diff --git a/configure b/configure
index 3e4ff75102..68aebc2675 100755
--- a/configure
+++ b/configure
@@ -4695,7 +4695,6 @@ $def_ebx_available
 $(ff_config_enable "$arch_all" "$arch" "ARCH")
 $(ff_config_enable "$subarch_all" "$subarch" "ARCH")
 
-#define HAVE_AMD3DNOW 0
 #define HAVE_MMX ARCH_X86
 #define HAVE_MMX2 ARCH_X86
 #define HAVE_SSE ARCH_X86
diff --git a/cpudetect.h b/cpudetect.h
index 51fd7bc50d..15a3fd11ce 100644
--- a/cpudetect.h
+++ b/cpudetect.h
@@ -35,8 +35,6 @@ typedef struct cpucaps_s {
     bool isX86;
     bool hasMMX;
     bool hasMMX2;
-    bool has3DNow;
-    bool has3DNowExt;
     bool hasSSE;
     bool hasSSE2;
     bool hasSSE3;
diff --git a/libmpcodecs/ad_liba52.c b/libmpcodecs/ad_liba52.c
index e6f0a7e7f7..505532af6b 100644
--- a/libmpcodecs/ad_liba52.c
+++ b/libmpcodecs/ad_liba52.c
@@ -189,10 +189,6 @@ static int init(sh_audio_t *sh_audio)
 #endif
   if(gCpuCaps.hasMMX) a52_accel|=MM_ACCEL_X86_MMX;
   if(gCpuCaps.hasMMX2) a52_accel|=MM_ACCEL_X86_MMXEXT;
-  if(gCpuCaps.has3DNow) a52_accel|=MM_ACCEL_X86_3DNOW;
-#ifdef MM_ACCEL_X86_3DNOWEXT
-  if(gCpuCaps.has3DNowExt) a52_accel|=MM_ACCEL_X86_3DNOWEXT;
-#endif
   a52_state=a52_init (a52_accel);
   if (a52_state == NULL) {
 	mp_msg(MSGT_DECAUDIO,MSGL_ERR,"A52 init failed\n");
diff --git a/libmpcodecs/dec_video.c b/libmpcodecs/dec_video.c
index baa8467c1e..1f4ce7497e 100644
--- a/libmpcodecs/dec_video.c
+++ b/libmpcodecs/dec_video.c
@@ -455,9 +455,7 @@ void *decode_video(sh_video_t *sh_video, struct demux_packet *packet,
 #if HAVE_MMX
     // some codecs are broken, and doesn't restore MMX state :(
     // it happens usually with broken/damaged files.
-    if (gCpuCaps.has3DNow) {
-        __asm__ volatile("femms\n\t":::"memory");
-    } else if (gCpuCaps.hasMMX) {
+    if (gCpuCaps.hasMMX) {
         __asm__ volatile("emms\n\t":::"memory");
     }
 #endif
diff --git a/libmpcodecs/pullup.h b/libmpcodecs/pullup.h
index 0db8f6f894..99aaa45dc2 100644
--- a/libmpcodecs/pullup.h
+++ b/libmpcodecs/pullup.h
@@ -21,8 +21,6 @@
 
 #define PULLUP_CPU_MMX 1
 #define PULLUP_CPU_MMX2 2
-#define PULLUP_CPU_3DNOW 4
-#define PULLUP_CPU_3DNOWEXT 8
 #define PULLUP_CPU_SSE 16
 #define PULLUP_CPU_SSE2 32
 
diff --git a/libmpcodecs/vf_filmdint.c b/libmpcodecs/vf_filmdint.c
index 9c34b4aa07..28eb1e77af 100644
--- a/libmpcodecs/vf_filmdint.c
+++ b/libmpcodecs/vf_filmdint.c
@@ -445,23 +445,6 @@ block_metrics_faster_c(unsigned char *a, unsigned char *b, int as, int bs,
 	    );								     \
     } while (--lines);
 
