Diagnosing Which Is Broken - PSU, GPU or MB?

grang on 23/10/2020 - 10:47
Last edited 23/10/2020 - 20:42

Been having on an off issues with a 8 year old PC that has been continiously upgraded.

Update 1: Seems GPU is fine.
—> Placed GTX 1650 low power card in problem PC. Still crashed. doesn't use 6 pin power
—> Placed GTX 960 in the full tower (usually a GTX 1070) and it worked fine on heavenly bench, steam bench, and playing Fortnite.

so must be PSU or MB. No used PSUs for sale… unsure if I should buy a good PSU to test/keep for future, as I may just buy a pre-built system (don't want to deal with component warranty issues).

Issue, on heavy GPU load, the screen blacks out. Display driver fails and PC has to be hard reset. Lately I run with both sides of the covers removed too, just in case, so likely not a case heating issue.

Put on new thermal paste for CPU and GPU and changed slots etc and for a short while that seemed to work, but now not working. I have 3 kids and they each have a PC, so I have another PC I can test things with.

My guess is it is either the PSU, or the GPU that is gone. (I have had a bad home Power quality issue for a few years that we didn't know about) One of the reasons I needed to replace the PSU a few years ago.

PSU: Coolermaster EX2-TREME 725
GPU: Nvidia 960 OC - 6 pin power
MB: ASRock Z68 Extreme 3 Gen 3 (with I5 2500k CPU)

PC works fine as is, until heavy load. So Minecraft rarely crashes, but fortnite or rocket league routinely crash, and often won't even load past opening screen.

How to diagnose? I was thinking of taking the 960 and put it in the Dell next to it which has a 1070 (Same driver) and see how well it runs, which should rule out whether the 960 or something else is the issue.

How can I test the PSU?

Its obvious an upgrade is due, but I want to get the new gen og graphics cards down in price and see the AMD reveal, so this baby has to limp on a bit longer. Kids also learn how to share again (uuuggghhhh).

Any suggestions welcome.

Computing Computer Component Motherboard Power Supply Unit Video Card

Comments

  • samyall on 23/10/2020 - 11:05
    +1

    In my experience, the issue has always been with the PSU. I never bought as fancy of a PSU as yours, but all mine only last max 3-5 years. They generally failed by not powering on at all (and not what you are describing) so take this advice with a grain of salt.

    Unfortunately, I think the only (safe) way to test a PSU is to replace it. You don't wanna mess around with opening it up unless you are an electrical engineer.

    You can rule out the MOBO if it has onboard graphics by taking out the graphics card and try running some CPU intensive stuff.

  • abb on 23/10/2020 - 11:10

    Can you swap the PSU with another machine? That's the simple way.

    Otherwise you'd be looking at measuring voltages with a multimeter or, preferably, an oscilloscope.

  • Jolakot on 23/10/2020 - 11:16

    Swap parts in from a working machine until it stops crashing, by the reviews on the PSU I'd start there. If it is the PSU, coolermaster has a 3 year warranty, so might be worth contacting them about it. If it's the GPU then just limp on for another few weeks, Nvidia is going to flood a bunch of 3080's right before the RX6000 announcement, seems to be an internal deadline right now.

  • ln28909 on 23/10/2020 - 11:31

    You could install something like MSI afterburner and check for GPU usage, clock, memory and power draw and see if there's any unexpected dip before crashing

    Then completely dial down both clock and memory, run it at full load (i prefer Heaven) then slowly dial it backup till it crash and check the power draw when that happens

    Then repeat on another GPU on the same system

    Then repeat all of the above on a different system

    This will test the PSU and GPU, since you're system is still running otw fine, I don't think it's a motherboard issue

    • grang on 23/10/2020 - 18:42

      yeah, I have installed MSI afterburner and massively upped the fan profile and moved power down to 80%
      I like your idea as a next step, to really bring performance down, and see what happens

  • petry on 23/10/2020 - 11:35

    almost certainly psu - borrow or use from another unit to see whats happens - then ramp up your gpu till you replicate failure.

    coolermaster build quality always questionable

  • skid on 23/10/2020 - 12:34

    It's the PSU (likely), RAM (if overclocked) or display driver (try older and newer versions). Had the same issues and the problem persisted even though I swapped out an old NVIDIA with a newer NVIDIA. I guess you could try using the internal GPU of the CPU too but not sure what that would prove if it stops crashing.

