Graphics Card not getting Detected

[prince262150]

Hello Everyone!

We are using Nvidia Graphics Card, Operating in Gen1 configuration on RHEL8.2, During Operating, Booting, Restart or Running High Processing Application, Graphics card communication with processor is getting lost and the system display gets Hanged while the processor is still running , so we access the system using SSH in other PC there we are seeing that in 'LSPCI' we are not able to see the detection of Graphics Card also when we are giving command 'nvidia-smi' it says unable to access device handle for GPU.
FYI
Attached Log Files.

 

[MikeWalsh]

From the sound of things, you're an enterprise organisation, yes? That being the case, I have to assume you have a support contract with RedHat.

Why don't you raise a support ticket via the appropriate channels and get help direct from them? Nobody's in a better position to understand the issue than RedHat themselves.....

(shrug...)


Mike.

 

[f33dm3bits]

Which gpu do you have installed in the system and which exact drivers did you install and how did you install them? And what's the output of the follow when you run it from the command-line?

lspci | qgrep -i vga

 

[prince262150]

Thanks for replying, we have contacted RHEL but they said they have limited resource for 3rd party driver support, hence i have come to this open forum.

 

[prince262150]

GPU - NVIDIA Quadro RTX 3000
Drivers - https://www.nvidia.com/Download/driverResults.aspx/226768/en-us/

Linux X64 (AMD64/EM64T) Display Driver​

Version:	550.90.07
Release Date:	2024.6.4
Operating System:	Linux 64-bit
Language:	English (US)
File Size:	293.33 MB

Installation - Through Run file provided by NVIDIA

The output of "lspci | grep -i vga" this command is

01:00.0 VGA compatible controller : NVIDIA corporation device 1f76 (rev a1)

 

[Brickwizard]

Have you tried Nvidia support at https://www.nvidia.com/

 

[f33dm3bits]

Installation - Through Run file provided by NVIDIA

That's not the recommended way to install them, just install them through the instructions provided by Nvidia.

1. Introduction — Installation Guide for Linux 12.5 documentation

The installation instructions for the CUDA Toolkit on Linux.

docs.nvidia.com

Basically all you have to do is the following.

1. Uninstall the driver you installed first
./NVIDIA-Linux-x86_64-550.90.07.run --uninstall

2. Install the repo and accept the gpg pubkey.
sudo dnf config-manager --add-repo https://developer.download.nvidia.com/compute/cuda/repos/rhel8/x86_64/cuda-rhel8.repo

3. Clear yum cache
sudo dnf clean expire-cache

4. Install the Nvidia driver.
sudo dnf module install nvidia-driver:latest-dkms

5. Install the cuda SDK kit if you need it (Optional)
sudo dnf install cuda-toolkit

6. Reboot

I would recommend you update the system from RHEL 8.2 to the most recent version, 8.10 because Nvidia drivers from that repo tend to only be supported with the most recent kernel version. I use that repo for a system running at home using an Nvidia Quadro P2000 gpu.

 

[MikeWalsh]

@f33dm3bits :-

In other words, use the accepted, documented, established method. Fair 'nuff.....

Actually, compiling via the .run file is the way we've always done this stuff under Puppy.....and that's because when one of our Japanese members wrote a Puppy-native utility for installing the Nvidia drivers many years ago, this was the route he chose.

Since it works so well, nobody's ever tried changing it. There simply hasn't been any need.

But then of course - as you're no doubt aware - we do most things differently to the 'mainstream' in Puppy. We don't use the 'standard' package managers, so the 'regular' instructions are right out the window for us. And even in the newest Puppies - which are starting to use apps like Synaptic now - the 'official', repo-maintained driver packages simply don't work for us......yet installing via the .run file always has.

(shrug...)


Mike.