VFIO blacklisted GPU temperatures

Hello. As a preface, I’m a novice when it comes to anything linux / virtualization / hardware related, so please forgive me if I ask any obvious questions.

Recently, I set up Proxmox on my system with a MacOS guest. I have 2 separate GPUs, an RX 580, and a 2080. The RX 580 is passed through to the MacOS guest, and the 2080 is blacklisted. When Proxmox boots up, the fans on the 2080 no longer function, which I assume is from the blacklisted drivers.

The main thing that I’m concerned about is: The card itself still feels very hot to the touch. Is it drawing enough power to cause it to overheat / cause damage? If so, would it be enough to simply remove the 2080 from the blacklist? The card comes with some RGB that I never figured out how to turn off, which is why I’m assuming it’s still drawing power.

On a somewhat unrelated note, if I tried to boot up my existing Windows install as a VM, and tried to pass the 2080 to it, would I need to remove the 2080 from Proxmox’s blacklist ?

when VFIO is holding a card, it puts it in the lowest power state. This causes very low power draw. It’ll obviously vary based on the GPU, but the other thing that happens is the fans do not spin. This means the heatsink can’t efficiently do what it’s intended to do.

So, to answer your question: no, it won’t cause damage. If the card is getting hot enough to cause damage, you’d be reporting that touching the card burned your hand. Remember, GPUs are capable of handling high temperatures. Not quite as high as CPUs, but still quite high. 70C is a typical operating temp for a stock-tuned GPU under a gaming load.

Now, if you’d like to know for sure, you can always just grab a thermal probe and test the card, to see.

Oh. I see. Thanks for the answer!

I have a few follow up questions if you don’t mind:

Is it possible to use something like lm_sensors / psensors to grab the temperature instead of a thermal probe? I’m not too sure how the aforementioned programs actually work, but if so, would they also be able to display the temperature data on the rest of the system while the guest is up and running?