    • grang on 23/10/2020 - 18:44

      Internal GPU (intel i5 2500k) no issue, and neither the NVIDIA GPU at low to medium loads… ony at higher graphics specs… so definitely when the power ramps up

      The feeling I have is that the crashes happen so quickly sometimes, the chips would not have properly got hot either, so I have this PSU view too… (although the bad power issue could have gone through the PSU and damaged the MB)

      • skid on 23/10/2020 - 21:22

        Unlikely to be a MB fault. Have you tried resetting your BIOS settings back to factory or standard settings?

  • abctoz on 23/10/2020 - 12:36

    try downclocking/downvolt a bit i have a few gpu's that aren't stable at stock, the symptom is driver crashes, but they're perfectly stable at say -100mhz core/mem just use afterburner to tweak it. what happens is the components on the gpu degrade over time so it could just be isolated to the gpu

    • grang on 23/10/2020 - 18:45

      ok, that will be an easy test. I have afterburner and only changed fan and power, but not core/mem mhz.

      thx

  • knightslay2 on 23/10/2020 - 13:32
    +1

    Yeah I'd try an integrated graphics or a working graphics card. If the PSU is working well, I think there is a chance the graphics card. I have a similar setup and my graphics card died as well.

    At the end of the day, its an old system it might be more afforable to replace the motherboard, cpu and ram. It will be too expensive to find similar components for your current setup. I have been wanting to upgrade as well to a new AMD processors.

  • DarthAntz on 23/10/2020 - 14:11

    Ram issue? Also check your temperatures. Use gpu-z to see your gpu temps, ram and cpu usage. Maybe a can of air through your pc can help with overheating problems. How dusty is it?

  • Crosscade on 23/10/2020 - 14:38
    +1

    I'm going to buck the trend and say Graphics Card.

    Try running something like furmark that puts a high sustained load on the GPU. Furmark usually stresses the GPU more than any game would, so if it crashes under that consistently that would be a good sign it could be the GPU or PSU.

    Put the GPU into other system if you can and test it, if in another system and get crashes you will know for sure it is the GPU.

    If you work out it is the GPU, it could just be a matter of dialing back the clocks a 100mhz and you will have 90-95% of the performance. Then can save up for a new GPU.

    I've had this issue with my 980ti and it has had to be running on lower memory clocks for about a year now. Still sticking in there.

    • grang on 23/10/2020 - 18:48

      This sounds sensible. (and a good combination of notes and recommendations above)

  • garetz on 23/10/2020 - 15:00

    Crashing on heavy load usually means its the gpu, if it was the psu then it would hard reset on power spikes, which can happen at any time.

    Or another option is both are faulty.

    just do an isolation test for each part, and you will have your answer.

  • jackinyourbox on 23/10/2020 - 16:32

    Replace easiest bit first and test

  • grang on 23/10/2020 - 19:41

    Now that is interesting.

    My third boy uses a mini Dell PC and I had to install a gtx1650 in there. No extra power.

    Took out the low profile bracket. Put it in the problem PC. Upgraded the drivers, loaded rocket league and bam. Crash.

    So maybe PSU or maybe MB

    Still going to test the 960 in the other full size tower PC

  • grang on 23/10/2020 - 20:25

    Update II:
    Typing this on problem PC with 1650 card temporarily installed. Can't run anything heavy, even the test function in MSI Afterburner. Just crashes to black screen and fans go max.

    The 960 seems fine when I test in the i7 (which normally has a 1070 card).

    So I think I can safely rule out GPU.

    I doubt it is the HyperX two white ram modules I have (8GB x2)
    So either PSU, MB?

    No second-hand PSUs on gumtree to test, and the Full machine is a dell with only 425 PSU, so not enough to run the problem PC.

    MMM

    I have a multimeter. What would I test for, as the problem is only on load….?

  • Diverse on 24/10/2020 - 00:52

    Most likely PSU but you can try using a stress test on RAM if you're unsure about that.

    Could also be a motherboard problem too.

  • the Unforgiven on 24/10/2020 - 17:56
    +1

    I would go PSU like samyall wrote

    What make is it? Is it a good model or a shit model? It must be yonks old by now. Time for a new one.
    Here is an EVGA 700 BQ, 80+ Bronze 700W, Semi Modular, FDB Fan, 5 Year Warranty, Power Supply 110-BQ-0700-V3 (UK) for $145
    https://www.amazon.com.au/EVGA-Bronze-Modular-Warranty-110-B…

    $145 for 5 years is not too bad. It is EVGA, which I have; it ain't a 1200W P2 thou :p
    It is semi-modular. It is 700W bronze. If you can use it, get it

Login or Join to leave a comment
Login Join