[Plugin] Nvidia-Driver

[ich777]

33 minutes ago, Thundermonk said:

Not sure if you are looking at a different motherboard bios?

The above screenshot is your USB boot device... You have to click on the Flash drive on the Main page as mentioned above in the Unraid WebUI and follow the steps from above to allow UEFI boot.

 

When the folder is named EFI- it is not possible for your BIOS to boot into UEFI mode AFAIK, this is what the Permit UEFI boot actually does in the Unraid WebUI, it changes the name from the EFI- folder to EFI

 

33 minutes ago, Thundermonk said:

 

Maybe this is why it's working, because you are allowing Legacy too and it eventually falls back to Legacy.

[Thundermonk]

33 minutes ago, ich777 said:

The above screenshot is your USB boot device... You have to click on the Flash drive on the Main page as mentioned above in the Unraid WebUI and follow the steps from above to allow UEFI boot.

 

When the folder is named EFI- it is not possible for your BIOS to boot into UEFI mode AFAIK, this is what the Permit UEFI boot actually does in the Unraid WebUI, it changes the name from the EFI- folder to EFI

 

Maybe this is why it's working, because you are allowing Legacy too and it eventually falls back to Legacy.

 

Apologies, totally misunderstood what you meant. I have changed it to permit UEFI boot mode and the folder is now just EFI so looks like that worked. Fingers crossed this works!

 

Once again, thank you for taking the time to help.

[Richard de Weerd]

is there any way to install a newer driver? I see that the plugin installed 520.56.06 for GeForce GTX 1050 Ti but apperently frigate needs 530 or highter

[ich777]

3 minutes ago, Richard de Weerd said:

is there any way to install a newer driver? I see that the plugin installed 520.56.06 for GeForce GTX 1050 Ti but apperently frigate needs 530 or highter

On what Unraid version are you? Diagnostics?

[Richard de Weerd]

5 minutes ago, ich777 said:

On what Unraid version are you? Diagnostics?

6.11.1

[ich777]

Just now, Richard de Weerd said:

6.11.1

Do you have Diagnostics?

You are a bit behind... Unraid 6.12.11 is the latest version.

[Richard de Weerd]

18 hours ago, ich777 said:

Do you have Diagnostics?

You are a bit behind... Unraid 6.12.11 is the latest version.

Updating helped, thanks for pointing me, I should have found that myself!

[mans_]

@ich777 Hi! I would like to know if you are using https://github.com/NVIDIA/nvidia-container-toolkit for this plugin? If yes, is this plugin up to date with the nvidia releases in the link shared? If yes, why is the nvidia container toolkit version not mentioned in the driver? 

I am planning to install pytorch and other python packages in a docker and wondering how reliable this plugin is or is it possible to install the driver directly on top unraid host?

Thank you! 

[ich777]

On 7/31/2024 at 3:44 PM, mans_ said:

@ich777 Hi! I would like to know if you are using https://github.com/NVIDIA/nvidia-container-toolkit for this plugin? If yes, is this plugin up to date with the nvidia releases in the link shared? If yes, why is the nvidia container toolkit version not mentioned in the driver? 

Yes otherwise noting would work for Unraid.

 

It's not mentioned in the plugin itself because these specific Unraid driver packages are made specifically with Docker support in mind otherwise it won't make much sense to release just a driver package for Unraid.

 

If you go through the source you can see that I add both libnvidia-container and nvidia-container-toolkit (also in my repositories <- including how they are compiled) :

https://github.com/ich777/unraid-nvidia-driver/blob/master/source/compile.sh#L45-L60

The slack-desc includes that:

nvidia-driver Package contents:

Nvidia-Driver v560.28.03
libnvidia-container v1.15.0
nvidia-container-toolkit v1.16.1



Custom nvidia-driver for Unraid Kernel v6.1.99 by ich777

but that is not visible to the user because the installation is done in the background.

 

Hope that helps.

 

  

EDIT: Sorry I completely missed that part:

On 7/31/2024 at 3:44 PM, mans_ said:

I am planning to install pytorch and other python packages in a docker and wondering how reliable this plugin is or is it possible to install the driver directly on top unraid host?

I don't know what you mean by how stable that plugin is, it is basically the Nvidia driver.

Maybe someone else can answer but I most of the times get requests here because of hardware incompatibilities, because of cards dying or some other issues but not about the driver(s) not working.

 

The latest driver version 560.28.03 has about 4500 downloads on it for Unraid version 6.12.11

So I assume everything is working.

 

Sure you can of course install the driver on top of Unraid but this is a really tedious process because you have to do that every time you reboot, basically this is why this plugin exists, I produce a driver package for every single Unraid version.

I compile the Production/New Feature/Beta and Legacy Branches in terms of driver packages and of course the latest driver version for each stable Unraid version.

 

As long as you stay on the latest version from Unraid you should always be or better speaking have always the option to upgrade to the latest Nvidia driver version.

