[GH-ISSUE #410] Couldn't start Wolf after updating host PC's Nvidia drivers - was my fix correct? #270

New issue

Open

opened 2026-10-15 12:59:03 +00:00 by bsakaguchi · 1 comment

bsakaguchi commented

2026-10-15 12:59:03 +00:00

Owner

Originally created by @TransparentDuck on GitHub (May 8, 2026).
Original GitHub issue: https://github.com/games-on-whales/wolf/issues/410

Hi all. I was having a problem with Nvidia driver versions which I seem to have fixed. I'm not sure if it's just because of something I did wrong, so I'm putting the problem and my workaround here in case it's of use to anybody.

Problem:
Couldn't start Wolf after updating my host PC's Nvidia drivers. Error:

208 0x55806b535f20 ERROR nvcodec plugin.c:171:plugin_init: Failed to init cuda, cuInit ret: 0x323: CUDA_ERROR_SYSTEM_DRIVER_MISMATCH: system has unsupported display driver / cuda driver combination

Details

Debian 13 PC, Nvidia 3080Ti
Original driver: 550.163.01 (most recent available from Debian repo)
Updated driver: 595.71.05/CUDA Version: 13.2) (most recent available from Nvidia repo)

I had this working pre-upgrade. Steam was fine, but the UI for both Heroic and Lutris were hugely laggy (but the were fine once a game was started). I've opened another issue for that here:
https://github.com/games-on-whales/wolf/issues/409

So, I thought upgrading my Nvidia driver might help. All went fine, but after starting Wolf I got that error:
208 0x55806b535f20 ERROR nvcodec plugin.c:171:plugin_init: Failed to init cuda, cuInit ret: 0x323: CUDA_ERROR_SYSTEM_DRIVER_MISMATCH: system has unsupported display driver / cuda driver combination

With the help of a friendly LLM (GPT-5 mini on duck.ai), I worked out that the Nvidia driver version in the container was 550 and the host was 595, hence the error.

I tried going through the Nvidia (Manual) installation steps a few times to recreate the Nvidia container, but no luck.

What ended up working is rebuilding the Nvidia container using the host driver version, and including some bin dirs in the Wolf container PATH variable. Details:

# 1) Show host driver version
sudo nvidia-smi
cat /sys/module/nvidia/version

# 2) List files inside the driver volume to inspect NVML/lib versions
sudo docker run --rm -v nvidia-driver-vol:/usr/nvidia alpine ls -l /usr/nvidia/lib /usr/nvidia/lib64 /usr/nvidia/lib32 || true

# 3) If the volume contains mismatched libraries: remove the existing volume
sudo docker volume rm nvidia-driver-vol

# 4) Rebuild the gow/nvidia-driver image using the current host driver version
curl https://raw.githubusercontent.com/games-on-whales/gow/master/images/nvidia-driver/Dockerfile \
  | sudo docker build -t gow/nvidia-driver:host -f - --build-arg NV_VERSION=$(cat /sys/module/nvidia/version) .

# 5) Recreate and populate the driver volume from the rebuilt image
sudo docker create --rm --mount source=nvidia-driver-vol,destination=/usr/nvidia gow/nvidia-driver:host sh

# 6) Restart Docker daemon
sudo systemctl restart docker

# 7) Recreate the wolf service with no dependencies (use Compose V2)
sudo docker compose up -d --force-recreate --no-deps wolf

# 8) Ensure PATH inside the service includes the mounted driver bin (if not already set in compose)
# Add to docker-compose.yml under service 'wolf' -> environment:
#   - PATH=/usr/nvidia/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin

# 9) Then recreate the service:
sudo docker compose up -d --force-recreate --no-deps wolf

I did try lots of other things before this, so not sure if any of those contributed to the fix as well.

Was this a problem just for me, or has this affected other as well?

