Wrong GPU Hot Spot Temperature since 7.22

Hi HWiNFO64 has always been reporting correct GPU Hot Spot Temperature until I upgraded to 7.22.

My GPU is 3090, I can tell its correct Hot Spot Temperature for almost a year now, since I started ETH mining with the GPU when the PC is not being used. The sensor is a key indicator, which normally reads between 104 to 110, and 108 is a sweet spot. If it's above 110, it means the GPU is heat throttled, which is not ideal, and 110 is the maximum reading, in which case I need to cool down the environment of my PC by adding fans to blow out heat from around the PC.

But since upgraded to 7.22, it shows only 64 degree, this is not right, this is GPU core temperature not the hot spot one. Another app iCUE (Corsair dashboard) still shows the correct GPU hot spot temperature:

1650623254037.png


Please verify the bug. Thanks!
 

Martin

HWiNFO Author
Staff member
Are you sure that this is due HWiNFO version 7.22? I'm surprised by this as there were no changes in this area recently.
Can you please confirm by running an older version of HWiNFO at the same time, what it will show?
 
This is indeed weird. I just tried 7.20 and 7.17, they both display 64 degree which should be 106~108 degree as you can see here on iCUE! and the last time I updated GPU firmware was 1 month ago - so nothing changed on GPU side!

That's really strange.

1650630192874.png


I have a 3080 GPU card too, i'll test on that box shortly!
 
Just check 3080 PC, it shows correct GPU Hot Spot Temperature of 104 degree, using the latest 7.23 beta!

I cold rebooted my 3090 PC and updated to the latest 7.23 beta too, still showing wrong value...

So no idea what's going on, my 3090 all workings fine... only hope another 3090 user can give it a go and report back if they are OK with the readings... for now i couldn't use HWiNFO64 to monitor my 3090 heat throttling...
 

Martin

HWiNFO Author
Staff member
Please attach the HWiNFO Debug File of the 3090 when running at the high temperature. I need to check some more details...
Are you sure iCUE is showing the correct Host Spot temperature and it's not the GPU Memory Junction Temperature? I can't see this value in HWiNFO, has you perhaps hidden it?
 

Martin

HWiNFO Author
Staff member
You might also try to completely "Reset Preferences" in HWiNFO if that will change something.
 

Martin

HWiNFO Author
Staff member
Interesting.. I think it had something to do with the hidden value which caused some configuration mismatch between the Hot Spot and Memory Junction temperatures.
 
Please attach the HWiNFO Debug File of the 3090 when running at the high temperature. I need to check some more details...
Are you sure iCUE is showing the correct Host Spot temperature and it's not the GPU Memory Junction Temperature? I can't see this value in HWiNFO, has you perhaps hidden it?
iCUE "GPU hotspot temperature" is indeed is my own tag for the sensor, not from iCUE itself (which only list sensors with # numeric counters), but i'm sure this value is true because 110 degree is the community recognised heat throttle threshhold top value and I have been tracking among different GPUs for more than 1 year now. When mining with GPU, we need to keep this value under 110. However I couldn't verify actual temperature using my infrared sensor gun, it has to be disassembled to measure I think (maybe even that it's not quite possible to measure)
 

Martin

HWiNFO Author
Staff member
IR guns won't provide reliable information in this case as junction temperatures are measured inside silicon. That can't be verified using tools measuring surface/case temperature.
 
Interesting.. I think it had something to do with the hidden value which caused some configuration mismatch between the Hot Spot and Memory Junction temperatures.
Ah, it's actually misleading :) it should read GPU Memory Junctions Temperature, somehow during my last upgrade, the sensor became invisible and only hotspot temperature being shown, hence I picked up my old iCUE dashboard with a wrong tag of my own (hotspot...)...

Now it's back to normal:

1650636512834.png

Thanks again for the tool, extremely useful for us to monitor our system!
 
Top