[mans_]

2 hours ago, ich777 said:

Yes otherwise noting would work for Unraid.

 

It's not mentioned in the plugin itself because these specific Unraid driver packages are made specificity with Docker support in mind otherwise it won't make much sense to release just a driver package.

 

If you go through the source you can see that I add both libnvidia-container and nvidia-container-toolkit (also in my repositories <- including how they are compiled) :

https://github.com/ich777/unraid-nvidia-driver/blob/master/source/compile.sh#L45-L60

The slack-desc includes that:

nvidia-driver Package contents:

Nvidia-Driver v560.28.03
libnvidia-container v1.15.0
nvidia-container-toolkit v1.16.1



Custom nvidia-driver for Unraid Kernel v6.1.99 by ich777

but that is not visible to the user because the installation is done in the background.

 

Hope that helps.

 

  

EDIT: Sorry I completely missed that part:

I don't know what you mean by how stable that plugin is, it is basically the Nvidia driver.

Maybe someone else can answer but I most of the times get requests here because of hardware incompatibilities, because of cards dying or some other issues but not about the driver(s) not working.

 

The latest driver version 560.28.03 has about 4500 downloads on it for Unraid version 6.12.11

So I assume everything is working.

 

Sure you can of course install the driver on top of Unraid but this is a really tedious process because you have to do that every time you reboot, basically this is why this plugin exists, I produce a driver package for every single Unraid version.

I compile the Production/New Feature/Beta and Legacy Branches in terms of driver packages and of course the latest driver version for each stable Unraid version.

 

As long as you stay on the latest version from Unraid you should always be or better speaking have always the option to upgrade to the latest Nvidia driver version.

 

Thank you for your work and reply. I just have a few more questions.

Will a 4090 work? (I see no reason why it won't but you say something about hardware incompatibilities, so just confirming)

If yes, is possible to pass through dual (2x4090) GPU to the SAME container with this plugin driver?

Edited July 31 by mans_

[ich777]

1 hour ago, mans_ said:

Will a 4090 work?

Yes, if the driver is new enough to support it, but the answer is basically yes, drivers should support a 4090

 

1 hour ago, mans_ said:

If yes, is possible to pass through dual (2x4090) GPU to the SAME container with this plugin driver?

Sure why not...? But keep in mind that depends on the container but I think most AI containers are capable of using multiple GPUs.

[ConnerVT]

On 7/31/2024 at 10:08 AM, ich777 said:

Hope that helps.

I love reading how the sausage is made.  Thanks for sharing, and for your outstanding work as well as your continued dedication.

 

(Sometime a guy needs to hear that what they do is appreciated...)

[shabash]

I have an Nvidia P1000 due to height limitations within my case (CS381)

 

Up until v545.29.06 my usual test of transcoding a source 4k x265 to an x264 1080p 10Mbps has worked wonders and it barely bothers the GPU

 

If I upgrade however to Unraid 7 beta 2, this Nvidia version is stripped from the plugin list and gives me a newer version automatically. But this causes Plex crashing

 

What are my options? Stay on Unraid 6? Buy a newer 1/2 height GPU? Can we keep v545.29.06 an option?

Edited August 11 by shabash

[ich777]

1 hour ago, shabash said:

If I upgrade however to Unraid 7 beta 2, this Nvidia version is stripped from the plugin list and gives me a newer version automatically. But this causes Plex crashing

Do you have Diagnostics? Troubleshooting is really hard without them.

The version is not stripped, it is most likely not available because you where on a really old version maybe...?

What version did you use before? What version do you use now?

 

A bit more information would be also nice, does it instantly crash? What kind of source material do you try to transcode AVC, HEVC, HDR,...?

 

1 hour ago, shabash said:

Can we keep v545.29.06 an option?

No, why?

As you can see here your GPU is still supported and my GPU (Nvidia T400) is still working just fine with Jellyfin and Unraid 7.0

 

What container are you using? From which maintainer?

 

I just checked and I see that there are about 750 downloads on the latest Nvidia Driver so I assume the driver is working for other users too.

[nielu]

Hi guys. First my apologies if this was asked before. I just could not find this specific problem and is asking for advice and help  please. I am running an unraid server (v6.12.11) on an ASUS ROG STRIX B560-A GAMING motherboard and Intel i7-11700K. It runs fine with Plex and a couple of other docker applications. I also had a Gigabyte GeForce RTX 3060 GPU to transcode plex media, etc. I took this out as the Intel processor's internal GPU works just as fine and besides, I just now build a new server on an AMD processor (ASRock X570 Phantom Gaming 4) with an AMD Ryzen 9 5900X processor. The Ryzen does not have an onboard GPU, so I had to plug in the GTX 3060.

The new server runs fine and will be replacing the older machine, why copyig all the data (Some 65Tb's worth). IT is still busy copying, etc...

