Two video cards on one motherboard. When one heats up, the fan of the other rotates. Geforce Nvidia RTX 2070 and RTX A4000. Ubuntu 22.04

The problem started about six months ago on Ubuntu 20.04 after a driver update. Before that, there were no issues. The 2070 and A4000 graphics cards were cooled independently. Then, when the A4000 was under load, the 2070’s fan would spin up. After installing Ubuntu 22.04, the problem was temporarily resolved by disabling the 2070 and then re-enabling it. This resulted in each card being cooled separately.

However, with a recent update, only the 2070 is being cooled, and when the A4000 heats up, it’s still the 2070 that’s being cooled. At the time of the driver update on Ubuntu, an xorg.conf file was created that resulted in no image being displayed on the monitor. After installing the NVIDIA drivers using the .run file and replacing the xorg.conf file, the display works. But the A4000 still causes the 2070’s fan to spin up when it gets hot.

Furthermore, when trying to control the fan speed in nvidia-settings, setting the percentage for the 2070 changes the value shown for the A4000, and vice versa. However, this doesn’t actually affect the fan speeds. The same behavior occurs when running fan speed control commands from the console. Each card works normally when the other is disabled.
After swapping the positions of the two graphics cards, the A4000 is still being cooled by the 2070’s fan. Previously, the monitor was connected to the A4000, but now it’s connected to the 2070. Reversing the physical installation of the two graphics cards has not resolved the issue; the A4000’s cooling continues to be achieved by activating the 2070’s fan. The display monitor, formerly connected to the A4000, is now connected to the 2070.

±----------------------------------------------------------------------------------------+
| NVIDIA-SMI 550.144.03 Driver Version: 550.144.03 CUDA Version: 12.4 |
|-----------------------------------------±-----------------------±---------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 NVIDIA RTX A4000 Off | 00000000:25:00.0 Off | Off |
| 41% 37C P8 6W / 140W | 17MiB / 16376MiB | 0% Default |
| | | N/A |
±----------------------------------------±-----------------------±---------------------+
| 1 NVIDIA GeForce RTX 2070 Off | 00000000:26:00.0 On | N/A |
| 46% 33C P5 13W / 175W | 925MiB / 8192MiB | 2% Default |
| | | N/A |
±----------------------------------------±-----------------------±---------------------+

The command watch -n 1 nvidia-smi shows an increase in the target fan speed, while the graphical utility displays an increase in the fan speed of another GPU, and the target fan speed value remains the same. The heating of the A4000 still leads to the fan speed increase of the 2070.

After swapping the PCI slots for the graphics cards, the A4000 started to cool down, while the 2070 is not cooling at all; its fan is not being controlled. On the other hand, the A4000 is heating up much more. Additionally, under load on the 2070, the A4000 gradually starts to heat up as well.