[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 Mads Johansen):

 Replying to [comment:3 Balling]:
 > >On windows I used the following command, changing cuda with the other
 options written above
 >
 > You are supposed to use -f null NUL instead of -f null - on windows. -
 (pipe) is not fast on Windows, I believe.
 >
 No, I'm not :)

 {{{
 ffmpeg -benchmark -hwaccel cuda -i mandelbrot-yuv420p-2160p.mov -f null
 NUL -  1>mandelbrot-yuv420p-2160p1cuda.txt 2>&1

 [AVFormatContext @ 000001a4062b92c0] Unable to choose an output format for
 'pipe:'; use a standard extension for the filename or specify the format
 manually.
 [out#1 @ 000001a46f7a1940] Error initializing the muxer for pipe:: Invalid
 argument
 Error opening output file -.
 Error opening output files: Invalid argument
 }}}


 > > The only conclusion I can come up with is that it doesn't matter what
 hwaccel you use.
 >
 > No true, I think nvdec (-hwaccel cuda) takes less energy and less memory
 than -c:v hevc_cuvid -i mandelbrot-yuv420p-2160p.mov
 >
 > you can see that in nvidia-smi

 I don't care about memory, but have the bench: maxrss lines if absolutely
 required.

 About the power consumption:

 {{{
 ffmpeg -benchmark -hwaccel cuda -i mandelbrot-yuv420p-2160p.mov -f null -
 > mandelbrot-yuv420p-2160p1cuda.txt 2>&1
 }}}
 68W

 {{{
 ffmpeg -benchmark -hwaccel d3d12va -i mandelbrot-yuv420p-2160p.mov -f null
 - > mandelbrot-yuv420p-2160p1d3d12va.txt 2>&1
 }}}
 68W

 {{{
 ffmpeg -benchmark -hwaccel vulkan -i mandelbrot-yuv420p-2160p.mov -f null
 - > mandelbrot-yuv420p-2160p1vulkan.txt 2>&1
 }}}
 68W

 {{{
 ffmpeg -benchmark -hwaccel dxva2 -i mandelbrot-yuv420p-2160p.mov -f null -
 > mandelbrot-yuv420p-2160p1dxva2.txt 2>&1
 }}}
 68W

 Conclusion about power consumption: It's identical. That means, again,
 that my original conclusion was right.
 ''It does not matter which hwaccel you use. EOD.''
-- 
Ticket URL: <https://trac.ffmpeg.org/ticket/11271#comment:4>
FFmpeg <https://ffmpeg.org>
FFmpeg issue tracker