FPS drop until reboot after overclocking

[Tomniverse]

The issue:

So I've been having a curious problem with my 1080 Ti. It's quite vague to describe, but sometimes (on stock clocks) my GPU suddenly has a strong FPS drop (30-60% decrease). This happens completely unpredictably and it doesn't matter whether this happens during gaming or during a benchmark, the FPS will remain lowered by 30-50% regardless of what application I run until I reboot the system. . 

 

I've found a way to reliably reproduce it, so maybe we can get to the bottom of this. If I raise the max voltages on MSI afterburner (+100 mv) while using my stable overclock (+60 core, +500 mem) or even stock clocks, after 2-3 minutes of a Heaven benchmark the FPS will halve and the power draw will drop from around 280-310 watts to 180-230 watts. Once this happens, the GPU will behave like this in any application until I reboot.

 

Now I don't think this has to do with instability of the overclocks. Going higher than +75 on the core clock crashes the application as can be expected, but at 0 to +60 mhz without overvolting it's completely stable except for that once in a week FPS and power drop. Temps shouldn't be the reason either as even with the +100 voltage and overclocks I rarely cross the 70°C threshold (max 84°C hotspot/VRAM temperature). Max 50% of the 16GB available RAM is usually used. VRAM usage is usually withing 2GB when the problem happens.

 

Things I've tried:

- Reinstall Nvidia drivers

- Check for parked cores during the issue

- Disable fast boot (both in bios and Windows)

- Set power settings to high performance in Windows and Nvidia control panel

- Unplug the USB controller (this was the culprit for someone else)

- Check task manager for memory leaks or some CPU-intensive background process

- Install Windows 11 (because why not)

- Reapply thermal paste

 

The weird thing is that this problem has been documented among varying Pascal GTX's the past 4 years and is not application-specific. It almost seems as if the whole GPU goes into a low power state that is only re-set by a full system reboot. It's not a memory leak, not a thermal issue, so what can it be? Maybe my 650W PSU isn't enough (but shouldn't it then just throttle or crash instead of going into this weird low power with normal clock speed state?). Any suggestions would be great! I'll provide whatever info or run whatever test you need!

 

My rig:

GPU - GTX 1080 Ti Gaming X 11GB

CPU - Ryzen 5 3600

RAM - G.Skill Ripjaws V F4-3200C16D-16GVKB

PSU - Corsair TX-M Series TX650M V2 

MOBO - Aorus B550m Pro

SSD - Kingston A2000 1TB

 

[frozensun]

I don't want even to read it,completly useless to OC just to get extra 50 Hz on core.

[Tomniverse]

Update: turned out it was a motherboard issue. When the new update came out (version 13), the issue was fixed.