-static inline struct metrics
-block_metrics_3dnow(unsigned char *a, unsigned char *b, int as, int bs,
-		    int lines, struct vf_priv_s *p, struct frame_stats *s)
-{
-    struct metrics tm;
-#if !HAVE_AMD3DNOW
-    mp_msg(MSGT_VFILTER, MSGL_FATAL, "block_metrics_3dnow: internal error\n");
-#else
-    static const unsigned long long ones = 0x0101010101010101ull;
-
-    BLOCK_METRICS_TEMPLATE();
-    __asm__ volatile("movq %%mm7, %0\n\temms" : "=m" (tm));
-    get_block_stats(&tm, p, s);
-#endif
-    return tm;
-}
-
 #undef PSUMBW
 #undef PSADBW
 #undef PMAXUB
@@ -797,9 +780,6 @@ static void diff_planes(struct vf_priv_s *p, struct frame_stats *s,
 	if (p->mmx2 == 1) {
 	    for (i = 0; i < w; i += 8)
 		block_metrics_mmx2(of+i, nf+i, os, ns, 4, p, s);
-	} else if (p->mmx2 == 2) {
-	    for (i = 0; i < w; i += 8)
-		block_metrics_3dnow(of+i, nf+i, os, ns, 4, p, s);
 	} else if (p->fast > 3) {
 	    for (i = 0; i < w; i += 8)
 		block_metrics_faster_c(of+i, nf+i, os, ns, 4, p, s);
@@ -1426,7 +1406,7 @@ static int vf_open(vf_instance_t *vf, char *args)
     p->dint_thres = 4;
     p->luma_only = 0;
     p->fast = 3;
-    p->mmx2 = gCpuCaps.hasMMX2 ? 1 : gCpuCaps.has3DNow ? 2 : 0;
+    p->mmx2 = gCpuCaps.hasMMX2;
     if (args) {
 	const char *args_remain = parse_args(p, args);
 	if (args_remain) {
@@ -1444,9 +1424,6 @@ static int vf_open(vf_instance_t *vf, char *args)
 	p->mmx2 = 0;
 #if !HAVE_MMX
     p->mmx2 = 0;
-#endif
-#if !HAVE_AMD3DNOW
-    p->mmx2 &= 1;
 #endif
     p->thres.odd  = p->thres.even;
     p->thres.temp = p->thres.noise;
diff --git a/libmpcodecs/vf_pp.c b/libmpcodecs/vf_pp.c
index 694fa09bf2..ae055f8069 100644
--- a/libmpcodecs/vf_pp.c
+++ b/libmpcodecs/vf_pp.c
@@ -49,8 +49,7 @@ static int config(struct vf_instance *vf,
 	unsigned int voflags, unsigned int outfmt){
     int flags=
           (gCpuCaps.hasMMX   ? PP_CPU_CAPS_MMX   : 0)
-	| (gCpuCaps.hasMMX2  ? PP_CPU_CAPS_MMX2  : 0)
-	| (gCpuCaps.has3DNow ? PP_CPU_CAPS_3DNOW : 0);
+	| (gCpuCaps.hasMMX2  ? PP_CPU_CAPS_MMX2  : 0);
 
     switch(outfmt){
     case IMGFMT_444P: flags|= PP_FORMAT_444; break;
diff --git a/libmpcodecs/vf_pullup.c b/libmpcodecs/vf_pullup.c
index 056f952839..83f2f88fb6 100644
--- a/libmpcodecs/vf_pullup.c
+++ b/libmpcodecs/vf_pullup.c
@@ -66,8 +66,6 @@ static void init_pullup(struct vf_instance *vf, mp_image_t *mpi)
 
 	if (gCpuCaps.hasMMX) c->cpu |= PULLUP_CPU_MMX;
 	if (gCpuCaps.hasMMX2) c->cpu |= PULLUP_CPU_MMX2;
-	if (gCpuCaps.has3DNow) c->cpu |= PULLUP_CPU_3DNOW;
-	if (gCpuCaps.has3DNowExt) c->cpu |= PULLUP_CPU_3DNOWEXT;
 	if (gCpuCaps.hasSSE) c->cpu |= PULLUP_CPU_SSE;
 	if (gCpuCaps.hasSSE2) c->cpu |= PULLUP_CPU_SSE2;
 
