PC restarting in GPU-intensive tasks

[TheBigToni]

For clarification, by "restarting", I mean completely powering off (no lights, no fans), then powering back up on its own. No BSOD, no error messages, no event logs.

 

Hi, my PC recently started restarting by itself in GPU-intensive tasks like gaming and downscaling 4K video using high-quality scalers (SSIM). I think I've narrowed it down to XMP being enabled, as I had 5 restarts on FurMark with it enabled, and 0 with it disabled after 10 runs. However, the problem is pretty random and this could entirely be coincidence. For instance, I went poking around inside and reseated my GPU and the power connectors. The issue appeared to be solved, until run 5 on FurMark.

 

I've had the same setup for at least 4 years, most of the time with XMP enabled. The CPU and mobo I've had since mid-2017. This is the first issue I've had like this. The only real change is that I had to replace both GPU fans a month ago, but this issue started just a few days ago.

 

I've monitored temps, they appear to be fine. GPU gets to a max of ~65 C. CPU gets in the mid-80s (only in Prime95, I know I have to re-paste it soon but I'm sure that's not the problem).

 

I've ran MemTest86, which showed 0 errors. I've downgraded GPU drivers to ones I know are stable. I ran multiple CPU stress tests/benchmarks, Prime95 and Cinebench with no problems (the crashing happens when nothing is going on with the CPU, anyway).

 

Windows 10 22H2

CPU: 7700K @ 4.2Ghz (stock, nothing changed)

GPU: EVGA Black 2080

RAM: G.Skill RipJaws V 3200Mhz (2133Mhz without XMP)

PSU: Corsair RM750x 750-watt

Motherboard: MSI Z270 SLI Plus, BIOS up-to-date (as much as possible)

 

What could this be? CPU or mobo showing their age? PSU starting to fail? I'm thinking of borrowing a multimeter to test my PSU, but nobody I knows has one.

Edited March 22, 2023 by TheBigToni
Clarification

[Ashleyyyy]

when you say "restart" do you mean a blue screen of death?

[TheBigToni]

2 minutes ago, Ashley MLP Fangirl said:

when you say "restart" do you mean a blue screen of death?

No, it just completely powers off and starts back up. No BSOD, nothing in event logs, no error checking on startup.

 

I'll add that to the post, my bad.

[Ashleyyyy]

1 minute ago, TheBigToni said:

No, it just completely powers off and starts back up. No BSOD, nothing in event logs, no error checking on startup.

 

I'll add that to the post, my bad.

that's interesting. do you have another computer? if you do you could swap parts from this system into that to see if any particular part is faulty. 

[TheBigToni]

1 minute ago, Ashley MLP Fangirl said:

that's interesting. do you have another computer? if you do you could swap parts from this system into that to see if any particular part is faulty. 

Unfortunately no. I was close with my former IT teacher, I'm thinking of e-mailing him to see if he has any spares.

[BetteBalterZen]

Maybe disable all OCs, then DDU, install latest driver and see how it works out? 

[TheBigToni]

1 minute ago, BetteBalterZen said:

Maybe disable all OCs, then DDU, install latest driver and see how it works out? 

I should've tried DDU'ing, yeah. I'll go do that real fast.

[Ashleyyyy]

1 minute ago, TheBigToni said:

Unfortunately no. I was close with my former IT teacher, I'm thinking of e-mailing him to see if he has any spares.

you could also try to narrow it down further by stressing only 1 part of the system and see when it crashes. 

[BetteBalterZen]

2 minutes ago, TheBigToni said:

I should've tried DDU'ing, yeah. I'll go do that real fast.

In the past week or so, I also had system freezes / lockups / shutdowns, I did however have logs in the Event Viewer pointing me towards Nvidia driver. I describe it in my latest status update. Anyhow, I fixed my issues by using DDU from Safemode and then installing the latest drivers again. 

[TheBigToni]

1 minute ago, Ashley MLP Fangirl said:

you could also try to narrow it down further by stressing only 1 part of the system and see when it crashes. 

