nikitam

Waterman, yes, the GPU temp went down to 50C in Furmark (fullscreen 1080p stress-test with default settings). However, my ambient temperature is also just 20C, this can contribute to the difference between our numbers too. My CPU temp in Cinebench R15 is 66-68C after repasting, in before it was jumping above 70C. I notice bigger effect from the GPU repasting, since the original pre-applied thermal compound had no contact with one of the corners of the GPU die. I am in fact surprised it worked at all, given that about 20-25 % of the chip had no contact with the cooler.

Hey Waterman, glad to hear your machine is now working properly! I would say keep an eye on the temps, but I've never heard that anyone would replace pastes frequently, like every year or so. Two years, as zguy85 suggests, would make sense I think. I also just re-pasted my GPU with a Noctua compound, GPU temperatures went down from 85C (and reboot) to just 50C in Furmark stress-test. zguy85, thanks for the advice!

Hey zguy85, thanks for sharing your experience! What thermals did you get before and after? Are there any things to pay extra attention to during the re-pasting? For instance, I understand that there is no heat spreader between the GPU die and the cooler, and I was a bit worried that I can crack it by applying to much mounting pressure.

I don't have the issue with internet browsing (yet?), are you sure your system is clean? I don't think the GPU is getting outdated this fast, no, a hardware/software problem is more likely. For me the GPU works all fine until around 90C (which is the thermal limit anyway), but it shouldn't actually get there with proper cooling. There are a few recent threads from users with similar or related issues, but I couldn't find exact reason for malfunction in these threads. My current guesses are 1. Bad GPU block-GPU die contact due to dry thermal paste that needs replacement 2. Gunk inside GPU cooling block, similar to what happens in custom water cooling loops after a while 3. Pump not working, even though Corsair Link shows that it does 4. Software error, maybe GPU should start throttling to prevent overheating instead of going straight to 90C? 5. Software error, maybe pump is OK but it should speed up instead of fluctuating around 2400 RPM? Maybe this all is a coincidence, but I started noticing the overheating after Windows feature update and Corsair Link 4.4.4.9 update, along with some recent NVidia drivers. Corsair Mint, thanks for the information! I contacted the Support team. P.S. This https://forum.corsair.com/v3/showthread.php?t=195961 might also be related to our problem.

Hi Waterman! You are definitely not alone here! I have a similar Corsair One with 1080 Ti and i7-7700K, purchased late 2017, and I noticed drastically worsened GPU thermals recently. In some games I get image artifacts all over the screen after playing for a while, then it goes black and the only way out is hard system shutdown. The machine is clean, I even removed the side panels to clean the dust which I couldn't vacuum-clean from the outside. I have the last Windows feature update, which required manually installing Corsair Link 4.4.4.9 from the link on the forum to prevent the blue screen during update due to Cpu-Z. I have performed stress-testing in Furmark, I attach a screenshot with thermals graph from Furmark and from Corsair Link. Main system fan on top of the case was set to Max RPM manually from the beginning. What I observe in Furmark is that GPU temp starts to climb up after a while, GPU fan ramps up, but the GPU pump stays at constant RPM around 2400 all the time. Eventually GPU temp gets to 85C at which point I stopped the test.

I have recently installed Ubuntu 16.04.3 LTS on my new Corsain One Pro (see the thread http://forum.corsair.com/v3/showthread.php?t=168371 ). It was quite simple, the only trick being that you have to add "nomodeset" option while booting the live installer from the USB and when booting the installed version of Ubuntu for the first time. This launches the system in some sort of a legacy graphics mode and allows you to update the NVidia drivers. WiFi and sound worked smoothly. My plan was to use the system for some machine learning projects, hence the need for Ubuntu. In fact, Corsair One is a very attractive option for this type of applications given its cooling system. So I've installed CUDA toolkit and other necessary software without any issues... But: the cooling system of my Corsair doesn't seem to adjust when PC is under heavy load in Ubuntu. Basically, I got the CPU up to 80C and the cooling still operates at its default settings. Of course, this completely forbids any kind of serious work in Ubuntu. And, while possible, machine learning on Windows is often much more problematic. So my question is: is there a way to resolve this problem? I've tried using ld-sensors, but it doesn't detect the fans or the pumps in Corsair. After doing some further googling, I found out that this problem (missing control over water cooling) also exists for other Corsair coolers. Ideally, a Corsair Link version for Linux would solve all these issues. But I would like to know what are the possible solutions now. For instance, can I adjust the pump & fan performance in BIOS?