GPU becomes excessively warm, causing performance reduction and further heating.
GPU becomes excessively warm, causing performance reduction and further heating.
Hey everyone, a few days ago I saw the GPU fans increasing noticeably in RPM. The issue is that the hotspot stays around 105 °C during heavy use. After removing the overclocking I was applying, I can't even re-enable it because the afterburner resets the overclock check. Today it got much worse, with the fan running at 3000 RPM or more. About six months ago I replaced the thermal paste and pads (since the VRM temps were really high then) and since then the GPU temps have stayed around 60°C, remaining stable. I don’t know why the hotspot forms or how to get more detailed sensor data—any tips would be great. Thanks in advance, Erik.
From my observations, the hotspot typically rises 15 to 20°C above average, not 40. This might indicate uneven heat distribution, insufficient thermal paste, or poor mounting pressure affecting cooling.
It might actually be true, I’ll look into it right away because it seems like a simple solution. It doesn’t seem so, as I added them about six months ago and they were fairly good quality pads (Thermalright 1.5mm 12.8 W/mK). The same applies to the spread and amount of thermal paste—I’m quite comfortable with my work in this area and used Thermal Grizzly Kryonaut. I’ll verify this after checking the mounting pressure. What surprises me most is how quickly the temperature rose and keeps getting worse, plus the failing OC Scanner in afterburner leaves a poor impression. Edited April 13, 2022 by Lokust
These issues aren't related to the GPU; they're coming from VRAM and other components, not the main card.
there are no error messages displayed. you might have seen them shown twice due to formatting or display issues. the fan is running at 50% as expected for normal operation. the "VRM 59C" reading appears low, but it’s likely within the manufacturer’s specifications. don’t worry, that’s not unusual.
It's true, I didn't see any VRM sensor data for my 2080 or 3080 models, which seems unusual.
Hi there, apologize for the delayed reply; I had been quite scheduled with work. Eventually, I managed to take the time to inspect the cooler and everything inside. Essentially, I discovered the issue and resolved it. A bit longer back, the thermal pads were still in good shape, properly placed and soft but not crumbly. The backplate was quite warm, which made me suspect the die was overheating—turns out it was. Probably over-tightened the screws while reattaching the cooler, causing the thermal compound to ooze out from one corner. I reapplied the pad, reassembled everything, and it worked perfectly. The hotspot dropped by 30°C and the rest remained stable. The GPU fans don’t have a fixed speed limit; as shown in the screenshot’s upper section, they increase from 1400 RPM to 3300 RPM, which seems to be the card’s max. The 50% setting might have been a misreport earlier. It looks like a recent update fixed the fan restriction issue. Regarding the chip/VRM readings appearing twice—no idea why, but possibly because the VRMs are split into two groups with only one sensor, causing duplicate data. Thanks for all your help! Have a great day!