48 minutes ago, samsausages said:

Is blackwell supposed to work? I just got a new card and it's not showing up. AI tells me blackwell isn't supported in unraid yet. Can someone confirm before I pull the card out?|||

Sorry, but I don't see a Nvidia GPU in your system, at least that what's lspci tells me.

Hey folks. I keep having this reoccuring issue were my nvidia 5050 will fail to be recognized by the driver after a few days. I'm not sure what the cause is, this never happened before with my 1660. I have reinstalled the driver plugin a few times, as well as made sure that the open source driver is being used. I have attached diagnostics to this message, appreciate any help I can get.

dingleberry-diagnostics-20260205-1850.zip

6 hours ago, DingusKahn said:

Hey folks. I keep having this reoccuring issue were my nvidia 5050 will fail to be recognized by the driver after a few days. I'm not sure what the cause is, this never happened before with my 1660. I have reinstalled the driver plugin a few times, as well as made sure that the open source driver is being used. I have attached diagnostics to this message, appreciate any help I can get.

dingleberry-diagnostics-20260205-1850.zip

This is your error:

Feb 5 18:47:49 Dingleberry kernel: NVRM: nvAssertOkFailedNoLog: Assertion failed: Out of memory [NV_ERR_NO_MEMORY] (0x00000051) returned from nvStatus @ message_queue_cpu.c:242

Feb 5 18:47:49 Dingleberry kernel: NVRM: _kgspInitRpcInfrastructure: GspMsgQueueInit failed

Feb 5 18:47:49 Dingleberry kernel: NVRM: kgspConstructEngine_IMPL: init RPC infrastructure failed

Feb 5 18:47:49 Dingleberry kernel: NVRM: osInitNvMapping: *** Cannot attach gpu

Feb 5 18:47:49 Dingleberry kernel: NVRM: RmInitAdapter: osInitNvMapping failed, bailing out of RmInitAdapter

Feb 5 18:47:49 Dingleberry kernel: NVRM: GPU 0000:2b:00.0: RmInitAdapter failed! (0x22:0x51:763)

Feb 5 18:47:49 Dingleberry kernel: NVRM: GPU 0000:2b:00.0: rm_init_adapter failed, device minor number 0

Feb 5 18:47:52 Dingleberry kernel: NVRM: nvCheckOkFailedNoLog: Check failed: Out of memory [NV_ERR_NO_MEMORY] (0x00000051) returned from memdescAllocInternal(pMemDesc) @ memdesc.c:1359

Please make sure that you have:

1. The latest BIOS version

2. Above 4G Decding enabled in your BIOS

3. Resizable BAR Support enabled in your BIOS

I also compiled a new Open Source driver yesterday that you can try: v590.48.01

Hi everyone, 👋

I already talked about this problem on the Discord server, and they suggested that I open a post here.

My issue is the following: I recently installed my first GPU, an RTX A2000 Ada Gen with 16GB. When the Nvidia Driver Plugin is not installed, my CPU and RAM usage are completely normal.

However, as soon as I install the Nvidia Driver Plugin in order to use the GPU, my RAM usage almost doubles, and one CPU core is constantly at 100% usage. Overall CPU load also increases significantly.

lom-001-diagnostics-20260206-2222.zip

Important is I tested it for the latest and the 580. driver!

I would really appreciate any help with this issue.

Thank you!
🙏🏻 Nicalion

8 hours ago, Nicalion said:

I would really appreciate any help with this issue.

Please try to uninstall the GPU Statistics plugin and see if that resolves your issue.

GPU Statistics calls nvidia-smi every N seconds (depending on how it's configured in the plugin) and that could cause your CPU spikes.

I can't tell much about the RAM usage, however please do keep in mind that Unraid runs from RAM and the driver is also installed to RAM, compressed it has about 400MB so to speak that should be about 700 to 800MB decompressed if I'm not mistaken and that will also go into RAM.

On 2/1/2026 at 7:52 AM, ConnerVT said:

If you are still using the Quadro P2000 you will want to stay on 580.xx.xx. Nvidia has obsoleted the Pascal based cards and is removing support for them from 590.xx.xx.

