[FFmpeg-user] Testing transcode speeds on Raspberry Pi 4

Ted Park kumowoon1025 at gmail.com
Fri Apr 17 11:57:57 EEST 2020


>>        ffmpeg -y -i SourcePath -b:v 9000k -c:v h264_omx DestPath
>> The difference is the -s flag.  Why would leaving that out result in an error?
> I guess your hardware does not support 4k encoding.
> Carl Eugen

> Makes sense, though until seeing your and Ted's notes I was taking for granted the false notion "if it can display 4k it can no doubt encode 4k"
> Never take anything for granted.
> The overall goal of these tests is to see if / when it makes financial sense to use RPIs as a render farm high res video.
> So far the answer points toward 'not likely'.

I think I did the same thing, or similar at least. I thought the specs said 4k decode & encode but I might have been  misreading 2 4k displays & 4k decode, for hardware encoding it says 1080p60 max. Apparently the SoC on the Pi 4 is basically the same as the one on the 3B+, with better thermal & power management, not a significant upgrade as I thought it was when I got it.

For anything up to 1920x1080 though, I feel like there must be a hardware accelerated scaler that can use the same format as the input for encoding, and if you have a way to manage workers and segment the transcoding job, organizing the units (basically some sort of chassis and power distribution solution) and a "mini-fleet" management, a portable mini render farm on a dolly is feasible for something like a "dailies farm in a backpack". (the "reference" hw accelerated encoding tool included in Raspbian produced output that didn't look as good as I expected from the bitrate)

Ted Park

More information about the ffmpeg-user mailing list