The problem is, the new server absolutely refuses to install the NVIDIA driver from the APPs. I installed the GPU Stats plugin, even the Intel GPU Top, but then removed it again as that was just to get the NVTop plugin to install, etc... After all of this, the driver refuses to install and gives me the following message:

 

Any advice or help will be much appreciated. Many thanks.

[ich777]

6 minutes ago, nielu said:

Any advice or help will be much appreciated. Many thanks.

Please post a screenshot from the full plugin page.

 

Are you actually sure that your server has full and unrestricted access to the Internet?

The plugin uses the GitHub API which is known to cause issues with Unifi hardware and AdBlockers.

 

Please make also sure that you have at least 300MB of free space on your USB Boot device for the driver.

[nielu]

Aha, a light went just up. My USB stick is only 1Gb in size and it complained that the stick was more than 90% full. I did not realise the driver was over 300mb. This is probably why it doesn't download. I have managed to install other plugins over the net with no issues, so I think this must be the problem. Let me change the USB to a larger one and try again. Thank you so much!

[crocker5731]

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

 

how do I fix it?

tower-diagnostics-20240816-0129.zip

Edited August 16 by crocker5731

[alturismo]

4 minutes ago, crocker5731 said:

how do I fix it?

may start with posting your diagnostics <<click<<

[crocker5731]

22 minutes ago, alturismo said:

may start with posting your diagnostics <<click<<

I got add the diagnostics file!!!

[ich777]

2 minutes ago, crocker5731 said:

I got add the diagnostics file!!!

I don't see a Nvidia GPU in your system, at least not from the output from lspci.

 

In addition to that, may I ask why you have the AMD-Vendor-Reset plugin installed, I don't see a AMD GPU in your system either...

[crocker5731]

7 minutes ago, ich777 said:

I don't see a Nvidia GPU in your system, at least not from the output from lspci.

 

In addition to that, may I ask why you have the AMD-Vendor-Reset plugin installed, I don't see a AMD GPU in your system either...

I had automatic-ripping-machine working at one time. But now I get win reinstalling it.

docker run
  -d
  --name='automatic-ripping-machine'
  --net='bridge'
  --privileged=true
  -e TZ="America/Los_Angeles"
  -e HOST_OS="Unraid"
  -e HOST_HOSTNAME="Tower"
  -e HOST_CONTAINERNAME="automatic-ripping-machine"
  -e 'NVIDIA_DRIVER_CAPABILITIES'='all'
  -e 'NVIDIA_VISIBLE_DEVICES'='GPU-9152780a-a0e6-3877-b6fd-01ec4b6c22be'
  -e 'ARM_UID'='1001'
  -e 'ARM_GID'='1001'
  -l net.unraid.docker.managed=dockerman
  -l net.unraid.docker.webui='http://[IP]:[PORT:8080]/'
  -l net.unraid.docker.icon='https://raw.githubusercontent.com/automatic-ripping-machine/automatic-ripping-machine/main/favicon.ico'
  -p '8088:8080/tcp'
  -v '/mnt/user/appdata/ARM/':'/home/arm':'rw'
  -v '/mnt/user/appdata/ARM/Music':'/home/arm/Music':'rw'
  -v '/mnt/user/appdata/ARM/config':'/etc/arm/config':'rw'
  -v '/mnt/user/appdata/ARM/logs':'/home/arm/logs':'rw'
  -v '/mnt/user/appdata/ARM/media':'/home/arm/media':'rw'
  --runtime=nvidia
  --gpus all
  --device /dev/sr0
  --device /dev/sg1 '1337server/automatic-ripping-machine:latest'

bcfc6a1d3e1c4ec2c84b57da2e9f8b0503e6b4d1fb62378ff4ac4a27019cbf4d
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 hook #0: error running hook: exit status 1, stdout: , stderr: Auto-detected mode as 'legacy'
nvidia-container-cli: initialization error: nvml error: driver not loaded: unknown.

The command failed.

 

how do I fix it???

 

What I'm trying to do automatic rip dvd to Plex-Media-Server.

Edited August 16 by crocker5731

[ich777]

Just now, crocker5731 said:

I had automatic-ripping-machine working at one time. But now I get win reinstalling it.

Are you sure you read my comment?

It seems you don't have a Nvidia GPU installed in your system.

 

If you want to get the container to work without the GPU you have to remove:

1 minute ago, crocker5731 said:

  --runtime=nvidia

 

[crocker5731]

5 minutes ago, ich777 said:

Are you sure you read my comment?

It seems you don't have a Nvidia GPU installed in your system.

 

If you want to get the container to work without the GPU you have to remove:

 

 

[ich777]

2 minutes ago, crocker5731 said:

 

Sorry but you came here that the GPU is not working and now you post a screenshot from the template.

 

You have to give me more information what you want, do you want to get the template going with the GPU or without the GPU?

Do you have a GPU in your system?