Best to manually select a 580.xx.xx version (latest one is fine).

Thanks again for all the help.
Managed to grab a NVIDIA RTX A400 and was able to successfully install it today. Everything is working well and updated to the new OS using the v590.48.01 drivers.

Appreciate the recommendations and assistance from everyone including @ich777 should be a nice GPU change for the coming years.

On 2/1/2026 at 7:52 AM, ConnerVT said:

If you are still using the Quadro P2000 you will want to stay on 580.xx.xx. Nvidia has obsoleted the Pascal based cards and is removing support for them from 590.xx.xx.

Best to manually select a 580.xx.xx version (latest one is fine).

@ich777 I am facing the same issue where my server "forgets" my device (RTX 5060 Ti).

I already checked:

• Latest BIOS version installed

• "Above 4G Decoding" enabled

• "Resizable BAR Support" enabled

i updated the Open Source driver today to the new version: v590.48.01. if the problem persists i share my logs

1 hour ago, value1338 said:

i updated the Open Source driver today to the new version: v590.48.01. if the problem persists i share my logs

Please post your Diagnostics after your GPU is "forgotten".

Thanks for releasing Nvidia Driver Version: 590.48.01. No more nvidia error (NV_ERR_NO_MEMORY) and missing GPU (5070ti) when quitting Gaming Windows VM. So it's really a fix from drivers

Edited February 10Feb 10 by sausuke

16 minutes ago, sausuke said:

Gaming Windows VM. So it's really a fix from drivers

Sorry, but this has nothing to do with the Nvidia Driver since if you are using it in a VM it is actually discouraged to install the Nvidia Driver Plugin (also see the first page).

33 minutes ago, ich777 said:

Sorry, but this has nothing to do with the Nvidia Driver since if you are using it in a VM it is actually discouraged to install the Nvidia Driver Plugin (also see the first page).

I just feedback based on my usage here. I'm using 3070 for several years and I didn't have problem using it that way (no passthrough via VM) until I upgraded to 5000 series I experienced these problems. I tried the same troubleshooting and those errors don't appear in logs anymore. It's just my observation 😁 I want to use it on both unraid and gaming VM, I don't want to passthrough. Glad this problem is fixed now hehe

Hello, hoping someone can help.

I upgraded my server from 6.12.13 to 7.2.3 last night. During the update I got some messages about the NVIDIA plugin I use for transcoding with my MSI GTX 1050 Ti. I followed the instructions, but after reboot I started getting errors from the plugin.

First error right after reboot:

NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

After that I’m now consistently seeing this:

Failed to initialize NVML: Driver/library version mismatch

NVML library version: 590.48

It looks like parts of the driver are on different versions and they’re not playing nice.

Here’s what I’ve tried so far:

• Switched between Latest, Production, and several of the Available Versions in the NVIDIA plugin.

• Rebooted after each driver change.

• Tried installing a single driver version, rebooting, then trying a different one.

• Confirmed the GTX 1050 Ti shows up under System Devices.

• Uninstalled the NVIDIA plugin, rebooted, reinstalled it, and rebooted again.

No matter what I do I still end up with the Driver/library version mismatch error when I check the plugin under Installed GPUs.

At this point I’m thinking something from the old driver is still hanging around after the OS upgrade, but I’m not sure how to fully clear it out without messing something else up.

Any ideas on how to properly clean this up and get the driver versions back in sync?

Thanks!

unraid-diagnostics-20260211-1538.zip

Edited February 12Feb 12 by Tbigs2011

17 hours ago, Tbigs2011 said:

Hello, hoping someone can help.

I upgraded my server from 6.12.13 to 7.2.3 last night. During the update I got some messages about the NVIDIA plugin I use for transcoding with my MSI GTX 1050 Ti. I followed the instructions, but after reboot I started getting errors from the plugin.

First error right after reboot:

NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

After that I’m now consistently seeing this:

Failed to initialize NVML: Driver/library version mismatch

NVML library version: 590.48

It looks like parts of the driver are on different versions and they’re not playing nice.

Here’s what I’ve tried so far:

• Switched between Latest, Production, and several of the Available Versions in the NVIDIA plugin.

