I have been having heat issues with my GPU. The problems started playing Overwatch where the computer would actually completely restart. A look at the minidump showed that it was the Nvidia driver which was crashing, and watching the temperatures I noticed the GPU Hotspot Temperature was getting very high – 115C. From what I read this is at the level the card may emergency shutdown to protect itself.
Before making any changes I ran HWiNFO and OCCT stress testing the GPU and watching the temperatures so I could have a repeatable test. I then pulled apart the computer and cleaned all the dust out and gave it a general clean. From there I took out the GPU, disassembled it and replaced the thermal pads and thermal paste.
The other thing I realised was that the case fans are connected to the Corsair controller. I had been using iCUE to control the fans, but I wanted more control in how they operated. Instead I used this Corsair Plugin for fancontrol, which gave me proper control over all of the fans. It was a bit of a pain because the Assisted Setup which is supposed to pair all inputs with fans doesn’t work, so you have to link them all up manually. What I did was set the GPU fans to be operated based on the GPU temp, the CPU fans on the CPU temp, and the case fans operate on a mix of the maximum of the two. I’m not sure if this is an optimum setup but it seems to work.
End result? GPU temperature reduced from 83C down to around 73C, with GPU hotspot temperature decreasing from 115C down to 80C. I’m not sure why the hotspot temperature drop is so dramatic, perhaps there were components that were not properly getting contact with the pad. Overall it was a bit of a scary project because I haven’t pulled the GPU apart before, but the end result meant I spent about $30 on pads and paste and got a huge drop in temperatures.
Finally I replaced all the custom software to control RGB with SignalRGB, which is a much better solution. Simple, lightweight and does the job perfectly.