- Apr 7, 2014
- 1,101
- 103
- 20,090
Is it me, or does the nVidia fans insist on trying to run a 30% regardless if I try to brute force them using either nVidia App, MSI Center or FanControl v226, or and MSI AfterBurner.
Notably, MSI AfterBurner if I move the slider to 50% it insists on reverting to 30%.
FanControl v226, in my attempt to create a fan curve of about 50% fans at 50c+ based on GPU temp, it seems to be carrying out a fight between 30% and 50%.
MSI Center seems to have some AI clever stuff in it but I cant get the GPU fans to set where I want it at 50%.
I have undervolted using MSI Afterburner to 70% to reduce power draw (max 560w at the wall), but doesn't impact performance as AI is all about the 32GB vram and cuda.
I have an MSI RTX 5090 Ventus and during the run, the memory temp maxes at 78c and GPU temp at 70c.
Left to the GPUs own mind, the fans vary between 30% and 34% to maintain a max 78c in my well vented case.
Ok, so, I am running ComfyUI running some AI that hammers the GPU but leaves the CPU idling. This could be running all day so I am trying to minimise stress on the mem temp to max longevity of the card.
Ultimately, I have two (three) questions here...
1. Am I obsessing over temps where 78c mem temp is totally normal, and I should just leave it to do its thing ?
2. Why are nVidia GPU fans hell bent on doing what they want regardless ?
3. Do you have any user experience of successfully setting GPU fans to a stable desired RPM ?
Thanks for reading, hope its not a TLDR.
Notably, MSI AfterBurner if I move the slider to 50% it insists on reverting to 30%.
FanControl v226, in my attempt to create a fan curve of about 50% fans at 50c+ based on GPU temp, it seems to be carrying out a fight between 30% and 50%.
MSI Center seems to have some AI clever stuff in it but I cant get the GPU fans to set where I want it at 50%.
I have undervolted using MSI Afterburner to 70% to reduce power draw (max 560w at the wall), but doesn't impact performance as AI is all about the 32GB vram and cuda.
I have an MSI RTX 5090 Ventus and during the run, the memory temp maxes at 78c and GPU temp at 70c.
Left to the GPUs own mind, the fans vary between 30% and 34% to maintain a max 78c in my well vented case.
Ok, so, I am running ComfyUI running some AI that hammers the GPU but leaves the CPU idling. This could be running all day so I am trying to minimise stress on the mem temp to max longevity of the card.
Ultimately, I have two (three) questions here...
1. Am I obsessing over temps where 78c mem temp is totally normal, and I should just leave it to do its thing ?
2. Why are nVidia GPU fans hell bent on doing what they want regardless ?
3. Do you have any user experience of successfully setting GPU fans to a stable desired RPM ?
Thanks for reading, hope its not a TLDR.
Last edited:
Facebook
Twitter
Instagram