diff --git a/libmpcodecs/vf_scale.c b/libmpcodecs/vf_scale.c
index 2d56840fd5..66434d0d56 100644
--- a/libmpcodecs/vf_scale.c
+++ b/libmpcodecs/vf_scale.c
@@ -661,8 +661,7 @@ float sws_lum_sharpen= 0.0;
 int get_sws_cpuflags(void){
     return
           (gCpuCaps.hasMMX   ? SWS_CPU_CAPS_MMX   : 0)
-	| (gCpuCaps.hasMMX2  ? SWS_CPU_CAPS_MMX2  : 0)
-	| (gCpuCaps.has3DNow ? SWS_CPU_CAPS_3DNOW : 0);
+	| (gCpuCaps.hasMMX2  ? SWS_CPU_CAPS_MMX2  : 0);
 }
 
 void sws_getFlagsAndFilterFromCmdLine(int *flags, SwsFilter **srcFilterParam, SwsFilter **dstFilterParam)
diff --git a/libmpcodecs/vf_tfields.c b/libmpcodecs/vf_tfields.c
index 4823ef463c..b4fa7a46bf 100644
--- a/libmpcodecs/vf_tfields.c
+++ b/libmpcodecs/vf_tfields.c
@@ -71,42 +71,6 @@ static void deint(unsigned char *dest, int ds, unsigned char *src, int ss, int w
 		fast_memcpy(dest, src, w);
 }
 
-#if HAVE_AMD3DNOW
-static void qpel_li_3DNOW(unsigned char *d, unsigned char *s, int w, int h, int ds, int ss, int up)
-{
-	int i, j, ssd=ss;
-	long crap1, crap2;
-	if (up) {
-		ssd = -ss;
-		fast_memcpy(d, s, w);
-		d += ds;
-		s += ss;
-	}
-	for (i=h-1; i; i--) {
-		__asm__ volatile(
-			"1: \n\t"
-			"movq (%%"REG_S"), %%mm0 \n\t"
-			"movq (%%"REG_S",%%"REG_a"), %%mm1 \n\t"
-			"pavgusb %%mm0, %%mm1 \n\t"
-			"add $8, %%"REG_S" \n\t"
-			"pavgusb %%mm0, %%mm1 \n\t"
-			"movq %%mm1, (%%"REG_D") \n\t"
-			"add $8, %%"REG_D" \n\t"
-			"decl %%ecx \n\t"
-			"jnz 1b \n\t"
-			: "=S"(crap1), "=D"(crap2)
-			: "c"(w>>3), "S"(s), "D"(d), "a"((long)ssd)
-		);
-		for (j=w-(w&7); j<w; j++)
-			d[j] = (s[j+ssd] + 3*s[j])>>2;
-		d += ds;
-		s += ss;
-	}
-	if (!up) fast_memcpy(d, s, w);
-	__asm__ volatile("emms \n\t" : : : "memory");
-}
-#endif
-
 #if HAVE_MMX2
 static void qpel_li_MMX2(unsigned char *d, unsigned char *s, int w, int h, int ds, int ss, int up)
 {
@@ -498,9 +462,6 @@ static int vf_open(vf_instance_t *vf, char *args)
 #endif
 #if HAVE_MMX2
 	if(gCpuCaps.hasMMX2) qpel_li = qpel_li_MMX2;
-#endif
-#if HAVE_AMD3DNOW
-	if(gCpuCaps.has3DNow) qpel_li = qpel_li_3DNOW;
 #endif
 	return 1;
 }
diff --git a/libvo/osd.c b/libvo/osd.c
index 08cda11a81..0c971e3c85 100644
--- a/libvo/osd.c
+++ b/libvo/osd.c
@@ -36,46 +36,29 @@ static const unsigned long long mask24lh  __attribute__((aligned(8))) = 0xFFFF00
 static const unsigned long long mask24hl  __attribute__((aligned(8))) = 0x0000FFFFFFFFFFFFULL;
 #endif
 
