Why is my GPU crashing about once a month?

[Michael_]

End of last year i upgraded my monitor and because of that i also had to "upgrade" the GPU because the old one was not
able to deliver 4k resolution.

This is my new GPU (Nvidia Geforce GT 1030):
https://www.mindfactory.de/product_...-OC-Passiv-PCIe-3-0-x16--Retail-_1247047.html

Well anyway shortly (maybe 1-2 weeks) after i installed the new GPU i clicked on a youtube video and the
screen went black. Win + Ctrl + Shift + B does nothing. Only a system restart helps.

Since then it rarely happens, about once a month i would say. Pretty much always when i click on a youtube video
or yesterday when i burned a video file to dvd (software uses gpu for processing).

I ran a lot of diagnostic and benchmark tools.
Furmark, Benchmark Heaven, CPUID HWMonitor, Prime95 & HeavyLoad.
I cant provoke it and i can't find something wrong.

Windows Error log says "LiveKernelEvent" Code 141.

Since it happens so rarely (about once a month) and i can't find something wrong i think its probably a faulty GPU?
Because i think if it was software/driver related it would happen all the time and not just once a month?
If it was lets say a faulty PSU i think Prime95 or Furmark would take it down?

Is there anything i can try to provoke it? Is there maybe a tool to change gpu load between Idle/WOT within seconds?
Can i diagnose this any further without it happening? It's hard to troubleshoot the issue like this.

FYI:
Before i had an AMD card. I just used Display Driver Unistaller
https://www.guru3d.com/files-get/display-driver-uninstaller-download,20.html
to fully remove the old drivers but that did not prevent it from happening again.

I have a still new in box AMD Radeon Pro WX 5100 which i probably will install if the issue should persist.
But i'm currently using a Mitsubishi Diamond Pro 2070SB (CRT) as my display and i'm not sure the WX 5100 supports refresh rates higher then 60hz (target is 85hz).
Officially no but that's also the case for the 1030 GT which in reality also does 100hz+ refresh rate.

 

[Michael_]

Push

 

[Michael_]

Push

 

[Darunion]

Its sounding like this is a really old pc, it is possible your power supply could be reaching the end of its life. I would use whatever you do when it happens as your means of testing. If that means 40 windows of youtube or rendering several videos at once.

Intermittent problems are always the worst and really does require a lot of awareness when it crashes as far as what all were you doing in what order. Also going through windows logs could help as well.

 

[Michael_]

Its sounding like this is a really old pc, it is possible your power supply could be reaching the end of its life.

Interestingly the +12V Reading in HWMonitor is just 10.427V
But i suspect that probably is just an inaccurate figure and also not going down under load.

The PC is about 10 Years old but i already replaced the PSU a couple years ago with a
500w fanless/passively cooled one from seasonic.

 

[GoodBoy]

Could be anything really. Video card and system memory are likely suspects.

 

[learners permit]

Yea I would look into that 12V rail by probing a molex connector with a reasonable quality digital voltmeter to see just how low it is as HW mon tends to be fairly accurate. I've used it diagnose failing psu's on several occasions over the years and followed up with the meter for confirmation. The molex yellow wire is 12v and black is ground and even inexpensive large diameter volt meter probes can be used on a molex with no risk of female connector pin tension damage.

 

[GoodBoy]

Yea I would look into that 12V rail by probing a molex connector with a reasonable quality digital voltmeter to see just how low it is as HW mon tends to be fairly accurate. I've used it diagnose failing psu's on several occasions over the years and followed up with the meter for confirmation. The molex yellow wire is 12v and black is ground and even inexpensive large diameter volt meter probes can be used on a molex with no risk of female connector pin tension damage.

Good idea.

If you can use a digital multimeter, and set it to record min/max on that 12v rail. The do testing, play games, whatever you were having a problem with, check the minimum observed voltage. If that rail is actually dropping below 11v, you need a new PSU, or RMA that one. This exact issue happened to me before, and I RMA'd a video card twice before figuring out that the PSU 12v output voltage was dropping under load into the 10v range. That is bad news for GPU's.
If the PSU stays at ~12v but the HWmonitor shows the GPU registering 10v still, that could be a GPU issue. The GPU does it's own voltage regulation even on the 12v supply, so that number is expected to be different than the PSU actual measured voltage. But at 10v it sounds too low, so could be an issue in the power stages on the card.