2 hours ago, Altheran said:

For anyone loosing access to their GPU randomly, here is what I did : Losing my GTX1660 GPU randomly X days/weeks ? after a reboot. - General Support - Unraid

I would have loved to have this post in the proper support thread, so to speak here, or at least a summary what your issue was.

Issue with that is that such information should be in the support thread and other plugin related issues like a driver not working otherwise it will be harder for other people to solve their issues when plugin related issues are spread accross the forums.

Just saying.

On 11/7/2025 at 2:02 PM, ich777 said:

The OpenSource driver is only compiled when a new Unraid version is released.

May I ask for what you are using the GPU for and on what Unraid version you are on?

Main unraid version or also the smaller one *.*.2?

Useing a RTX5050, mainly used for encoding. But nvidia has some trouble with atleast current driver, so when useing tune uhq, it segfaults on some sources

On version 7.2 :)

Edited November 10, 2025Nov 10 by raysv
clarification

18 hours ago, raysv said:

Useing a RTX5050, mainly used for encoding. But nvidia has some trouble with atleast current driver, so when useing tune uhq, it segfaults on some sources

And what driver do you need?

https://www.nvidia.com/en-us/drivers/unix/

On 11/10/2025 at 1:19 PM, Altheran said:

For anyone loosing access to their GPU randomly, here is what I did : Losing my GTX1660 GPU randomly X days/weeks ? after a reboot. - General Support - Unraid

Just wanted to say thank you this fixed it for me.

On 11/10/2025 at 3:48 PM, ich777 said:

I would have loved to have this post in the proper support thread, so to speak here, or at least a summary what your issue was.

Issue with that is that such information should be in the support thread and other plugin related issues like a driver not working otherwise it will be harder for other people to solve their issues when plugin related issues are spread accross the forums.

Just saying.

Just wanted to let you know this i was getting the same NVRM: GPU 0000:09:00.0: RmInitAdapter failed! message in my logs for the last couple months on driver 580.82.09 running a Quadro T1000 8Gb which made the GPU unrecognized in nvidia-smi and the rest of the system.

I followed what was in Altheran's quoted comment. BIOS settings were mostly already what ChatGPT had advised. The only thing that needed changing was PCIE Gen3.0 from Auto. ASPM was disabled in BIOS.

Added the pcie_aspm=off line to syslinux.cfg

and

# Ensure NVIDIA driver is initialized & kept warm between clients

modprobe nvidia

/usr/bin/nvidia-smi -pm 1

# Optional warm-up query (fails fast if device absent)

/usr/bin/nvidia-smi -L || true

to the bottom of /boot/config/go

Been up for 5 days and have not seen the RmInitAdapter failed! message since. FYI

9 minutes ago, averagejoeshmoe said:

# Ensure NVIDIA driver is initialized & kept warm between clients

modprobe nvidia

This is not necessary because the driver is load on boot.

10 minutes ago, averagejoeshmoe said:

/usr/bin/nvidia-smi -pm 1

# Optional warm-up query (fails fast if device absent)

/usr/bin/nvidia-smi -L || true

Again, please don't use this commands, this is the old way of enabling persistenced and pretty much outdated, please use this command instead:

nvidia-persistenced

that's all you need in your go file.

Not sure if this is driver related, but I've been having some strange issues with my 5060 Ti. I was on 7.1.4 and it was working fine, but the issue popped up about a week or two ago. I just updated to 7.2.0 and with it the Nvidia driver (was on v575.57.08, now on v580.95.05) to see if that would fix it, but no dice.

Before the issue popped up, the GPU would idle at 5-6W at P8 with power throttling on. Suddenly, it is now "idling" at 10-22W, with the fan spun up to 42%, the load between 0-2%, and P0 power state. I check nvidia-smi and nvtop and they show nothing is using the GPU. SMI shows 10W, but the GPU statistics widget on the dashboard shows 22W.

nvidia-smi
Sun Nov 16 22:20:27 2025       
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 580.95.05              Driver Version: 580.95.05      CUDA Version: 13.0     |
+-----------------------------------------+------------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  NVIDIA GeForce RTX 5060 Ti     Off |   00000000:01:00.0 Off |                  N/A |
| 45%   26C    P0             10W /  180W |       0MiB /  16311MiB |      0%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+
+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI              PID   Type   Process name                        GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|  No running processes found                                                             |
+-----------------------------------------------------------------------------------------+

The strange thing with nvtop, though, is that when I run it and leave it running, the GPU utilization goes back to the regular 5W idling. Once I exit nvtop or the terminal, it pops back to 22W.

