEL_8x8]   = mc_copy_w8;
     pf->copy[PIXEL_4x4]   = mc_copy_w4;
 
-    pf->plane_copy = plane_copy;
+    pf->plane_copy = x264_plane_copy_c;
     pf->hpel_filter = hpel_filter;
 
     pf->prefetch_fenc = prefetch_fenc_null;
diff --git a/common/x86/mc-a2.asm b/common/x86/mc-a2.asm
index 245c09f..3d0b491 100644
--- a/common/x86/mc-a2.asm
+++ b/common/x86/mc-a2.asm
@@ -598,22 +598,19 @@ cglobal x264_sfence
     ret
 
 ;-----------------------------------------------------------------------------
-; void x264_plane_copy_mmxext( uint8_t *dst, int i_dst,
-;                              uint8_t *src, int i_src, int w, int h)
+; void x264_plane_copy_core_mmxext( uint8_t *dst, int i_dst,
+;                                   uint8_t *src, int i_src, int w, int h)
 ;-----------------------------------------------------------------------------
-cglobal x264_plane_copy_mmxext, 6,7
+; assumes i_dst and w are multiples of 16, and i_dst>w
+cglobal x264_plane_copy_core_mmxext, 6,7
     movsxdifnidn r1, r1d
     movsxdifnidn r3, r3d
-    add    r4d, 3
-    and    r4d, ~3
-    mov    r6d, r4d
-    and    r6d, ~15
-    sub    r1,  r6
-    sub    r3,  r6
+    sub    r1,  r4
+    sub    r3,  r4
 .loopy:
     mov    r6d, r4d
-    sub    r6d, 64
-    jl     .endx
+    cmp    r6d, 64
+    jl .loop16
 .loopx:
     prefetchnta [r2+256]
     movq   mm0, [r2   ]
@@ -635,10 +632,9 @@ cglobal x264_plane_copy_mmxext, 6,7
     add    r2,  64
     add    r0,  64
     sub    r6d, 64
-    jge    .loopx
-.endx:
+    jg .loopx
     prefetchnta [r2+256]
-    add    r6d, 48
+    cmp    r6d, 16
     jl .end16
 .loop16:
     movq   mm0, [r2  ]
@@ -648,20 +644,19 @@ cglobal x264_plane_copy_mmxext, 6,7
     add    r2,  16
     add    r0,  16
     sub    r6d, 16
-    jge    .loop16
+    jg .loop16
 .end16:
-    add    r6d, 12
-    jl .end4
-.loop4:
-    movd   mm2, [r2+r6]
-    movd   [r0+r6], mm2
-    sub    r6d, 4
-    jge .loop4
-.end4:
+    mov    r6d, r1d
+    ; fill the space between lines too, because partial cachelines fail to write-combine
+.loop_pad:
+    movntq [r0  ], mm0
+    movntq [r0+8], mm0
+    add    r0,  16
+    sub    r6d, 16
+    jg .loop_pad
     add    r2, r3
-    add    r0, r1
     dec    r5d
-    jg     .loopy
+    jg .loopy
     sfence
     emms
     RET
diff --git a/common/x86/mc-c.c b/common/x86/mc-c.c
index b3683a3..fd04392 100644
--- a/common/x86/mc-c.c
+++ b/common/x86/mc-c.c
@@ -88,7 +88,8 @@ extern void x264_mc_chroma_ssse3( uint8_t *src, int i_src_stride,
 extern void x264_mc_chroma_ssse3_cache64( uint8_t *src, int i_src_stride,
                                   uint8_t *dst, int i_dst_stride,
                                   int dx, int dy, int i_width, int i_height );
-extern void x264_plane_copy_mmxext( uint8_t *, int, uint8_t *, int, int w, int h);
+extern void x264_plane_copy_core_mmxext( uint8_t *, int, uint8_t *, int, int w, int h);
+extern void x264_plane_copy_c( uint8_t *, int, uint8_t *, int, int w, int h);
 extern void *x264_memcpy_aligned_mmx( void * dst, const void * src, size_t n );
 extern void *x264_memcpy_aligned_sse2( void * dst, const void * src, size_t n );
 extern void x264_memzero_aligned_mmx( void * dst, int n );
@@ -339,10 +340,23 @@ void x264_hpel_filter_ssse3( uint8_t *dsth, uint8_t *dstv, uint8_t *dstc, uint8_
 #else
 HPEL(16, sse2, sse2, sse2, sse2)
 HPEL(16, ssse3, ssse3, ssse3, ssse3)
-
 #endif
 HPEL(16, sse2_misalign, sse2, sse2_misalign, sse2)
 
+static void x264_plane_copy_mmxext( uint8_t *dst, int i_dst, uint8_t *src, int i_src, int w, int h)
+{
+    if( w < 256 ) { // tiny resolutions don't want non-temporal hints. dunno the exact threshold.
+        x264_plane_copy_c( dst, i_dst, src, i_src, w, h );
+    } else if(i_src > 0) {
+        // have to use plain memcpy on the last line (in memory order) to avoid overreading src
+        x264_plane_copy_core_mmxext( dst, i_dst, src, i_src, (w+15)&~15, h-1 );
+        memcpy( dst+i_dst*(h-1), src+i_src*(h-1), w );
+    } else {
+        memcpy( dst, src, w );
+        x264_plane_copy_core_mmxext( dst+i_dst, i_dst, src+i_src, i_src, (w+15)&~15, h-1 );
+    }
+}
+
 void x264_mc_init_mmx( int cpu, x264_mc_functions_t *pf )
 {
     if( !(cpu&X264_CPU_MMX) )