Hi,
GPU: ZOTAC RTX 3080 LHR
When gaming, my hot spot temp reaches 103.3 °C (While the GPU Core temp is ≈75 °C and memory junction is about ≈70 °C) I have repasted my GPU (before, hot spot temp was 104.3 °C) but this did not bring a noticeable improvement (to the Hot Spot temps, it decrased core temps by ≈10 °C). This makes my fans spin up quite often and quite loudly (I may also have an issue with fan whine). I created a custom fan curve which makes the fans louder over-all but eliminates fan spikes. I have added a support strut to the GPU in case the plate is getting separated from the radiator because of possible sag, which also did not bring any improvement.
While idle, the core and junction temps are roughly the same (hot spot is ≈10 °C hotter than core/junction)
I also have conflicting information on the desired hot spot temperatures. On one hand I've heard that it is fine to be up to 110 °C, and on the other sources have pointed to a temperature that should be lower.
TL;DR:
Edit:
More research indicates possible problems, but further inquiry is required:
GPU: ZOTAC RTX 3080 LHR
When gaming, my hot spot temp reaches 103.3 °C (While the GPU Core temp is ≈75 °C and memory junction is about ≈70 °C) I have repasted my GPU (before, hot spot temp was 104.3 °C) but this did not bring a noticeable improvement (to the Hot Spot temps, it decrased core temps by ≈10 °C). This makes my fans spin up quite often and quite loudly (I may also have an issue with fan whine). I created a custom fan curve which makes the fans louder over-all but eliminates fan spikes. I have added a support strut to the GPU in case the plate is getting separated from the radiator because of possible sag, which also did not bring any improvement.
While idle, the core and junction temps are roughly the same (hot spot is ≈10 °C hotter than core/junction)
I also have conflicting information on the desired hot spot temperatures. On one hand I've heard that it is fine to be up to 110 °C, and on the other sources have pointed to a temperature that should be lower.
TL;DR:
- Idle Core 40 °C, Mem junction 40 °C, hot spot 50 °C (C-Hs Δ ≈ 10 °C)
- Load Core 75 °C, Mem junction 70 °C hot spot 90-100 °C (C-Hs Δ ≈ 30 °C)
- Repasted, effect: Core ↓ ≈15°C, Hot Spot ↓ ≈1°C (all above results are after repaste)
- Gpu is supported by a strut to prevent sagging
- The side panel is open and it was quite chilly in the room when test was performed (Ambient ≈ 17 °C)
- I have not exchanged the memory thermal pads, and due to their cost I am not willing to, unless absolutely sure that they are the problem
Edit:
More research indicates possible problems, but further inquiry is required:
- Reassembly might've screwed with thermal pads (dust might've gotten on them, making them uneven?)
- Screws might be tightened unevenly (They have springs, I'm not sure how that affects tightening)
- Thermal paste used might not be good for bare die application (It did bring down the core temp tho), I used thermal grizzly kryonaut.
- Cannot determine if the hot spot is in the memory or the die, that info might shine more light on the problem, but I am unsure how to check that
Last edited: