DeviceRequests=null Devices=null

Posted by poopooskin1

My question: DeviceRequests=null Devices=null SparkBox AI replied: It’s not working in Jellyfin yet. The CUDA test proves Docker can give a test container access to the RTX 5070 Ti. But DeviceRequests=null and Devices=null show the Jellyfin container itself wasn’t given GPU access. Doctor also reports the NVIDIA driver isn’t available inside the WSL distro. For now, to avoid playback failing when a video needs transcoding, open Jellyfin’s Dashboard → Playback → Transcoding, set Hardware acceleration to None, and save. Don’t install an Ubuntu NVIDIA kernel driver inside WSL based on Doctor’s generic suggestion—WSL gets GPU support from Windows. I can’t tell from these checks why SparkBox didn’t attach the GPU to Jellyfin under WSL. If you want to pursue NVIDIA transcoding there, use Share to demox so Tom can investigate that specific setup. What I still need help with: I am trying to get hardware acceleration working on Jellyfin. I'm using WSL2 with a RTX 5070ti. Let me know of any specific information you might need in order to help me. Thank you in advance for any help!

3 replies

Chris wrote:

Got it — this is an automatic acknowledgment from Chris, your AI support representative. Your post is in the support queue, and we’ll get to it as soon as we can. Legend members receive priority replies. If you’d like faster support while helping fund the project’s goals and continued development, consider becoming a Legend backer. <!-- chris-support-queue-ack:v1 --

Chris wrote:

I'm sorry—the earlier answer treated the CUDA test as proof that GPU passthrough should work in Jellyfin and told you to change Jellyfin settings before we knew the cause. SparkBox only adds the NVIDIA GPU to Jellyfin when its NVIDIA hardware-acceleration option is enabled and the Linux environment running Docker has both a loaded NVIDIA driver and Docker's NVIDIA runtime; the test container alone doesn't establish those conditions, and I can't tell yet which one is missing in your WSL setup. One check: what SparkBox version are you running, and is this Docker Desktop with WSL integration or Docker Engine installed inside WSL? Please don't install a Linux driver or change Jellyfin settings yet. I've corrected what I know so the unsupported conclusion isn't repeated.

Chris wrote:

An update: we found why Jellyfin wasn't getting your card, and it was on our side. WSL2 never loads a Linux NVIDIA driver, because the card comes from the Windows driver. SparkBox was checking for a Linux driver before handing the card to Jellyfin, so on WSL2 it always gave up and used the processor. Your CUDA test working was the right sign. SparkBox 1.6.915, out now, checks for the card the way WSL2 provides it. The checkup also no longer tells you to install a Linux driver, which you should never do inside WSL. This applies when Docker runs inside your WSL distro with NVIDIA's container toolkit, which your CUDA test suggests. In WSL: 1. sudo sparkbox upgrade 2. Make sure Settings → Hardware Transcoding is set to NVIDIA, then run sudo sparkbox up 3. Check: sudo docker inspect -f '{{.HostConfig.Runtime}}' sb-jellyfin-media should print nvidia After that, Jellyfin usually switches itself to NVIDIA encoding a few minutes later. If Dashboard → Playback → Transcoding still says None (for example because you changed it earlier), set it to NVIDIA NVENC and save. Legend boxes get 1.6.915 right away. Other boxes get it within about a week. If step 3 prints something other than nvidia, paste that line and the output of sudo sparkbox version here.