diff --git a/encoder/encoder.c b/encoder/encoder.c
index 9326a39..243361a 100644
--- a/encoder/encoder.c
+++ b/encoder/encoder.c
@@ -677,7 +677,7 @@ x264_t *x264_encoder_open   ( x264_param_t *param )
     x264_reduce_fraction( &h->param.i_fps_num, &h->param.i_fps_den );
 
     /* Init x264_t */
-    h->i_frame = 0;
+    h->i_frame = -1;
     h->i_frame_num = 0;
     h->i_idr_pic_id = 0;
 
@@ -1057,9 +1057,6 @@ static inline void x264_reference_update( x264_t *h )
 {
     int i;
 
-    if( h->fdec->i_frame >= 0 )
-        h->i_frame++;
-
     if( !h->fdec->b_kept_as_ref )
     {
         if( h->param.i_threads > 1 )
@@ -1426,6 +1423,7 @@ int     x264_encoder_encode( x264_t *h,
         }
     }
 
+    h->i_frame++;
     if( h->frames.current[0] == NULL )
     {
         int bframes = 0;
@@ -1892,6 +1890,17 @@ void    x264_encoder_close  ( x264_t *h )
         }
     }
 
+    if( h->param.i_threads > 1)
+    {
+        x264_t *thread_prev;
+
+        thread_prev = h->thread[ h->i_thread_phase % h->param.i_threads ];
+        x264_thread_sync_ratecontrol( h, thread_prev, h );
+        x264_thread_sync_ratecontrol( thread_prev, thread_prev, h );
+        h->i_frame = thread_prev->i_frame + 1 - h->param.i_threads;
+    }
+    h->i_frame++;
+
     /* Slices used and PSNR */
     for( i=0; i<5; i++ )
     {
diff --git a/encoder/ratecontrol.c b/encoder/ratecontrol.c
index af0a00a..ac677a9 100644
--- a/encoder/ratecontrol.c
+++ b/encoder/ratecontrol.c
@@ -483,18 +483,19 @@ int x264_ratecontrol_new( x264_t *h )
             x264_log( h, X264_LOG_WARNING, "2nd pass has fewer frames than 1st pass (%d vs %d)\n",
                       h->param.i_frame_total, rc->num_entries );
         }
-        if( h->param.i_frame_total > rc->num_entries )
+        if( h->param.i_frame_total > rc->num_entries + h->param.i_bframe*4 + X264_THREAD_MAX )
         {
             x264_log( h, X264_LOG_ERROR, "2nd pass has more frames than 1st pass (%d vs %d)\n",
                       h->param.i_frame_total, rc->num_entries );
             return -1;
         }
 
-        rc->entry = (ratecontrol_entry_t*) x264_malloc(rc->num_entries * sizeof(ratecontrol_entry_t));
-        memset(rc->entry, 0, rc->num_entries * sizeof(ratecontrol_entry_t));
+        /* FIXME: ugly padding because VfW drops delayed B-frames and frames delayed by multithreading */
+        rc->entry = (ratecontrol_entry_t*) x264_malloc((rc->num_entries + h->param.i_bframe*4 + X264_THREAD_MAX) * sizeof(ratecontrol_entry_t));
+        memset(rc->entry, 0, (rc->num_entries + h->param.i_bframe*4 + X264_THREAD_MAX) * sizeof(ratecontrol_entry_t));
 
         /* init all to skipped p frames */
-        for(i=0; i<rc->num_entries; i++)
+        for(i=0; i<rc->num_entries + h->param.i_bframe*4 + X264_THREAD_MAX; i++)
         {
             ratecontrol_entry_t *rce = &rc->entry[i];
             rce->pict_type = SLICE_TYPE_P;
@@ -522,7 +523,7 @@ int x264_ratecontrol_new( x264_t *h )
             }
             e = sscanf(p, " in:%d ", &frame_number);
 
-            if(frame_number < 0 || frame_number >= rc->num_entries)
+            if(frame_number < 0 || frame_number >= rc->num_entries + h->param.i_bframe*4 + X264_THREAD_MAX)
             {
                 x264_log(h, X264_LOG_ERROR, "bad frame number (%d) at stats line %d\n", frame_number, i);
                 return -1;
@@ -804,7 +805,7 @@ void x264_ratecontrol_start( x264_t *h, int i_force_qp )
     if( h->param.rc.b_stat_read )
     {
         int frame = h->fenc->i_frame;
-        assert( frame >= 0 && frame < rc->num_entries );
+        assert( frame >= 0 && frame < rc->num_entries + h->param.i_bframe*4 + X264_THREAD_MAX );
         rce = h->rc->rce = &h->rc->entry[frame];
 
         if( h->sh.i_type == SLICE_TYPE_B
@@ -1022,7 +1023,7 @@ int x264_ratecontrol_slice_type( x264_t *h, int frame_num )
     x264_ratecontrol_t *rc = h->rc;
     if( h->param.rc.b_stat_read )
     {
-        if( frame_num >= rc->num_entries )
+        if( frame_num >= rc->num_entries + h->param.i_bframe*4 + X264_THREAD_MAX )
         {
             /* We could try to initialize everything required for ABR and
              * adaptive B-frames, but that would be complicated.