bullseye with nvidia driver does not find all cards

[mannebk]

Hi folks,

I changed my OS from Ubuntu to debian Bullseye, installed the the nvidea driver package

NVIDIA-SMI reports 470.161.03 Driver Version: 470.161.03 CUDA Version: 11.4

but the driver only finds the first 4 PCIE cards, 5-9 wont be found.

I got 9 units of P106-100 cards in that rig.

The first 4 just work fine and flawlessly.

I installed from current netinst AMD iso via USB, only ssh and systools, nothing else, did add contrib and non-free as source and installed the nvidia-detect, did run it, found all 9 cards, then installed as recommended by detect the nvidia-driver-packet.

output from detect:

Code: Alles auswählen

Detected NVIDIA GPUs:
01:00.0 3D controller [0302]: NVIDIA Corporation GP106 [P106-100] [10de:1c07] (rev a1)
02:00.0 3D controller [0302]: NVIDIA Corporation GP106 [P106-100] [10de:1c07] (rev a1)
03:00.0 3D controller [0302]: NVIDIA Corporation GP106 [P106-100] [10de:1c07] (rev a1)
05:00.0 3D controller [0302]: NVIDIA Corporation GP106 [P106-100] [10de:1c07] (rev a1)
08:00.0 3D controller [0302]: NVIDIA Corporation GP106 [P106-100] [10de:1c07] (rev a1)
09:00.0 3D controller [0302]: NVIDIA Corporation GP106 [P106-100] [10de:1c07] (rev a1)
0a:00.0 3D controller [0302]: NVIDIA Corporation GP106 [P106-100] [10de:1c07] (rev a1)
0b:00.0 3D controller [0302]: NVIDIA Corporation GP106 [P106-100] [10de:1c07] (rev a1)
0c:00.0 3D controller [0302]: NVIDIA Corporation GP106 [P106-100] [10de:1c07] (rev a1)

edit: and this is smi output

Code: Alles auswählen

manne@P106-100:~$ sudo nvidia-smi -B
No excluded devices found.
manne@P106-100:~$ sudo nvidia-smi -L
GPU 0: NVIDIA P106-100 (UUID: GPU-931ed796-b3a9-3167-7ccc-27f18254784e)
GPU 1: NVIDIA P106-100 (UUID: GPU-351e9d32-155a-79f6-42a8-317be6d8c3a6)
GPU 2: NVIDIA P106-100 (UUID: GPU-0c498c0e-ed5c-d671-eba7-7da3a9d63135)
GPU 3: NVIDIA P106-100 (UUID: GPU-b1ced8e9-352d-6d20-f6c2-960b778bc23f)
manne@P106-100:~$ 

In Ubuntu I had no issue, all 9 working, but I just have debian everywhere, so I thought I change that rig to debian, but basically my real motivation was, I wanted to get rid from one of the monitoring OS that wont give me root on my own hardware and charge me 2 bucks (about 8-20% of the return of that rig each month) to just give me a web UI to look at the results of an API that hands out a simple jason I absolutely can monitor myself just as well for free and even get very specific stats for stuff I want to know...

Btw. I dont recall the driver used in Ubuntu, i never checked, but I have tried going back, and it works with Ubuntu, so im sure its not my BIOS.

Does anyone have a suggestion how to go ahead in finding the culprit?

Or does this issue ring any bells?

Thanks
Manne