Why does bitrate not affect encoding speed?

[Chengbin]

As the title, why does bitrate not affect encoding speed? I find that very weird.

[Dark Shikari]

$ x264 SOCCER_352x288_30_orig_02.yuv --bitrate 100 -o /dev/null
encoded 300 frames, 93.66 fps, 105.93 kb/s

$ x264 SOCCER_352x288_30_orig_02.yuv --bitrate 500 -o /dev/null
encoded 300 frames, 61.74 fps, 526.91 kb/s

$ x264 SOCCER_352x288_30_orig_02.yuv --bitrate 2000 -o /dev/null
encoded 300 frames, 39.67 fps, 2076.50 kb/s

$ x264 SOCCER_352x288_30_orig_02.yuv --bitrate 5000 --qpmin 0 -o /dev/null
encoded 300 frames, 26.23 fps, 5150.22 kb/s

What did you say again?

[Chengbin]

Thanks for clearing that up.

I find it funny that I didn't realize the xxx.xx kb/s number is actually the overall bitrate of the video so far until recently. This just further confirms that I'm an idiot.

[linyx]

Wow, I never realized it was that drastic. If you don't mind, why is it so much faster with lower bitrates?

[thewebchat]

Dumb guess: If you throw away more data at quantization, you don't have to feed as much into the entropy coder.

[Dark Shikari]

Wow, I never realized it was that drastic. If you don't mind, why is it so much faster with lower bitrates?More skipped macroblocks, more decimated 8x8 blocks, fewer IDCTs, fewer dequants, fewer zigzags, more early terminations, less time spent in bit cost calculation, fewer bits sent to CABAC...

[linyx]

Thanks, most of that makes sense.:p

[popper]

$ x264 SOCCER_352x288_30_orig_02.yuv --bitrate 100 -o /dev/null
encoded 300 frames, 93.66 fps, 105.93 kb/s

$ x264 SOCCER_352x288_30_orig_02.yuv --bitrate 500 -o /dev/null
encoded 300 frames, 61.74 fps, 526.91 kb/s

$ x264 SOCCER_352x288_30_orig_02.yuv --bitrate 2000 -o /dev/null
encoded 300 frames, 39.67 fps, 2076.50 kb/s

$ x264 SOCCER_352x288_30_orig_02.yuv --bitrate 5000 --qpmin 0 -o /dev/null
encoded 300 frames, 26.23 fps, 5150.22 kb/s

What did you say again?

interesting to see it so visually obvious layed out like that.

and what difference if any, do you see when you run these same high speed/action football footage i assume! through your current x264 "Macroblock Tree Ratecontrol" testing version?

CIF and widescreen 16/9 aspect versions of low bitrates are interesting for your ISP home upload AVC streamers too.

[Shevach]

Generally speaking bitrate should impact on encoding performance. But if an excerpt consists of near-identical spatially smooth pictures then a change of quants (equivalently a change in expected bit-rate) does not affect on performance.
Indeed, in such case almost all MBs in Inter picture are Skip regardless quant value.