The main thing to watch is temperature. Make sure the GPU is not getting too hot.
If your PC hangs or crashes because your overclock is too high, it shouldn't damage anything, just restart your PC.
You can enable OSD in Presicion to monitor GPU and memory clock, gpu utilization, temperature, etc during stress testing.
Set power target and temp target to max.
Set the fan curves a little aggressive to make sure they keeps temps down as the GPU is stressed.
On my FTW3, I have them reach 90% by ~75C. I do not see temps above 65C in a warm room with the GPU fully stressed.
You will need a good benchmark program to stress the card.
I have found that EVGA OC Scanner is not sufficient, since I can pass a test on it but the OC settings will fail in other situations.
Unigine has free options such as Valley or Superposition. 3DMark is also a good one.
The idea is to raise the GPU clock one step at a time (say, by 15-25 MHz), then run the stress test/benchmark.
If it runs smoothly (no freezes, crashes, or artifacts on screen), then up the clock a bit and try again.
This is an iterative process to find the point where it finally crashes.
When you reach that point, back off the clock speed ~25MHz and retest.
If this passes again, then try the settings in your favorite games or other GPU intensive tasks to make sure there are no problems.
On the 1080ti's, max GPU clock seems to be just over 2000MHz. Mine went to 2062MHz.
You can now go through the same process with the memory clock.
When raising memory clock, watch the scores reported by the stress test/benchmark.
You may notice that, as you raise memory clock, the benchmark completes without crashing but the scores start going down.
This indicates you have found the best performance memory clock.
It sounds like a lot but once you get the hang of it, it's pretty easy to do.
post edited by Opolis - 2017/06/02 08:57:38