Hi, we did change some of out temp logic and temp sensor readings a while back, might have forgotten to make the nvidia-smi output friendly for endusers to understand…
In essence, we introduced T(limit) as the distance to the temp, a SW slowdown will kick in, so 15° C in your case…
T(limit)=0 would indicate SW slowdown to kick in, 2° hotter, HW slowdown will kick in, and 5° hotter HW shutdown will happen to protect the GPU…
Generally, any current/average temp readings up to high 80s are still fully within certification limits! = no reason to worry.
If you have a simple way to keep the GPU cooler, fine, any slowdown will kick in only later, and overall, lifetime of cooler electronic components increases over hotter components.
That said, Nvidia and their OEM partners guarantee full functionality of the GPU in a certified chassis and a specified environmental temperature for the full warrantied time of the product…
I have requested input from product management to consider making the nvidia-smi readings easier to understand for endusers….
hope this calms your worries..
regards
-Frank