-#define CONFIG_RUNTIME_CPUDETECT 1
-
-//Note: we have C, X86-nommx, MMX, MMX2, 3DNOW version therse no 3DNOW+MMX2 one
+//Note: we have C, X86-nommx, MMX, MMX2
 //Plain C versions
-#if !HAVE_MMX || CONFIG_RUNTIME_CPUDETECT
 #define COMPILE_C
-#endif
 
 #if ARCH_X86
 
-#if (HAVE_MMX && !HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT
 #define COMPILE_MMX
-#endif
-
-#if HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT
 #define COMPILE_MMX2
-#endif
-
-#if (HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT
-#define COMPILE_3DNOW
-#endif
 
 #endif /* ARCH_X86 */
 
 #undef HAVE_MMX
 #undef HAVE_MMX2
-#undef HAVE_AMD3DNOW
 #define HAVE_MMX 0
 #define HAVE_MMX2 0
-#define HAVE_AMD3DNOW 0
 
 #if ! ARCH_X86
 
 #ifdef COMPILE_C
 #undef HAVE_MMX
 #undef HAVE_MMX2
-#undef HAVE_AMD3DNOW
 #define HAVE_MMX 0
 #define HAVE_MMX2 0
-#define HAVE_AMD3DNOW 0
 #define RENAME(a) a ## _C
 #include "osd_template.c"
 #endif
@@ -87,10 +70,8 @@ static const unsigned long long mask24hl  __attribute__((aligned(8))) = 0x0000FF
 #undef RENAME
 #undef HAVE_MMX
 #undef HAVE_MMX2
-#undef HAVE_AMD3DNOW
 #define HAVE_MMX 0
 #define HAVE_MMX2 0
-#define HAVE_AMD3DNOW 0
 #define RENAME(a) a ## _X86
 #include "osd_template.c"
 #endif
@@ -100,10 +81,8 @@ static const unsigned long long mask24hl  __attribute__((aligned(8))) = 0x0000FF
 #undef RENAME
 #undef HAVE_MMX
 #undef HAVE_MMX2
-#undef HAVE_AMD3DNOW
 #define HAVE_MMX 1
 #define HAVE_MMX2 0
-#define HAVE_AMD3DNOW 0
 #define RENAME(a) a ## _MMX
 #include "osd_template.c"
 #endif
@@ -113,37 +92,19 @@ static const unsigned long long mask24hl  __attribute__((aligned(8))) = 0x0000FF
 #undef RENAME
 #undef HAVE_MMX
 #undef HAVE_MMX2
-#undef HAVE_AMD3DNOW
 #define HAVE_MMX 1
 #define HAVE_MMX2 1
-#define HAVE_AMD3DNOW 0
 #define RENAME(a) a ## _MMX2
 #include "osd_template.c"
 #endif
 
-//3DNOW versions
-#ifdef COMPILE_3DNOW
-#undef RENAME
-#undef HAVE_MMX
-#undef HAVE_MMX2
-#undef HAVE_AMD3DNOW
-#define HAVE_MMX 1
-#define HAVE_MMX2 0
-#define HAVE_AMD3DNOW 1
-#define RENAME(a) a ## _3DNow
-#include "osd_template.c"
-#endif
-
 #endif /* ARCH_X86 */
 