I've tested this through multiple reboots, disabling/re-enabling docker, shutting off my Home Assistant VM, and nothing moves the dial. Even with the array stopped its still sitting at that state.

Diagnostics below, I did a quick scan for any issues and couldn't really find anything:

https://drive.proton.me/urls/WRRR0D83H4#ofF0g6IG4OBx

3 hours ago, dumbestsmarthouse said:

Not sure if this is driver related, but I've been having some strange issues with my 5060 Ti. I was on 7.1.4 and it was working fine, but the issue popped up about a week or two ago. I just updated to 7.2.0 and with it the Nvidia driver (was on v575.57.08, now on v580.95.05) to see if that would fix it, but no dice.

This was always the behavior, the GPU, at least after booting Unraid will always stay at P0 and will only go int P8 when something was using the GPU before is now idle again.

However you can circumvent that by putting that line into your go file:

nvidia-persistenced

Hey !
So I'm trying to work with Nvidia GPUDirect RDMA on unraid 7.2.0 but when looking at the kernel module loaded I see that I have:

root@Tower:~# lsmod | grep "nvidia"
nvidia_uvm           1847296  4
nvidia_drm            131072  0
nvidia_modeset       1638400  2 nvidia_drm
nvidia              104067072  29 nvidia_uvm,nvidia_modeset
drm_ttm_helper         16384  1 nvidia_drm
video                  69632  1 nvidia_modeset
drm_kms_helper        217088  2 drm_ttm_helper,nvidia_drm
drm                   655360  6 drm_kms_helper,nvidia,drm_ttm_helper,nvidia_drm,ttm
backlight              20480  3 video,drm,nvidia_modeset
i2c_core              106496  7 drm_kms_helper,i2c_algo_bit,igb,nvidia,i2c_smbus,i2c_piix4,drm

But when looking in the folder: /lib/modules/6.12.54-Unraid/kernel/drivers/video/

I have nvidia-peermem.ko

The issue is that when I run:

modprobe nvidia-peermem -v

I get the following error:

root@Tower:~# modprobe nvidia-peermem -v
insmod /lib/modules/6.12.54-Unraid/kernel/drivers/video/nvidia-peermem.ko
modprobe: ERROR: could not insert 'nvidia_peermem': Invalid argument

Do you guys have any idea what could be going wrong (I have nothing in dmesg) ?

And if so how to fix it !

3 minutes ago, space192 said:

Do you guys have any idea what could be going wrong (I have nothing in dmesg) ?

What card do you have?

Without Diagnostics I can't say anything.

Just now, ich777 said:

What card do you have?

Without Diagnostics I can't say anything.

I have:

• 1 RTX 3090

• 2 Tesla P40

Oh and completely forgot about the diagnostics sorry

tower-diagnostics-20251119-2312.zip

1 minute ago, space192 said:

I have:

Sorry but you have too much customized things on your server:

Nov 19 17:18:40 Tower emhttpd: /usr/local/emhttp/plugins/user.scripts/backgroundScript.sh "/tmp/user.scripts/tmpScripts/gpu_power_usage/script" >/dev/null 2>&1/usr/local/emhttp/plugins/user.scripts/backgroundScript.sh "/tmp/user.scripts/tmpScripts/patch nvidia/script" >/dev/null 2>&1

BTW, please don't enable both, you have the old way of enabling persistenced and also the new way of just issuing nvidia-persistenced from what I see from your Diagnostics, the latter is the only one that's needed usually.

However I don't think that you can use peermem on your GPUs since you have a Consumer and a Datacenter card installed, peermem works only on Datacenter cards AFAIK and also I'm a bit surprised that it's in cluded in the default non Datacenter driver package.

As said, I don't think you even can enable peermem when you have a Consumer card installed, please see here:

https://docs.nvidia.com/cuda/gpudirect-rdma/

Hi, I just updated my unraid to 7.2.0, but now my nvidia driver stopped working. it automatically replaced the existing driver with v580.105.08, instead of the 470.xxx that i need for my gtx 760. I don't see any option to select such legacy driver and i can't find a download.

