[SOLVED] GPU maxing out

[roijjer]

I have evga GTX 970 graphics card with latest drivers. I just switched to a case with vertical mount for graphics card. It connects to MB with a riser cable. Since then, my gpu has been maxing out when I am in any game. If I restart my computer it acts normal sometimes. I'm having a hard time troubleshooting because it doesn't seem consistent and I couldn't figure out what triggers it. I looked at the task manager with the game open. The game is the process using all of the GPU. So, I don't think it's a mining virus. GPU use is nominal with no games running. As an example, when everything is normal, when I play world of tanks, I get constant 60fps with gpu at about 75%. When it's acting weird I have 20fps in game with 100% gpu. It just doesn't make sense to me. Could the interference in riser cable be causing this?

Thanks.

 

[mintycms]

are you using a coin mining software? if so, that maxes out when set to mine all the time instead of just when at idle

 

[roijjer]

I am not using coin mining software. To clarify. GPU use is low and normal when I'm not in game. I did a fresh reinstall of windows 10 and I only have FIFA 19 and World of tanks installed right now. When I start any of these games after a computer restart, it is all normal. GPU use about 75% and game have >60fps. BUT, sometimes, in game FPS is about 20+ and GPU use is 100% when I check in task manager. The game process is the only one taking up GPU use. I still can't pinpoint what activity or setting triggers it.

 

[roijjer]

First image with high gpu use. 2nd picture is when it's normal. I was able to consistently reproduce the issue now. If I shutdown the computer and turn it back on, it have this high gpu use and lag. If i restart, it acts normal.

 

[roijjer]

I kinda found a work around. It has to do with the riser cable. For some reason, it appears that the graphics card doesn't get the correct clock settings when I cold boot when using riser cable. I installed EVGA precision xoc and put overclock value a few mhz above or below the stock clock and that seems to have alleviated the issue. I feel like it's still not matching the performance from restart. But, I think it's acceptable for now.

EDIT: I think I really got it figured out now.. Motherboard have a PCIE setting with Gen1.0 and Gen 2.0 options. It was set to AUTO. I think when i cold boot, the graphics card isn't responsive enough. So, it may select Gen 1.0 automatically which likely reduce the bandwidth of the slot. I set it to Gen2.0 manually and GPU is consistently about 75% on World of tanks garage screen as it should be.