I’ve been trying to setup Tdarr to transcode using my AMD integrated GPU instead of my CPU, but all I’m finding online is people using nvidia cards…well, I don’t have one of those, but I have an AMD CPU with integrated gpu, so I wanted to use that, but apparently that’s extremely uncommon and I can’t find any working solutions.
Edit: I’m running Tdarr in a docker container on my OMV media server.
Edit 2: I’ve gotten it working, but the compression is nonexisting. A h264 -> h265 transcode increases file size by ~5%.
Needed to add my Tdarr container to the render group and pass through the dev/dri/renderD128 folder.
I have AMD hardware acceleration working for Plex in an LXC container with an AMD APU so I’d assume it’s possible.
Tdarr seems to use ffmpeg under the covers, so I’d focus on getting that working with amd. If I remember I had to install the mesa drivers and pass in the /dev/dri folder. Then you can check ffmpeg for the amf encoders (AMD media framework).
I actually just got it working, after spending all day yesterday. The speed is magnificent, but the compression rate is absolutely abysmal. When i use the integrated GPU an h264 encoded file end up taking up 5-10% more space when converted to h265, whereas when just use CPU an h264->h265 is around a 45% reduction in size. I have no use for speed if it doesn’t reduce file size at all.
Besides maybe confusing the codecs, hardware encoders, especially the AMD ones, are always less space efficient than software encoders.
If you want to convert video for long-term storage, please use a software encoder.
Fair enough, I just saw so many using GPUs and how much faster it was and assumed it was the same size-wise with the files. I’m getting ~40fps with my CPU, so it’s going to take forever to do all of it.
Uh… What?
GPU you are converting from 265 to 264 and expecting smaller file sizes, but CPU you are going from 264 to 265?
If compression methods/codecs are equal, the hardware shouldn’t affect compression
Maybe I worded it poorly, I’m doing h264 -> h265 in both cases, but it increases file size when using iGPU.