Tweaking the fan speeds is also essential if you want to optimize clocks for consistency (NVIDIA's GPU Boost start to throttle after the GPU enters the 70C range by default, although you can adjust that a bit with the temperature curve). Power and temperature targets are kind of whatever is necessary. I personally use 3DMark Port Royal, since it tests RTX as well, and 3DMark itself is configured to error out and tell you if there are any issues.

Keep raising the dot until it gets unstable and then pull it back down. Test in Unigine or Furmark or 3DMark or whatever your preference is.

If you want to undervolt, do it on a lower voltage. Screenshot of the GPU usage during the last step of the scan:Open up the voltage curve, see and note down at what voltage the curve starts to flatten, and then in the main window, drop the core clock by -100 to -200 and then go back into the voltage curve and raise the dot at the voltage you noted before as high as you can, and apply. So far I have already reinstalled the drivers, the next step would be running DDU. I am aware that the scanner relies on the Nvidia API but I would like to know if the low GPU usage during the scan is normal on 30 series cards, I wonder if swapping the card in could have impacted something in the scanning procedure, leading to low GPU usage during the test.

Games that use 70-80% of the GPU are fine but heavy ones that peg the usage near 100% lead to the display driver crashing and recovering. This leads to instability in games when using the curve generated by the scan. While running the OC scanner on the 3080, I have noticed that the GPU usage during the scan is rather strange.īasically it begins the scan with 50% GPU usage, then on the next step of the frequency curve it lowers to 25% usage and on the last step at 1075 mV the test is carried out at 2% GPU usage. I have recently replaced my faulty 2080 Ti with a 3080 Ti.

Afterburner oc scanner