Graphics card crashes, CPU or Motherboard?

[omnichad]

I have a computer that's crashing out of higher-end games like Fortnite or PUBG at random points after at least 10-15 minutes of play (AMD driver crashes). This is on a fresh install of Windows 10 with the latest drivers. Worst of all, I can't make it crash myself with random clicking around - but customer playing games does crash.

I've tested the RAM, swapped out power supply, and suspect storm surge/brownout damage due to a bad hard drive that had to be replaced just prior to discovering this issue.

The GPU is a Radeon RX 480 and it works fine in a test system I set up with no crashes. So now I'm down to either the CPU or Motherboard.

I know that PCIe lanes are more or less a direct connection between the CPU and the graphics card. But that there also might be a PCIe switch on the motherboard itself. I usually prefer to replace the CPU and motherboard together regardless, but if you were going to do it one step at a time, which would you replace first? I'm inclined to just have the customer replace both as the extra inconvenience and additional testing isn't worth it for the cost savings.

 

[Slaters Kustum Machines]

I would suspect MB, if you think a power issue is the cause. However, are you able to install another video card and have them test it for a day or two?

 

[omnichad]

The card works fine in another system. They're testing/using that system for now. To put in an equivalent card would cost me $300+ on something I probably couldn't use if it turned out not to be needed. A weaker card might not stress it enough to reproduce. Is there a reason to try another card if that card is proven to work fine?

 

[Slaters Kustum Machines]

The card works fine in another system. They're testing/using that system for now. To put in an equivalent card would cost me $300+ on something I probably couldn't use if it turned out not to be needed. A weaker card might not stress it enough to reproduce. Is there a reason to try another card if that card is proven to work fine?

No, if that card is working fine in another system, I would be looking at replacing the MB in the original system.

 

[DataMedics]

Is anything overclocked? Could be the customer is just pushing it too far.

 

[omnichad]

No OC. Was working fine before the hard drive failed and was replaced. So there is likely over/undervoltage damage somewhere.

 

[fincoder]

Worst of all, I can't make it crash myself with random clicking around - but customer playing games does crash.

For cases like this I use Furmark to stress-test the graphics system. Tried anything like that?

The card working OK in another system might indicate an OS or driver issue in the first system. Maybe a clean install of Windows on a spare drive might be worth a try.

 

[GTP]

You say you've swapped out the PSU? Did you swap it out for a higher wattage PSU?
I've seen this many times.
This type of crash can (mostly) be traced back to a faulty PSU that dies under load.

If it crashes after replacing the PSU with a better one, then I agree with @fincoder on running a benchtest/stresstest on it..

 

[omnichad]

When I replaced the failed drive on this computer, it got a fresh install of Windows. It's on the same driver version as the second PC that I set up.

The card itself is definitely fine, it managed to pass MSI Kombustor on the *bad* computer - ran it at full load at 90 degrees celsius for over 10 minutes with no glitch. Didn't run Furmark, but I can't remember the name of the other stress tester I used. These tests use a lot of processing power on the GPU but not a lot of bus bandwidth I don't think. I think it's heavy load on the PCIe lanes that trigger this issue. I even did a VRAM memory test and it passed perfectly well on the bad motherboard/cpu. It's only in real-world gameplay that the problem is reproduced.

 

[omnichad]

I used the original PSU on the second computer for testing just to prove to myself that it wasn't a factor. It's a midrange 600W power supply, but not a high-draw CPU or motherboard and only a single graphics card.

I should also say that this PC ran fine for 2 years before a power surge knocked out the hard drive and whatever else is failing here.

 

[fincoder]

it got a fresh install of Windows. It's on the same driver version as the second PC that I set up.

The card itself is definitely fine, it managed to pass MSI Kombustor on the *bad* computer - ran it at full load at 90 degrees celsius

OK, looks like you've already considered my ideas. This really is a tricky problem!

 

[GTP]

before a power surge knocked out the hard drive and whatever else is failing here.

I reiterate, swap the PSU for a new one and see what happens.

 

[mikeroq]

Worst of all, I can't make it crash myself with random clicking around - but customer playing games does crash.

Ok, can you clarify?
By "random clicking around" do you mean you are doing random stuff or actually opening the game?

You say the customer can crash it, are you working on it in your shop and they are trying at home, or are you both trying at same location same time?

 

[nlinecomputers]

I reiterate, swap the PSU for a new one and see what happens.

Gotta agree with this. If there is ANY reason to think a power surge transitioned that PSU I would replace it as a matter of course. Flaky power can do weird things and older equipment can be more sensitive to dirty power than a brand new motherboard.