It only crashes when I have XMP enabled and perform a GPU stress test (or during gaming). I've completed multiple more tests since I made this post with no crashes.

[Ashleyyyy]

Just now, TheBigToni said:

It only crashes when I have XMP enabled and perform a GPU stress test (or during gaming). I've completed multiple more tests since I made this post with no crashes.

it could be the memory or mobo aging then if it's only an issue with XMP on. 

 

since your system is quite old and intel based, XMP shouldn't affect performance that much. a simple fix for this then would be just turning XMP off and running the system that way. 

[johnno23]

XMP over the years has giving me more issues when gaming than anything else in the systems I have built.

I stopped using XMP about 7 years ago now.

looking at the XMP profile and then manually entering the timings does seem to help for some people but I simply never bothered as I really dont need that extra few fps as I play for fun and never in a competitive scenario..

[TheBigToni]

22 minutes ago, BetteBalterZen said:

In the past week or so, I also had system freezes / lockups / shutdowns, I did however have logs in the Event Viewer pointing me towards Nvidia driver. I describe it in my latest status update. Anyhow, I fixed my issues by using DDU from Safemode and then installing the latest drivers again. 

Unfortunately performing a DDU and installing latest drivers did not fix the issue with XMP enabled.

 

15 minutes ago, johnno23 said:

XMP over the years has giving me more issues when gaming than anything else in the systems I have built.

I stopped using XMP about 7 years ago now.

looking at the XMP profile and then manually entering the timings does seem to help for some people but I simply never bothered as I really dont need that extra few fps as I play for fun and never in a competitive scenario..

Yeah, it's just that, I mean, I paid for 3200Mhz RAM (not too expensive, but still) and for it to just randomly stop letting me use it at 3200Mhz is kinda weird. And I have the problem of "if it can go high, I want it high".

 

I'll just keep XMP disabled until I can actually diagnose the issue using some spare parts or tools. It's not a huge deal. I'm just hoping fixing it won't be expensive.

[TheBigToni]

Update: XMP was not the problem. I don't know why my PC decided to pretend like it was for a day, but it has returned to crashing.

[TheBigToni]

9 minutes ago, TechlessBro said:

Just set the power limit on the gpu down 10% or more. Limit boost clocks.

test

if it works then likely PSU is having issues.

 

Do the same with cpu set the power limit down and disable and boost.

test

this could be mobo power or the psu still

 

Very good advice, thank you. Such an obvious thing that I glanced over.

 

The only issue is that, yes, it seems to work on a lower power limit... But then I set it back to 100% and it still works. 220W on 100% limit.

 

I'll keep testing and let you know. Thanks again for the advice.

[TheBigToni]

18 minutes ago, TechlessBro said:

Just set the power limit on the gpu down 10% or more. Limit boost clocks.

test

if it works then likely PSU is having issues.

 

Do the same with cpu set the power limit down and disable and boost.

test

this could be mobo power or the psu still

 

Okay, so even with the GPU on 58% power limit, I crash almost instantly when putting a load on it. I guess I'll try pulling my GPU and trying to do the same load on the CPU only.

[TheBigToni]

50 minutes ago, TechlessBro said:

Hang on do you have the old gpu fans still? Check the specs for them and the new ones.

it is possible the fans are over drawing power on the gpu, since it will be a small circuit and could effect something else on the card.

 

can you limit the fan speed on the gpu? Set it to 40% max and test.

Unfortunately the GPU stopped powering on for some reason. No matter what I try, it just won't power on: Swapping the PCIe slot, swapping the cables, etc. I'm taking it as a sign to rebuild my rig from scratch. It was about time I upgraded, anyway.

 

Thanks for your help, though.

 

*there have been countless other issues with this rig that have seemingly been undiagnosable, but have "fixed" themselves in the past- just to come up again, then disappear again, like a cycle. There is clearly more wrong with it than I care enough to fix. I'd rather just get a new PC as a whole and maybe use this one as an unimportant home server.