After a gaming session today I was reviewing my temp logs and noticed ICX reporting mem3 at around 90 degrees. My card is waterblocked and all other ICX temps read mid 50s or lower. This leads me to believe there is bad thermal pad contact on the memory closest to the pcie connector, or to the lower right side of the gpu core.

Is that the correct position of mem3? Also with it being a 3090, how likely is it that the bad contact could actually be on the backside of the card with the memeory modules and the backplate? Can ICX pick up a problem in that region through the pcb? Or is it more likely to be on the front side of the card?

I assume I'm looking for pads with little to no compression/indentation from the memory modules, but in the event that the contact looks good, what else could cause this on that one sensor?

I plan to take it apart later this week when I need to open up the loop for a replacement flow sensor anyways, just wondering how much of a tear down I should expect on the gpu. Backplate is quick, taking the whole block off is not hard, but just much longer of a process.

Buggy icx readings?

Odd. The temps you posted shouldn’t have caused throttling.

Right. I initially thought the driver crashed when fps tanked, but when I checked the event logs there was nothing. Thats when I checked the temps in HWiNFO64 and saw the max 90ishC peak on the iCX sensors and "Performance limit - Thermal" listed as "yes". I also saw that it showed yes for throttle because of "SLI GPUBoost Sync"? Which is odd cause its a single card. The gpu core max temp was 42c. Not sure how that is even close to throttle.

That is my understanding as well. The thing is only one sensor spiked each time and neither should have triggered anything, which is why I am confused things are happening.

Yeah I'm gonna reduce clocks and test more, I'm just still confused as to why thermals are being reported the way they are unless erroneous thermal readings are a result of stability issues. Didn't know that was a thing.