Edited November 19, 2025Nov 19 by HansieAndersje

6 hours ago, HansieAndersje said:

I don't see any option to select such legacy driver and i can't find a download.

Please see this recommended post on top:

Anyone hit this after upgrading from 7.2.0 to 7.2.1?
Since upgrading from 7.2.0 to 7.2.1 I can't use the --runtime=nvidia extra parameter in docker apps - General Support - Unraid

Edited November 22, 2025Nov 22 by DataBitz
Remove Diag file

5 minutes ago, DataBitz said:

Anyone hit this after upgrading from 7.2.0 to 7.2.1?

Did you read my comment:

1 hour ago, ich777 said:

Did you read my comment:

Thanks, added now.

5 minutes ago, DataBitz said:

Thanks, added now.

I really can't tell why it's not working since it should because nothing changed between 7.2.0 and 7.2.1 in terms of the driver.

Can you please try to switch to another driver, maybe one version before the one you are currently on and reboot afterwards.

I have a similar issue.

After upgrade to 7.2.1, I cannot get the plugin to work. I uninstalled, rebooted and tried to reinstall. I keep seeing "Can't download Nvidia Driver Package v580.105.08". And also "plugin: run failed: '/bin/bash' returned 1".

I have issues with the plugin at every upgrade, but it typically works when reinstalling. But this time somehow does not.

4 minutes ago, steve1977 said:

I have a similar issue.

After upgrade to 7.2.1, I cannot get the plugin to work. I uninstalled, rebooted and tried to reinstall. I keep seeing "Can't download Nvidia Driver Package v580.105.08". And also "plugin: run failed: '/bin/bash' returned 1".

I have issues with the plugin at every upgrade, but it typically works when reinstalling. But this time somehow does not.

Do you have Diagnostics or anything that can maybe help?

Where does it say run bash failed?

What does not work or better speaking what was not working before you uninstalled the plugin, was the card showing up or not, was it working in containers or not?

Let me get a diagnostic log.

I thought the "upgrade issue" was a known bug. It happens at every update and I posted about this. After upgrading Unraid, the plugin is still installed, but the nvidia driver is no longer installed. I will find the thread and post.

What is different this time is that I cannot reinstall the plugin. After removing it, I cannot install the plugin at all. It gives me the errors above while attempting to install the plugin. Let met get the diagnostics.

Just now, steve1977 said:

I thought the "upgrade issue" was a known bug. It happens at every update and I posted about this. After upgrading Unraid, the plugin is still installed, but the nvidia driver is no longer installed. I will find the thread and post.

No, this is not a known bug.

Do you notice the notifications about the drivers when upgrading to a new Unraid version, if yes, what do the notifications say?

1 minute ago, steve1977 said:

What is different this time is that I cannot reinstall the plugin. After removing it, I cannot install the plugin at all. It gives me the errors above while attempting to install the plugin.

How are you installing the plugin, through CA or directly via a GitHub RAW link, a screenshot would also help.

Troubleshooting without nothing more than "is not working" and "bash return 1" is really hard. :D

However the notifications that you see should look something like that (you have to watch it to the end):

https://github.com/ich777/unraid-plugin_update_helper?tab=readme-ov-file#update-notifications

1 hour ago, ich777 said:

I really can't tell why it's not working since it should because nothing changed between 7.2.0 and 7.2.1 in terms of the driver.

Can you please try to switch to another driver, maybe one version before the one you are currently on and reboot afterwards.

Yes, switching from v580.105.08 to v580.95.05 fixed it.

Just now, DataBitz said:

Yes, switching from v580.105.08 to v580.95.05 fixed it.

Can you please try to upgrade to 580.105.08 again and reboot, since it should also have the working container tools.

12 minutes ago, ich777 said:

Can you please try to upgrade to 580.105.08 again and reboot, since it should also have the working container tools.

Yes, also working on 580.105.08 now.
Thank you.
FYI, I had upgraded to 580.105.08 a couple of weeks ago, while on 7.2.0.

Just now, DataBitz said:

Yes, also working on 580.105.08 now.
Thank you.
FYI, I had upgraded to 580.105.08 a couple of weeks ago, while on 7.2.0.

The driver is the same for 7.2.0 and 7.2.1 so I really don't know what's going on here, maybe the package went bad.