A while ago I saw @wendell’s video where he mentioned that he probably ruined his 3090FE because it was running at higher temperatures than it should have. Which is why, he is now using his MSI 3090 Suprim. Does someone know which temperatures the 3090FE’s memory was at?
As a result I had a look at the memory temperatures of my Asus 3090 Strix and saw the following when playing Guardian’s of the Galaxy:
So, as you can see the memory hit 90°C, but I often checked the OSD while gaming and it never was above 88°C, so I think a maximum value of 88°C is also fair to assume. Currently, I have capped the FPS at 60, so I haven’t checked how far the memory temperature goes when it is allowed to run as fast as it can.
When I had a look at the memory temperatures while gaming, they were usually between 84-88°C, unless I did things in the inventory screen or things like that, then it dropped. I could try gaming until the GPU is hot enough and then uncap the FPS and see how far it goes to draw any kind of conclusion. Likewise, I could try other games like Cyberpunk 2077 and see how it handles that.
Should I try to decrease the temperatures further or is this fine? At the moment, there is no precise knowledge of safe temperatures for GDDR6X, only that Ampere thermal throttles at 110°C. I have already undervolted the GPU so there is no leeway there. I could try to increase the fan speed of both GPU and/or case fans. However, the speed of both set of fans would just be increased, since the former checks the GPU’s temperature and the latter are bound to the CPU or the mainboard.
I think my fan and GPU placement is good to offer the backplate as much cooling as possible,
so it’s odd that I have still those high temperatures.
So, I uncapped the FPS and had a look at the temperatures while constantly being power limited and it seems they don’t go above 90°C, which is something at least.
With DLSS on Quality - I forgot to activate DLSS previously - and FPS capped to 60 I get temperatures between 78-80°C, which seems much more reasonable. I also need to check Cyberpunk 2077 for memory temperatures, but I do feel safer already
Anyway, I assume it is best to not touch that power slider, is it?
So, regard this as the final post here since apparently nobody is really interested in this kind of subject. That being said, if someones Googles this issue and stumbles over this thread my experiences might provide some kind of help.
I bought a couple of smaller heatsink as well as a double-sided thermally conductive tape and put them onto the backplate of my GPU like so:
I didn’t attach any more of those heatsinks - although I have eight more pieces - because I didn’t want to get too low on the left side of the GPU (the heatsinks are lower due to RGB stuff^^).
Anyway, I tried both Cyberpunk 2077 and Guardians of the Galaxy without any FPS capping and the highest temperature measured was 88°C, with temperatures usually between 84°C and 86°C (GotG) and 82°C - 84°C (Cyberpunk 2077). When I decided to limit GotG to 60 FPS the highest temperature measured was 78°C.
Hence, I was able to reduce my temperatures by 2°C - 4°C, which is not too much to be honest. I might still try to increase my fan speeds a bit because I noticed the GPU itself being 2°C - 4°C lower, which might have resulted in less fan speed of the GPU.
I have also discovered someone else doing something similar as I have done with his liquid cooled EVGA 3090 here:
