Just now, Alfredo said:

Below is a script for the User Scripts plugin. It ensures that the CDI runtime is properly configured and that Docker is aware of the new devices, whether during a fresh boot or after an update.

Please don't do that.

This is not necessary! The plugin is now fixed!

BTW, I would strongly recommend that you use:

--gpus=all

@Alfredo & @xinud did you post something here? It seems I can't find your posts anymore?

17 minutes ago, ich777 said:

@Alfredo & @xinud did you post something here? It seems I can't find your posts anymore?

Yes, I deleted it — everything is now working with --device nvidia.com/gpu=all without my script, which is odd since both the plugin and Unraid were already up to date. Yesterday Plex was crashing, today it just works.

Thanks for your work on the plugin.

On 5/12/2026 at 3:09 AM, ich777 said:

Maybe but I can't tell, it would be best to create your own Forum entry there.

Are you sure, who says that? I'm not sure if the keyboard really hangs because he mentions that he thinks that it's hanging. However from what I see this is an Xid error 79 (GPU falls from the Bus - most of the times a incompatible BIOS setting or even outdated BIOS.

I don't think that you have an Xid error or do you?

https://docs.nvidia.com/deploy/xid-errors/analyzing-xid-catalog.html

just to update here my problem zfs mirror data corruption fixed i don't really sure what causing it, i try different sata port still same corruption report, no more corruption complain when i change docker system from cache0 to cache1

-tried rebuild docker system on cache0 still have corruption report

-move docker system to cache1 fixed

Note i have appdata and docker system on cache0 zfs mirror overlay2 data corruption complain on both drive.

2 hours ago, Gvon said:

just to update here my problem zfs mirror data corruption fixed i don't really sure what causing it, i try different sata port still same corruption report, no more corruption complain when i change docker system from cache0 to cache1

Probably this would be best to report in the General subforums since if this is a general issue that should be reported and this is the wrong place to report such things.

6 minutes ago, ich777 said:

Probably this would be best to report in the General subforums since if this is a general issue that should be reported and this is the wrong place to report such things.

sorry about that You're absolutely right.

On 5/13/2026 at 5:34 PM, ich777 said:

@Alfredo & @xinud did you post something here? It seems I can't find your posts anymore?

I was going to say that I got mine working using "--gpus=all" which is great.

but then was going to ask a follow-up question about how to specify a specific GPU for a container. However, further research I found I could use --gpus='"device=1"' which seems to work for me now as well

Thank you!

GTX 1660 Ti - --runtime=nvidia broken after upgrading to 7.3.0 (no HW transcoding anymore)

After upgrading from 7.2.4 to 7.3.0, my GTX 1660 Ti no longer works with Plex using --runtime=nvidia.

•  Container fails to start with “bad parameter” / “command failed” when using --runtime=nvidia or --gpus all

•  Works fine with no GPU parameters (software transcoding only)

•  docker run --rm --gpus all nvidia/cuda... test works successfully

•  Tested on drivers 580.159.03, 590.48.01, and 595.71.05

•  Plex-Media-Server container from Previous Apps

It was working perfectly on 7.2.4 (I think this is the version I left to go directly to 7.3.0). Seems like a compatibility issue with the new Docker version in 7.3.

Anyone else with cards (16xx series) having this issue?

2 hours ago, marlinz357 said:

Anyone else with cards (16xx series) having this issue?

As pointed oit in the previous threads use --gpus=all

Hello,

I'm trying to upgrade my server from Unraid 7.2.4 to 7.2.6 and I am getting these error messages telling me not to reboot:

Any idea why this is happening?
I have a GTX 1080 in the server.

50 minutes ago, bschaeff18 said:

Any idea why this is happening?

I would recommend to reboot, please note that your next reboot will take a bit longer since the plugin tries to download the driver and after the reboot you have to visit the plugin page and select for your GTX1080 the legacy driver version 580.x, click Update & Download, wait for the download to finish and then reboot once more.

30 minutes ago, ich777 said:

I would recommend to reboot, please note that your next reboot will take a bit longer since the plugin tries to download the driver and after the reboot you have to visit the plugin page and select for your GTX1080 the legacy driver version 580.x, click Update & Download, wait for the download to finish and then reboot once more.

That worked, thanks!

Having a similar issue, but with 2x 3060 gpus. After upgrading to 7.2.6, dockers with GPU acceleration ON no longer start.

Environment

• Unraid 7.3.0

• 2 × RTX 3060 (seen fine by host)

• Nvidia‑Driver plugin: 580.159.03, nvidia-smi shows CUDA 13.0

Problem

• On host, nvidia-smi works and both GPUs look healthy.

• Any Docker container with GPU access fails during init with:

  docker: Error response from daemon: failed to create task for container: failed to create shim task: OCI runtime create failed: runc create failed: unable to start container process: error during container init: error running createContainer hook #N: exit status 2, stdout: , stderr: /bin/sh: 0: cannot open /proc/self/fd/12: No such file

What I’ve already tried

• I tried removing the plugin, restarting and then installing the driver again with different branches but same issue 580.159.03, 590.48.01, and 595.71.05.

• Tried starting whisper-asr-webservice or ollama using --runtime=nvidia or --gpus all in Extra Parameters but ran into same error on hook

• Restarted Docker and the whole server multiple times.

Ask

• Is this a known issue with NVIDIA Container Toolkit / libnvidia-container on Unraid 7.3.0 (GPU works on host but all GPU Docker containers die in the Nvidia hook)?

• What’s the correct way on Unraid to fully reset/reinstall the Nvidia container runtime config so these hooks stop failing ?

6 hours ago, viniciustp said:

Ask

Please post your Diagnostics.

4 hours ago, ich777 said:

Please post your Diagnostics.

Here it is. Thank you in advance for the help!

diagnostics-20260516-1858.zip

Help please. I run two 3090's with ollama. Both were working fine yesterday. Today updated to 7.3, now when I load an llm, it only works in one of my gpus. It should span across both like it did before. If I load a second llm, it will load on the other card, but one will not load across like it was yesterday. If I load the one llm, itis crawling, going to system ram. No idea why. the ollama has all gpus tag, and again, it was fine yesterday. both show up in nvidia smi. Both show up and are checked in the nvida plug in.loki-diagnostics-20260518-0110.zip

Any thoughts? ideas? is this a 7.3 issue? or?

Thanks in advance.

1 hour ago, Renew said:

Any thoughts? ideas? is this a 7.3 issue? or?

Just to male sure, Ollama didnmt receive an update correct? Just to rule put that possibility since Ollama receives updates often.

Do you also have the docker run conmand from your Ollama container?

I'll look into your Diagnostics later.

EDIT: I don't see anything obvious from your Diagnostics @Renew. Can you issue nvidia-smi in the container and see if both GPUs are showing up.

You can also try to create an environment variable with the Key: CUDA_VISIBLE_DEVICES and as Value set: 0,1 in your Ollama Docker template.

cant downgrade to 580 driver: Can't download Nvidia Driver Package v580.159.03. How to do it manually ?

UPDATE: NVMND i reinstalled the plugin and it works

Edited May 19May 19 by Gee1

On 5/16/2026 at 12:30 AM, ich777 said:

As pointed oit in the previous threads use --gpus=all

--gpus=all causes the container to fail to start.

The container only starts if Extra Parameters is left blank, with Nvidia_visible_devices set to my GPU UUID and Nvidia_driver_capabilities set to all (which worked fine with --runtime=nvidia before going to 7.3 from 7.2.4).

So without an extra parameter (hw) transcoding does not happen, only slams my processor h.264.

2 hours ago, marlinz357 said:

--gpus=all causes the container to fail to start.

Diagnostics please when the driver is installed and after trying to run a container.

On 5/17/2026 at 12:27 PM, viniciustp said:

Here it is. Thank you in advance for the help!

From the diagnostics that you've posted I see anything obvious.