 void vo_draw_alpha_yv12(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){
-#if CONFIG_RUNTIME_CPUDETECT
 #if ARCH_X86
 	// ordered by speed / fastest first
 	if(gCpuCaps.hasMMX2)
 		vo_draw_alpha_yv12_MMX2(w, h, src, srca, srcstride, dstbase, dststride);
-	else if(gCpuCaps.has3DNow)
-		vo_draw_alpha_yv12_3DNow(w, h, src, srca, srcstride, dstbase, dststride);
 	else if(gCpuCaps.hasMMX)
 		vo_draw_alpha_yv12_MMX(w, h, src, srca, srcstride, dstbase, dststride);
 	else
@@ -151,29 +112,13 @@ void vo_draw_alpha_yv12(int w,int h, unsigned char* src, unsigned char *srca, in
 #else
 		vo_draw_alpha_yv12_C(w, h, src, srca, srcstride, dstbase, dststride);
 #endif
-#else //CONFIG_RUNTIME_CPUDETECT
-#if HAVE_MMX2
-		vo_draw_alpha_yv12_MMX2(w, h, src, srca, srcstride, dstbase, dststride);
-#elif HAVE_AMD3DNOW
-		vo_draw_alpha_yv12_3DNow(w, h, src, srca, srcstride, dstbase, dststride);
-#elif HAVE_MMX
-		vo_draw_alpha_yv12_MMX(w, h, src, srca, srcstride, dstbase, dststride);
-#elif ARCH_X86
-		vo_draw_alpha_yv12_X86(w, h, src, srca, srcstride, dstbase, dststride);
-#else
-		vo_draw_alpha_yv12_C(w, h, src, srca, srcstride, dstbase, dststride);
-#endif
-#endif //!CONFIG_RUNTIME_CPUDETECT
 }
 
 void vo_draw_alpha_yuy2(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){
-#if CONFIG_RUNTIME_CPUDETECT
 #if ARCH_X86
 	// ordered by speed / fastest first
 	if(gCpuCaps.hasMMX2)
 		vo_draw_alpha_yuy2_MMX2(w, h, src, srca, srcstride, dstbase, dststride);
-	else if(gCpuCaps.has3DNow)
-		vo_draw_alpha_yuy2_3DNow(w, h, src, srca, srcstride, dstbase, dststride);
 	else if(gCpuCaps.hasMMX)
 		vo_draw_alpha_yuy2_MMX(w, h, src, srca, srcstride, dstbase, dststride);
 	else
@@ -181,29 +126,13 @@ void vo_draw_alpha_yuy2(int w,int h, unsigned char* src, unsigned char *srca, in
 #else
 		vo_draw_alpha_yuy2_C(w, h, src, srca, srcstride, dstbase, dststride);
 #endif
-#else //CONFIG_RUNTIME_CPUDETECT
-#if HAVE_MMX2
-		vo_draw_alpha_yuy2_MMX2(w, h, src, srca, srcstride, dstbase, dststride);
-#elif HAVE_AMD3DNOW
-		vo_draw_alpha_yuy2_3DNow(w, h, src, srca, srcstride, dstbase, dststride);
-#elif HAVE_MMX
-		vo_draw_alpha_yuy2_MMX(w, h, src, srca, srcstride, dstbase, dststride);
-#elif ARCH_X86
-		vo_draw_alpha_yuy2_X86(w, h, src, srca, srcstride, dstbase, dststride);
-#else
-		vo_draw_alpha_yuy2_C(w, h, src, srca, srcstride, dstbase, dststride);
-#endif
-#endif //!CONFIG_RUNTIME_CPUDETECT
 }
 
 void vo_draw_alpha_uyvy(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){
-#if CONFIG_RUNTIME_CPUDETECT
 #if ARCH_X86
 	// ordered by speed / fastest first
 	if(gCpuCaps.hasMMX2)
 		vo_draw_alpha_uyvy_MMX2(w, h, src, srca, srcstride, dstbase, dststride);
-	else if(gCpuCaps.has3DNow)
-		vo_draw_alpha_uyvy_3DNow(w, h, src, srca, srcstride, dstbase, dststride);
 	else if(gCpuCaps.hasMMX)
 		vo_draw_alpha_uyvy_MMX(w, h, src, srca, srcstride, dstbase, dststride);
 	else
@@ -211,29 +140,13 @@ void vo_draw_alpha_uyvy(int w,int h, unsigned char* src, unsigned char *srca, in
 #else
 		vo_draw_alpha_uyvy_C(w, h, src, srca, srcstride, dstbase, dststride);
 #endif
-#else //CONFIG_RUNTIME_CPUDETECT
-#if HAVE_MMX2
-		vo_draw_alpha_uyvy_MMX2(w, h, src, srca, srcstride, dstbase, dststride);
-#elif HAVE_AMD3DNOW
-		vo_draw_alpha_uyvy_3DNow(w, h, src, srca, srcstride, dstbase, dststride);
-#elif HAVE_MMX
-		vo_draw_alpha_uyvy_MMX(w, h, src, srca, srcstride, dstbase, dststride);
-#elif ARCH_X86
-		vo_draw_alpha_uyvy_X86(w, h, src, srca, srcstride, dstbase, dststride);
-#else
-		vo_draw_alpha_uyvy_C(w, h, src, srca, srcstride, dstbase, dststride);
-#endif
-#endif //!CONFIG_RUNTIME_CPUDETECT
 }
 
