[FFmpeg-trac] #11271(undetermined:closed): How to choose the best hwaccel?

[FFmpeg]

#11271: How to choose the best hwaccel?
-------------------------------------+-------------------------------------
             Reporter:  Mads         |                    Owner:  (none)
  Johansen                           |
                 Type:  enhancement  |                   Status:  closed
             Priority:  normal       |                Component:
                                     |  undetermined
              Version:  git-master   |               Resolution:  fixed
             Keywords:               |               Blocked By:
             Blocking:               |  Reproduced by developer:  0
Analyzed by developer:  0            |
-------------------------------------+-------------------------------------
Comment (by Balling):

 Oh, and also there is a bug on Linux vs windows for real time decoding,
 not like in your case for as fast as possible, Nvidia-smi takes more
 energy on Linux, I hope they fix it soon. (GPU is stuck at P2 state, while
 on Windows it goes to P5, and P3.)

 https://github.com/NVIDIA/open-gpu-kernel-
 modules/issues/333#issuecomment-2646549118


 It does go to P0 when you decode as fast as possible, i.e. what you
 proposed (slightly wrong on windows, pipe "-" is not supposed to be used,
 instead use NT kernel's abstraction NUL (this is like /dev/null or like ..
 and . in linux folders)): ffmpeg -hwaccel cuda -i video.mp4 -an -c:v
 rawvideo -f null NUL

 In fact you can get perfect 100% utilisation if you decode with -c:v
 h264_cuvid on a file with too much slices (usually Blu-ray has 4 slices,
 while that file I tested has 41). Samples in my issue here:
 https://code.videolan.org/videolan/vlc/-/issues/26918
-- 
Ticket URL: <https://trac.ffmpeg.org/ticket/11271#comment:8>
FFmpeg <https://ffmpeg.org>
FFmpeg issue tracker