Inconsistent fan rotation on some machines

Graphics / Linux Linux
1

Hi all,

I have some identical machines, same hardware and all running Debian 9.1 with the nvidia-driver (375.66) from the repository.

  • They all run a GTX 970 card
  • Same xorg.conf
  • Same X11 folder (tested with rsync -ac)

On some of them, the GPU fan is spinning at low speed continuously, it’s exactly what I want.
On the others, the fan is OFF and is spinning fast a few seconds every 20 seconds which is a bit annoying.

When I stops the graphical manager on those machines, the fan starts to spin normally and do the yoyo again when I reload lightdm, even on the logon screen.

I tried to activate coolbits = 4 in xorg, I can then change the fan speed from nvidia-settings but when I set a low level, it keeps doing the yoyo with more or less time between each peak.

So, if you have any idea of what’s wrong :-)
Thanks in advance !

2

Different brands? Different VBios versions?

3

Same cards, same motherboard, same CPU, RAM and all.
nvidia-driver 375.66-2~deb9u1 amd64 (from debian repository)
Debian 9

machine1# cat /proc/driver/nvidia/gpus/0000:03:00.0/information
Model: GeForce GTX 970
IRQ: 73
GPU UUID: GPU-e9dc6cfd-fcf5-365d-5c98-9137df27168b
Video BIOS: 84.04.36.00.68
Bus Type: PCIe
DMA Size: 40 bits
DMA Mask: 0xffffffffff
Bus Location: 0000:03:00.0
Device Minor: 0

machine2# cat /proc/driver/nvidia/gpus/0000:03:00.0/information
Model: GeForce GTX 970
IRQ: 73
GPU UUID: GPU-1f7da3d6-2605-a4e5-985e-f20de68508cc
Video BIOS: 84.04.36.00.68
Bus Type: PCIe
DMA Size: 40 bits
DMA Mask: 0xffffffffff
Bus Location: 0000:03:00.0
Device Minor: 0

A few months ago I tried with an updated driver from nvidia website but same issue.

4

Your vendor could have had different revisions of your card which explains different fan profiles.

By default NVIDIA drivers do not intervene with the fan management built into your cards. So again it all boils down to your hardware.

5

I can remember such an incident, the reason wasn’t due to different revisions but bad thermal compound used on a batch.

6

but as soon as I switch to console mode without X, the fan stops it’s yoyo.