• Rebooted after each driver change.

• Tried installing a single driver version, rebooting, then trying a different one.

• Confirmed the GTX 1050 Ti shows up under System Devices.

• Uninstalled the NVIDIA plugin, rebooted, reinstalled it, and rebooted again.

No matter what I do I still end up with the Driver/library version mismatch error when I check the plugin under Installed GPUs.

At this point I’m thinking something from the old driver is still hanging around after the OS upgrade, but I’m not sure how to fully clear it out without messing something else up.

Any ideas on how to properly clean this up and get the driver versions back in sync?

Thanks!

unraid-diagnostics-20260211-1538.zip

Oddly enough I had the server running for a little while (5 or so hours) with the production option chosen and I randomly got a notification that version v580.126.09 was found and downloaded and that I'd have to restart my server.

I did that and now my GPU is recognized properly in the plugin. Not sure what the deal was but I'm happy it's fixed. I really hope I don't have to deal with this when I replace this GPU in a few weeks after an upgrade I have planned.

Recently I installed an old Nvidia 1080 graphics card, and the system became unstable. So unstable that it started either hanging completely or rebooting multiple times per day. Everything pointed to the problem being with the NVIDIA drivers, including the monitoring app consuming 100% of at least one CPU. I could only tell when I could login to the actual machine, which I don't have it really setup for, unfortunately.

I removed the NVIDIA driver, and disabled (for now) the one app that used it, and my system has been stable for the past 15 hours, so without a lot of data to back that up, I'm fairly convinced it's something to do with the NVIDIA driver support.

I've attached a troubleshooting dump, if anyone has the time to take a peak.

Thanks.

tower-diagnostics-20260211-1924.zip

2 hours ago, hhubris said:

Everything pointed to the problem being with the NVIDIA drivers, including the monitoring app consuming 100% of at least one CPU.

Most of the times it is simply enough to Uninstall the GPU Statistics plugin or set the query interval to something reasonable like 5000ms or even 10000ms.

Please go also yourself through your syslog, it seems you have a massive issue with the UPS plugin too.

I recently installed a 5070 card and moved over to the Open Source version of the driver and NVIDIA-SMI can't find the card. Any other suggestions?

Unraid 7.2.3

tower-diagnostics-20260213-1729.zip

On 2/14/2026 at 12:40 AM, ryan8382 said:

I recently installed a 5070 card and moved over to the Open Source version of the driver and NVIDIA-SMI can't find the card. Any other suggestions?

Sorry, I missed your message, however it seems your card is bound to VFIO, please unbind it an reboot:

Feb 13 17:00:16 Tower kernel: NVRM: GPU 0000:03:00.0 is already bound to vfio-pci.

Feb 13 17:00:16 Tower kernel: NVRM: The NVIDIA probe routine was not called for 1 device(s).

Feb 13 17:00:16 Tower kernel: NVRM: This can occur when another driver was loaded and

Feb 13 17:00:16 Tower kernel: NVRM: obtained ownership of the NVIDIA device(s).

Feb 13 17:00:16 Tower kernel: NVRM: Try unloading the conflicting kernel module (and/or

Feb 13 17:00:16 Tower kernel: NVRM: reconfigure your kernel without the conflicting

Feb 13 17:00:16 Tower kernel: NVRM: driver(s)), then try loading the NVIDIA kernel module

Feb 13 17:00:16 Tower kernel: NVRM: again.

Feb 13 17:00:16 Tower kernel: NVRM: No NVIDIA devices probed.

Feb 13 17:00:16 Tower kernel: nvidia-nvlink: Unregistered Nvlink Core, major device number 239

Edit - Never mind, figured it out... forgot I was running a custom kernel...