 void vo_draw_alpha_rgb24(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){
-#if CONFIG_RUNTIME_CPUDETECT
 #if ARCH_X86
 	// ordered by speed / fastest first
 	if(gCpuCaps.hasMMX2)
 		vo_draw_alpha_rgb24_MMX2(w, h, src, srca, srcstride, dstbase, dststride);
-	else if(gCpuCaps.has3DNow)
-		vo_draw_alpha_rgb24_3DNow(w, h, src, srca, srcstride, dstbase, dststride);
 	else if(gCpuCaps.hasMMX)
 		vo_draw_alpha_rgb24_MMX(w, h, src, srca, srcstride, dstbase, dststride);
 	else
@@ -241,29 +154,13 @@ void vo_draw_alpha_rgb24(int w,int h, unsigned char* src, unsigned char *srca, i
 #else
 		vo_draw_alpha_rgb24_C(w, h, src, srca, srcstride, dstbase, dststride);
 #endif
-#else //CONFIG_RUNTIME_CPUDETECT
-#if HAVE_MMX2
-		vo_draw_alpha_rgb24_MMX2(w, h, src, srca, srcstride, dstbase, dststride);
-#elif HAVE_AMD3DNOW
-		vo_draw_alpha_rgb24_3DNow(w, h, src, srca, srcstride, dstbase, dststride);
-#elif HAVE_MMX
-		vo_draw_alpha_rgb24_MMX(w, h, src, srca, srcstride, dstbase, dststride);
-#elif ARCH_X86
-		vo_draw_alpha_rgb24_X86(w, h, src, srca, srcstride, dstbase, dststride);
-#else
-		vo_draw_alpha_rgb24_C(w, h, src, srca, srcstride, dstbase, dststride);
-#endif
-#endif //!CONFIG_RUNTIME_CPUDETECT
 }
 
 void vo_draw_alpha_rgb32(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){
-#if CONFIG_RUNTIME_CPUDETECT
 #if ARCH_X86
 	// ordered by speed / fastest first
 	if(gCpuCaps.hasMMX2)
 		vo_draw_alpha_rgb32_MMX2(w, h, src, srca, srcstride, dstbase, dststride);
-	else if(gCpuCaps.has3DNow)
-		vo_draw_alpha_rgb32_3DNow(w, h, src, srca, srcstride, dstbase, dststride);
 	else if(gCpuCaps.hasMMX)
 		vo_draw_alpha_rgb32_MMX(w, h, src, srca, srcstride, dstbase, dststride);
 	else
@@ -271,19 +168,6 @@ void vo_draw_alpha_rgb32(int w,int h, unsigned char* src, unsigned char *srca, i
 #else
 		vo_draw_alpha_rgb32_C(w, h, src, srca, srcstride, dstbase, dststride);
 #endif
-#else //CONFIG_RUNTIME_CPUDETECT
-#if HAVE_MMX2
-		vo_draw_alpha_rgb32_MMX2(w, h, src, srca, srcstride, dstbase, dststride);
-#elif HAVE_AMD3DNOW
-		vo_draw_alpha_rgb32_3DNow(w, h, src, srca, srcstride, dstbase, dststride);
-#elif HAVE_MMX
-		vo_draw_alpha_rgb32_MMX(w, h, src, srca, srcstride, dstbase, dststride);
-#elif ARCH_X86
-		vo_draw_alpha_rgb32_X86(w, h, src, srca, srcstride, dstbase, dststride);
-#else
-		vo_draw_alpha_rgb32_C(w, h, src, srca, srcstride, dstbase, dststride);
-#endif
-#endif //!CONFIG_RUNTIME_CPUDETECT
 }
 
