At this time, we compare encoding of video with large dimensions and large number of frames per second. For the test we used video Crowd Run with resolution 3840×2160 and frame rate 50 FPS. Video consists of 500 video frames. The video itself can be downloaded here: Crowd Run 2160p @ 50 (~ 5.8 GB).
Extremely low bitrate Is used to encode video for such parameters: 4000kbps.
Encoding type: 1 pass
Encoding parameters: –bitrate 4000 –input-res 3840×2160 –fps 50 –frame-threads 2 –ref 16 –bframes 16 –merange 48 –max-merge 5 –me 2 –subme 7 –rd 4 –aq-mode 2 –min-keyint 1 –keyint 25 –b-adapt 2 –rc-lookahead 40 –no-scenecut –early-skip –rdpenalty 2 –tu-intra-depth 2 –tu-inter-depth 2 –ssim –input-depth 8
Version: core 142
Encoding type: 2 прохода
Encoding parameters: –bitrate 4000 –input-res 3840×2160 –fps 50 –level 5.2 –preset placebo –slow-firstpass
SSIM Mean Y: 0.7172763 (5.486 dB)*
Encoding speed: 0.14 FPS
Total encoding time: 59 minutes
Peak memory usage: not less than 8 GB
Encoded file size: 4932 KB
SSIM Mean Y: 0.6073204 (4.060db)
Encoding speed: 1.32 FPS (1st pass) и 2.24 FPS (2nd pass)
Total encoding time: 10 minutes
Peak memory usage: ~2 GB
Encoded file size: 4995 KB
* – SSIM is in range from 0 (terrible quality) to 1 (perfect quality). The more the better.
x265 – part of frame 100%
x264 – part of frame 100%
Observations and conclusions
– Even at reduced screenshots we can see that quality which gives x265 much better than x264. We can not achieve the same quality at comparable bit rate on the previous generation codec x264.
– Time encoding and amount of consumed resources at x265 still at times costly than the x264. x265 encoding time 6 time more than placebo x264 (rarely used in practice because of the run-time).