I currently getting the following errors in the install log:

Downloading plugin
plugin: installing: nvidia-driver.plg
Executing hook script: CA_preHook
Clearing Community Applications plugin cache
Executing hook script: pre_plugin_checks
plugin: downloading: nvidia-driver-2026.01.16.txz ...
plugin: downloading: nvidia-driver-2026.01.16.txz ... 100%
plugin: downloading: nvidia-driver-2026.01.16.txz ... done


+========================================================================================+
| Installing new package /boot/config/plugins/nvidia-driver/nvidia-driver-2026.01.16.txz |
+========================================================================================+

Verifying package nvidia-driver-2026.01.16.txz.
Installing package nvidia-driver-2026.01.16.txz:
PACKAGE DESCRIPTION:
Package nvidia-driver-2026.01.16.txz installed.

-----ERROR - ERROR - ERROR - ERROR - ERROR - ERROR - ERROR - ERROR - ERROR------
---Can't get latest Nvidia driver version and found no installed local driver---
Removing package: nvidia-driver-2026.01.16
plugin: run failed: '/bin/bash' returned 1
Executing hook script: CA_postHook
Clearing Community Applications plugin cache
Executing hook script: post_plugin_checks
Plugin installed

Any suggestions? I've deleted any old remnants of previous driver//app installs. Card is an RTX3090 and not isolated via VFIO.

Edited February 22Feb 22 by clanger

I highly doubt this, but other then downgrading my Unraid OS, is there anyway to use the previous drivers? I have a Nvidia GT 730 and it uses linux driver version 470.256.02.

I remember seeing it somewhere i may have to setup up the driver file to use that version and kernel that unraid is currently using. If that's the case, is there a tutorial to follow? I have never done that before.

47 minutes ago, sneville83 said:

I remember seeing it somewhere i may have to setup up the driver file to use that version and kernel that unraid is currently using. If that's the case, is there a tutorial to follow? I have never done that before.

May I ask for what do you use the card actually?

And yes, currently the old legacy driver doesn't compile against recent Kernels which all currently available Unraid versions are using.

However there is a patch out there for the old legacy driver to make them work again, however I already tested that and I'm pretty certain you don't want to crash your server often correct? These patches make the driver work but the server becomes completely unstable.

Sorry but the blame goes to Nvidia since they now deprecated these cards.

39 minutes ago, ich777 said:

May I ask for what do you use the card actually?

I have a Nvidia GT 730 in my server right now. Been trying to find a relatively cheap that is more up to date for transcoding.

1 minute ago, sneville83 said:

Been trying to find a relatively cheap that is more up to date for transcoding.

If you are searching something cheap then please look maybe for a Nvidia T400/T600/T1000, you could also get away with a Intel Arc Card even an A380 will do the job just fine and the A380 is also capable of transcoding AV1, as long as the software you are using is capable of transcoding AV1, with the benefit of not having to install a driver.
However the idle power draw is a bit higher in idle with the Intel card from what I know.

Just now, ich777 said:

If you are searching something cheap then please look maybe for a Nvidia T400/T600/T1000, you could also get away with a Intel Arc Card even an A380 will do the job just fine and the A380 is also capable of transcoding AV1, as long as the software you are using is capable of transcoding AV1, with the benefit of not having to install a driver.
However the idle power draw is a bit higher in idle with the Intel card from what I know.

The transcoding is basically done through Plex. I was also thinking of using the GPU to create subtitles through Whisper.

I tried setting up transcoding videos "manually" to save space on my media server, but got frustrated trying to find the best settings to use that save space and still be usable on plex.

Used T400 are available on eBay for as low as $50USD.

I recently bought a new A400 for $180, delivered. Replaced a P400. Figure I'm good for several years now.

There really isn't anything a GT730 can be used for anymore. Even Win10 required a GT740 as the minimum GPU it would support.

On 1/22/2026 at 9:55 AM, ich777 said:

I compile the Open Source drivers when a new Unraid version is released.

So to speak, when the next Unraid version is released you should see the Open Source 590.x series driver.

Any way we can install an older version of the Open Source drivers?

I'm trying to use an RTX 5060 ti (which requires the OpenSource driver) with a GTX 1660 Super, splitting the GPUs between different containers, but with the current Open Source driver, the 1660 doesn't support the Compute module, whereas with the prodction driver, the 5060 isn't even supported.

Downgrading the driver might solve the 1660 issue (see here), but I need to downgrade to another Open Source version.

Thanks

Edited February 25Feb 25 by gshlomi