 #ifdef FAST_OSD_TABLE
@@ -304,13 +188,10 @@ void vo_draw_alpha_init(void){
 //FIXME the optimized stuff is a lie for 15/16bpp as they aren't optimized yet
 	if( mp_msg_test(MSGT_OSD,MSGL_V) )
 	{
-#if CONFIG_RUNTIME_CPUDETECT
 #if ARCH_X86
 		// ordered per speed fasterst first
 		if(gCpuCaps.hasMMX2)
 			mp_msg(MSGT_OSD,MSGL_INFO,"Using MMX (with tiny bit MMX2) Optimized OnScreenDisplay\n");
-		else if(gCpuCaps.has3DNow)
-			mp_msg(MSGT_OSD,MSGL_INFO,"Using MMX (with tiny bit 3DNow) Optimized OnScreenDisplay\n");
 		else if(gCpuCaps.hasMMX)
 			mp_msg(MSGT_OSD,MSGL_INFO,"Using MMX Optimized OnScreenDisplay\n");
 		else
@@ -318,19 +199,6 @@ void vo_draw_alpha_init(void){
 #else
 			mp_msg(MSGT_OSD,MSGL_INFO,"Using Unoptimized OnScreenDisplay\n");
 #endif
-#else //CONFIG_RUNTIME_CPUDETECT
-#if HAVE_MMX2
-			mp_msg(MSGT_OSD,MSGL_INFO,"Using MMX (with tiny bit MMX2) Optimized OnScreenDisplay\n");
-#elif HAVE_AMD3DNOW
-			mp_msg(MSGT_OSD,MSGL_INFO,"Using MMX (with tiny bit 3DNow) Optimized OnScreenDisplay\n");
-#elif HAVE_MMX
-			mp_msg(MSGT_OSD,MSGL_INFO,"Using MMX Optimized OnScreenDisplay\n");
-#elif ARCH_X86
-			mp_msg(MSGT_OSD,MSGL_INFO,"Using X86 Optimized OnScreenDisplay\n");
-#else
-			mp_msg(MSGT_OSD,MSGL_INFO,"Using Unoptimized OnScreenDisplay\n");
-#endif
-#endif //!CONFIG_RUNTIME_CPUDETECT
 	}
 }
 
diff --git a/libvo/osd_template.c b/libvo/osd_template.c
index 103d169b32..16fe4cff11 100644
--- a/libvo/osd_template.c
+++ b/libvo/osd_template.c
@@ -24,11 +24,7 @@
 #undef PREFETCHW
 #undef PAVGB
 
-#if HAVE_AMD3DNOW
-#define PREFETCH  "prefetch"
-#define PREFETCHW "prefetchw"
-#define PAVGB	  "pavgusb"
-#elif HAVE_MMX2
+#if HAVE_MMX2
 #define PREFETCH "prefetchnta"
 #define PREFETCHW "prefetcht0"
 #define PAVGB	  "pavgb"
@@ -37,12 +33,7 @@
 #define PREFETCHW " # nop"
 #endif
 
-#if HAVE_AMD3DNOW
-/* On K6 femms is faster of emms. On K7 femms is directly mapped on emms. */
-#define EMMS     "femms"
-#else
 #define EMMS     "emms"
-#endif
 