To fully reinstall the driver:
1. Remove the plugin

2. Reboot

3. Install the plugin again

4. Select your preferred driver

5. Reboot

On 5/19/2026 at 12:01 AM, ich777 said:

Diagnostics please when the driver is installed and after trying to run a container.

Thank you for your help

diagnostics-20260519-2054.zip

Hi all,

I’m having trouble getting my NVIDIA GeForce RTX 5060 Ti (GB206, Blackwell) working with the Nvidia Driver plugin on Unraid.

System:

• Unraid OS: 7.2.4

• GPU: NVIDIA GeForce RTX 5060 Ti (lspci shows GB206)

• Motherboard: ASUS X99‑A/USB 3.1 (LGA2011‑v3)

• Current Nvidia Driver (per plugin): 595.71.05 (open‑source)

Symptoms:

• The Nvidia Driver plugin page shows:

  • Nvidia Driver Version: 595.71.05

  • Open Source Kernel Module: Yes

  • Installed GPU(s): “No devices were found”

• lspci clearly sees the card:

01:00.0 VGA compatible controller: NVIDIA Corporation GB206 [GeForce RTX 5060 Ti] (rev a1)

01:00.1 Audio device: NVIDIA Corporation GB206 High Definition Audio Controller (rev a1)

lsmod shows NVIDIA modules loaded (no nouveau):

nvidia_uvm

nvidia_drm

nvidia_modeset

nvidia

nvidia-smi on the host returns:

No devices were found

Key NVRM errors from dmesg/syslog:

I’m seeing repeated initialization failures like:

NVRM: GPU0 kbusVerifyBar2_GB202: MMUTest BAR2 readback VA = 0x0 returned garbage ...

NVRM: GPU0 nvAssertOkFailedNoLog: Assertion failed: Generic memory error [NV_ERR_MEMORY_ERROR]

NVRM: GPU0 RmInitNvDevice: *** Cannot initialize the device

NVRM: GPU 0000:01:00.0: RmInitAdapter failed! (0x24:0x72:1344)

NVRM: GPU 0000:01:00.0: rm_init_adapter failed, device minor number 0

...

NVRM: GPU0 _kgspBootGspRm: unexpected WPR2 already up, cannot proceed with booting GSP

NVRM: GPU0 RmInitAdapter: Cannot initialize GSP firmware RM

NVRM: GPU 0000:01:00.0: RmInitAdapter failed! (0x62:0x40:2168)

NVRM: GPU 0000:01:00.0: rm_init_adapter failed, device minor number 0

These repeat many times during boot and whenever the driver appears to attempt initialization.

What I’ve already tried:

• Used the Nvidia Driver plugin to switch driver versions within the 595.x family (open‑source and production where available).

• Uninstalled the Nvidia Driver plugin, rebooted, reinstalled it from Apps, selected the recommended driver, clicked Update & Download, and rebooted again.

• Verified the card is visible via lspci, that nvidia modules are loaded, and that nouveau is not present.

• Confirmed no VMs are configured to pass through this GPU.

The behavior is always the same: modules load, lspci sees the GPU, but RmInitAdapter fails and nvidia-smi reports no devices.

Why I think this might be a driver/Blackwell issue:

Searching around, I found similar reports on Linux for new RTX 50‑series / Blackwell cards with errors like _kgspBootGspRm: unexpected WPR2 already up, cannot proceed with booting GSP and RmInitAdapter failed (e.g., RTX 5090/5080 threads on NVIDIA’s forums and an issue titled “RmInitAdapter failed on Fedora 43 for RTX 5060 Ti”). These look very similar to what I’m seeing on Unraid

23 hours ago, marlinz357 said:

Thank you for your help 168.24 kB · 1 download

diagnostics-20260519-2054.zip

ich777, please let me know when you get a chance to view my diagnostics above.

19 hours ago, marlinz357 said:

ich777, please let me know when you get a chance to view my diagnostics above.

Please read the first post (big red text):

May 18 21:24:42 Gandalf emhttpd: /usr/local/emhttp/plugins/user.scripts/backgroundScript.sh "/tmp/user.scripts/tmpScripts/unlock nvidia/script" >/dev/null 2>&1