Hardware decoding with Nvidia GPU in Docker container

[ripcityhandyman]

I recently put an Nvidia GTX 1060 6GB in my server. Up until now, I've been using QSV on my Intel Pentium Gold G5400.

I have followed the guide to allow Channels DVR to access the GPU from within the container: Hardware transcoding inside Docker using Nvidia GPU

Unfortunately, although NVENC is being used for hardware encoding, the decoding is using software decoding instead, and pins the CPU (which results in worse performance than QSV, which was used for the entire transcode).

I FileFlows setup in another Docker container for video conversion tasks, and it is able to use the GPU for both encoding and decoding.

Is lack of NVDEC support a known limitation of Channels or is there something wrong with my setup?

[tmm1]

Known limitation. What kind of content are you seeing this issue with?

[ripcityhandyman]

Here are the details for a couple recordings I've tried.

1080p H.264

4k H.265

The GTX 1060 has hardware decoding support for both codecs, so it would be great if Channels could make use of it. If not, I will swap back to QSV for now.

Thanks!