 static inline void RENAME(vo_draw_alpha_yv12)(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){
     int y;
@@ -324,12 +315,6 @@ static inline void RENAME(vo_draw_alpha_rgb32)(int w,int h, unsigned char* src,
     dstbase++;
 #endif
 #if HAVE_MMX
-#if HAVE_AMD3DNOW
-    __asm__ volatile(
-        "pxor %%mm7, %%mm7\n\t"
-        "pcmpeqb %%mm6, %%mm6\n\t" // F..F
-        ::);
-#else /* HAVE_AMD3DNOW */
     __asm__ volatile(
         "pxor %%mm7, %%mm7\n\t"
         "pcmpeqb %%mm5, %%mm5\n\t" // F..F
@@ -337,48 +322,11 @@ static inline void RENAME(vo_draw_alpha_rgb32)(int w,int h, unsigned char* src,
         "psllw $8, %%mm5\n\t" //FF00FF00FF00
         "psrlw $8, %%mm4\n\t" //00FF00FF00FF
         ::);
-#endif /* HAVE_AMD3DNOW */
 #endif /* HAVE_MMX */
     for(y=0;y<h;y++){
         register int x;
 #if ARCH_X86 && (!ARCH_X86_64 || HAVE_MMX)
 #if HAVE_MMX
-#if HAVE_AMD3DNOW
-    __asm__ volatile(
-	PREFETCHW" %0\n\t"
-	PREFETCH" %1\n\t"
-	PREFETCH" %2\n\t"
-	::"m"(*dstbase),"m"(*srca),"m"(*src):"memory");
-    for(x=0;x<w;x+=2){
-     if(srca[x] || srca[x+1])
-	__asm__ volatile(
-		PREFETCHW" 32%0\n\t"
-		PREFETCH" 32%1\n\t"
-		PREFETCH" 32%2\n\t"
-		"movq	%0, %%mm0\n\t" // dstbase
-		"movq	%%mm0, %%mm1\n\t"
-		"punpcklbw %%mm7, %%mm0\n\t"
-		"punpckhbw %%mm7, %%mm1\n\t"
-		"movd	%1, %%mm2\n\t" // srca ABCD0000
-		"paddb	%%mm6, %%mm2\n\t"
-		"punpcklbw %%mm2, %%mm2\n\t" // srca AABBCCDD
-		"punpcklbw %%mm2, %%mm2\n\t" // srca AAAABBBB
-		"movq	%%mm2, %%mm3\n\t"
-		"punpcklbw %%mm7, %%mm2\n\t" // srca 0A0A0A0A
-		"punpckhbw %%mm7, %%mm3\n\t" // srca 0B0B0B0B
-		"pmullw	%%mm2, %%mm0\n\t"
-		"pmullw	%%mm3, %%mm1\n\t"
-		"psrlw	$8, %%mm0\n\t"
-		"psrlw	$8, %%mm1\n\t"
-		"packuswb %%mm1, %%mm0\n\t"
-		"movd %2, %%mm2	\n\t" // src ABCD0000
-		"punpcklbw %%mm2, %%mm2\n\t" // src AABBCCDD
-		"punpcklbw %%mm2, %%mm2\n\t" // src AAAABBBB
-		"paddb	%%mm2, %%mm0\n\t"
-		"movq	%%mm0, %0\n\t"
-		:: "m" (dstbase[4*x]), "m" (srca[x]), "m" (src[x]));
-	}
-#else //this is faster for intels crap
     __asm__ volatile(
 	PREFETCHW" %0\n\t"
 	PREFETCH" %1\n\t"
@@ -430,7 +378,6 @@ static inline void RENAME(vo_draw_alpha_rgb32)(int w,int h, unsigned char* src,
 		:: "m" (dstbase[4*x]), "m" (srca[x]), "m" (src[x]), "m" (bFF)
 		: "%eax");
 	}
-#endif
 #else /* HAVE_MMX */
     for(x=0;x<w;x++){
         if(srca[x]){
diff --git a/loader/win32.c b/loader/win32.c
index 33d3303ea6..3f1ef66a49 100644
--- a/loader/win32.c
+++ b/loader/win32.c
@@ -1020,8 +1020,6 @@ static void WINAPI expGetSystemInfo(SYSTEM_INFO* si)
 	    PF[PF_XMMI_INSTRUCTIONS_AVAILABLE] = TRUE;
 	if (gCpuCaps.hasSSE2)
 	    PF[PF_XMMI64_INSTRUCTIONS_AVAILABLE] = TRUE;
-	if (gCpuCaps.has3DNow)
-	    PF[PF_AMD3D_INSTRUCTIONS_AVAILABLE] = TRUE;
 
 		cachedsi.dwProcessorType = PROCESSOR_INTEL_PENTIUM;
 		cachedsi.wProcessorLevel = 5;