Cheers

Originally created by @TransparentDuck on GitHub (May 8, 2026). Original GitHub issue: https://github.com/games-on-whales/wolf/issues/410 Hi all. I was having a problem with Nvidia driver versions which I seem to have fixed. I'm not sure if it's just because of something I did wrong, so I'm putting the problem and my workaround here in case it's of use to anybody. Problem: Couldn't start Wolf after updating my host PC's Nvidia drivers. Error: > 208 0x55806b535f20 ERROR nvcodec plugin.c:171:plugin_init: Failed to init cuda, cuInit ret: 0x323: CUDA_ERROR_SYSTEM_DRIVER_MISMATCH: system has unsupported display driver / cuda driver combination ### Details Debian 13 PC, Nvidia 3080Ti Original driver: 550.163.01 (most recent available from Debian repo) Updated driver: 595.71.05/CUDA Version: 13.2) (most recent available from Nvidia repo) I had this working pre-upgrade. Steam was fine, but the UI for both Heroic and Lutris were hugely laggy (but the were fine once a game was started). I've opened another issue for that here: https://github.com/games-on-whales/wolf/issues/409 So, I thought upgrading my Nvidia driver might help. All went fine, but after starting Wolf I got that error: 208 0x55806b535f20 ERROR nvcodec plugin.c:171:plugin_init: Failed to init cuda, cuInit ret: 0x323: CUDA_ERROR_SYSTEM_DRIVER_MISMATCH: system has unsupported display driver / cuda driver combination With the help of a friendly LLM (GPT-5 mini on duck.ai), I worked out that the Nvidia driver version in the container was 550 and the host was 595, hence the error. I tried going through the Nvidia (Manual) installation steps a few times to recreate the Nvidia container, but no luck. What ended up working is rebuilding the Nvidia container using the host driver version, and including some bin dirs in the Wolf container PATH variable. Details: ``` # 1) Show host driver version sudo nvidia-smi cat /sys/module/nvidia/version # 2) List files inside the driver volume to inspect NVML/lib versions sudo docker run --rm -v nvidia-driver-vol:/usr/nvidia alpine ls -l /usr/nvidia/lib /usr/nvidia/lib64 /usr/nvidia/lib32 || true # 3) If the volume contains mismatched libraries: remove the existing volume sudo docker volume rm nvidia-driver-vol # 4) Rebuild the gow/nvidia-driver image using the current host driver version curl https://raw.githubusercontent.com/games-on-whales/gow/master/images/nvidia-driver/Dockerfile \ | sudo docker build -t gow/nvidia-driver:host -f - --build-arg NV_VERSION=$(cat /sys/module/nvidia/version) . # 5) Recreate and populate the driver volume from the rebuilt image sudo docker create --rm --mount source=nvidia-driver-vol,destination=/usr/nvidia gow/nvidia-driver:host sh # 6) Restart Docker daemon sudo systemctl restart docker # 7) Recreate the wolf service with no dependencies (use Compose V2) sudo docker compose up -d --force-recreate --no-deps wolf # 8) Ensure PATH inside the service includes the mounted driver bin (if not already set in compose) # Add to docker-compose.yml under service 'wolf' -> environment: # - PATH=/usr/nvidia/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin # 9) Then recreate the service: sudo docker compose up -d --force-recreate --no-deps wolf ``` I did try lots of other things before this, so not sure if any of those contributed to the fix as well. Was this a problem just for me, or has this affected other as well? Cheers

bsakaguchi commented

2026-10-15 12:59:04 +00:00

Author

Owner

@devexa commented on GitHub (May 29, 2026):

Happened to me when upgrading from 595.71.05 to 610.43.02.

OS: UNRAID 7.3.1
GPU: RTX 3080 FE
Installation method: Using official GoW UNRAID plugin

Applying the official fix from the docs fixed it for me :

https://games-on-whales.github.io/wolf/stable/user/troubleshooting.html#_recreating_the_nvidia_driver_volume

@devexa commented on GitHub (May 29, 2026): Happened to me when upgrading from 595.71.05 to 610.43.02. OS: UNRAID 7.3.1 GPU: RTX 3080 FE Installation method: Using official GoW UNRAID plugin Applying the official fix from the docs fixed it for me : https://games-on-whales.github.io/wolf/stable/user/troubleshooting.html#_recreating_the_nvidia_driver_volume