Unable to enable GPU fan control - MSI 4090

Graphics / Linux Linux
1

Hi there, I have a MSI 4090 running on Ubuntu 20.04 that I am trying to set up for AI work. I have Nvidia driver 535 installed (non-open version).

I am unable to control the fans on the GPU so when the GPU hits 90C it ramps up to 100% and stays there even when the temps come back down. When I click on enable fan control in Xserver, it immediately defaults back to GPU fan control disabled

Screenshot from 2024-02-18 20-21-56

I have tried editing Xwrapper.config to have needs_root_rights=yes and allowed_users=anybody. I have tried enabling cool-bits=28 but nothing seems to have worked.

Any ideas on what I should try next?

vastai-2@VastAI-2:~$ nvidia-settings -a ‘GPUTargetFanSpeed=50’

ERROR: Error assigning value 50 to attribute ‘GPUTargetFanSpeed’
(VastAI-2:0[fan:0]) as specified in assignment ‘GPUTargetFanSpeed=50’
(Unknown Error).
ERROR: Error assigning value 50 to attribute ‘GPUTargetFanSpeed’
(VastAI-2:0[fan:1]) as specified in assignment ‘GPUTargetFanSpeed=50’
(Unknown Error).`

2

Here’s what nvidia-smi shows

Screenshot from 2024-02-18 20-35-07
Screenshot from 2024-02-18 20-35-07805×476 53 KB

3

The fan is already in error state, likely due to overheated memory. In that state, it’s locked at 100%.

4

It is in that error state even after a reboot or when the GPU is cold. Are you saying that the memory overheated at some point and this error is permanent?

5

If it’s in ERR! state right after boot, it’s not overheating at all but the fan control is malfunctioning in general. Please try upgrading driver to latest.

6

I’ve tried 525 and 535 drivers now, is there something even more recent?

7

545/550beta

8

550 beta did the trick! Thanks for the tip!!!

9

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.