diff --git a/src/radeon_video.c b/src/radeon_video.c
index d4e1f0a..9ee75ff 100644
--- a/src/radeon_video.c
+++ b/src/radeon_video.c
@@ -2434,6 +2434,8 @@ RADEONDisplayVideo(
     CARD32 scale_cntl;
     double dsr;
     int tap_set;
+    int predownscale=0;
+    int src_w_d;
 
     is_rgb=0;
     switch(id){
@@ -2509,6 +2511,8 @@ RADEONDisplayVideo(
     step_by_uv = step_by_y;
 
     /* if the source width was larger than what would fit in overlay scaler increase step_by values */
+#if 0
+    /* XXX this does not appear to work (at least on rv250) */
     i=src_w;
     while(i>pPriv->overlay_scaler_buffer_width){
     	step_by_y++;
@@ -2516,9 +2520,18 @@ RADEONDisplayVideo(
 	h_inc >>=1;
 	i=i/2;
     	}
+#else
+    /* predownscale instead (yes this majorly sucks quality-wise) - will
+       only work for widths up to 2 times the overlay_scaler_buffer_width */
+    src_w_d = src_w;
+    if (src_w>pPriv->overlay_scaler_buffer_width) {
+	src_w_d /= 2; /* odd widths? */
+	predownscale = 1;
+    }
+#endif
 
 
-    h_inc_d = src_w;
+    h_inc_d = src_w_d;
     h_inc_d = h_inc_d/drw_w;
     /* we could do a tad better  - but why
        bother when this concerns downscaling and the code is so much more
@@ -2587,7 +2600,8 @@ #endif
 
     RADEONWaitForFifo(pScrn, 10);
     OUTREG(RADEON_OV0_H_INC, h_inc | ((h_inc_uv >> 1) << 16));
-    OUTREG(RADEON_OV0_STEP_BY, step_by_y | (step_by_uv << 8));
+    OUTREG(RADEON_OV0_STEP_BY, step_by_y | (step_by_uv << 8) |
+	predownscale << 4 | predownscale << 12);
 
     x_off = 8;
     y_off = 0;