Goldeye but stuck on old nvidia drivers?

Apps and Virtualization
1

Hey fam, My 5060ti is crashing my apps.

I yoinked the B50 and replaced it with a 5060ti. The consensus seems to be that nvdia and goldeye is the way to go…

The card is in, and it shows up as an nvidia vga device in lspci output, and shows up as a gpu that i can isolate for vms in the advanced settings and as a card I can use for my apps in the individual app settings, but not in nvidia-smi output.

When I select it for use in my apps (jellyfin/immich/ollama) the apps wont start, and throw this error:

EFAULT] Failed ‘up’ action for ‘immich’ app, please check /var/log/app_lifecycle.log

checking that log:

(ERROR) app_lifecycle.compose_action():56 - Failed ‘up’ action for ‘immich’ app: Network ix-immich_default Creating\n Network ix-immich_default Created\n Container ix-immich-permissions-1 Creating\n Container ix-immich-permissions-1 Created\n Container ix-immich-redis-1 Creating\n Container ix-immich-machine-learning-1 Creating\n Container ix-immich-pgvecto_upgrade-1 Creating\n Container ix-immich-redis-1 Created\n Container ix-immich-machine-learning-1 Created\n Container ix-immich-pgvecto_upgrade-1 Created\n Container ix-immich-pgvecto-1 Creating\n Container ix-immich-pgvecto-1 Created\n Container ix-immich-server-1 Creating\n Container ix-immich-server-1 Created\n Container ix-immich-permissions-1 Starting\nError response from daemon: failed to create task for container: failed to create shim task: OCI runtime create failed: could not apply required modification to OCI specification: error modifying OCI spec: failed to inject CDI devices: unresolvable CDI devices runtime.nvidia.com/gpu=GPU-56022208-69bf-9f59-0ce1-1406e51f9b36: unknown\n [2025/12/09 23:26:29] (ERROR) app_lifecycle.compose_action():56 - Failed ‘up’ action for ‘immich’ app: Network ix-immich_default Creating\n Network ix-immich_default Created\n Container ix-immich-permissions-1 Creating\n Container ix-immich-permissions-1 Created\n Container ix-immich-machine-learning-1 Creating\n Container ix-immich-pgvecto_upgrade-1 Creating\n Container ix-immich-redis-1 Creating\n Container ix-immich-pgvecto_upgrade-1 Created\n Container ix-immich-pgvecto-1 Creating\n Container ix-immich-redis-1 Created\n Container ix-immich-machine-learning-1 Created\n Container ix-immich-pgvecto-1 Created\n Container ix-immich-server-1 Creating\n Container ix-immich-server-1 Created\n Container ix-immich-permissions-1 Starting\nError response from daemon: failed to create task for container: failed to create shim task: OCI runtime create failed: could not apply required modification to OCI specification: error modifying OCI spec: failed to inject CDI devices: unresolvable CDI devices runtime.nvidia.com/gpu=GPU-56022208-69bf-9f59-0ce1-1406e51f9b36: unknown\n truenas_admin@truenas[~]$

I did a bunch of AI guided troubleshooting because I’m a midwit, and it looks like i might still be running the old drivers?

# Check firmware versions ls -la /lib/firmware/nvidia/ga102/gsp/*.bin zsh: command not found: # total 30083 drwxr-xr-x 2 root root 6 Nov 14 16:45 . drwxr-xr-x 7 root root 7 Nov 14 16:45 .. -rw-r–r-- 1 root root 59768 May 28 2025 booter_load-535.113.01.bin -rw-r–r-- 1 root root 39544 May 28 2025 booter_unload-535.113.01.bin -rw-r–r-- 1 root root 20588 May 28 2025 bootloader-535.113.01.bin -rw-r–r-- 1 root root 38061600 May 28 2025 gsp-535.113.01.bin zsh: command not found: # -rw-r–r-- 1 root root 59768 May 28 2025 /lib/firmware/nvidia/ga102/gsp/booter_load-535.113.01.bin -rw-r–r-- 1 root root 39544 May 28 2025 /lib/firmware/nvidia/ga102/gsp/booter_unload-535.113.01.bin -rw-r–r-- 1 root root 20588 May 28 2025 /lib/firmware/nvidia/ga102/gsp/bootloader-535.113.01.bin -rw-r–r-- 1 root root 38061600 May 28 2025 /lib/firmware/nvidia/ga102/gsp/gsp-535.113.01.bin truenas_admin@truenas[~]$

I might have upgraded to goldeye before yanking my old 1080ti, but that shouldn’t matter, should it?

I went into the app settings and did the “un-check install nvidia drivers, reboot, re-check install nvidia drivers” dance a couple of times, sometimes throwing in extra reboots just for flavor, but selecting the nvidia card in my app settings still halts my current and newly installed apps from starting.

At this point, he Machine Spirit has started to recommend fixes involving rm -rf, and other things those nice men on the podcast seem to frown on, signaling I was now deep enough to beg help from real humans…

2

What on earth makes you think that. I would have said the B50 is a far superior card for transcoding, which is what I assume you are doing

3

Eh, a few reasons.

  1. Ease of implementation. 1080ti was fully plug and play and I had been pulling my hair out with the B50 for a week. Boy do I feel silly now.
  2. The B series cards wont be supported in truenas until the next major release in the best case scenario, and I couldnt pass the B50 through to a VM using a newer kernel for motherboard reasons I dont fully understand.
  3. The A series cards are better for transcode and supported now, but the cost save isnt worth the performance hit on ollama related stuff. I hate that CUDA is so OP, but it is OP…
  4. Last but not least: I got a good microcenter deal on the 5060ti.