GPU seems to be acting up, don't know the cause. - Hardware & Software

Hi everyone, So recently I've run into a major issue. It used to be minor, a bit annoying, but now it's become a full on problem and I'm having to tip toe and try not to trigger the issue again. About two or three months ago, I'd encounter random freezing while most commonly playing games, and less commonly using Chrome or just doing whatever on my desktop. These freezes would be hard crashes, and I'd figure that out by mashing caps lock on my keyboard to see if the light would turn on or off. It wouldn't. Unsure of what to do, I'd restart my computer and it wouldn't happen again -- sometimes not for a span of 6 or more hours, sometimes not for days, sometimes up to a week. Then it would happen randomly, once again. I figured out it was my graphics card causing the problem. Crossing my fingers and hoping that it was a software or driver related issue, I completely uninstalled AMD's Radeon Adrenalin drivers from my system, booted into Safe Mode, and reinstalled a newer version of the drivers. No dice, and it would happen again later on. This time, though, instead of being able to restart my computer seamlessly and boot back into Windows, my GPU fans would spin to 100%, with my motherboard letting out a few POST beeps indicating that it wasn't detecting a graphics card. I'd like to start troubleshooting the problem, but I'm at a loss of where to start. I'm not getting artifacting, my card's temperatures are great, my power supply has both a high enough wattage and is rated high enough to sustain the power demand of my system (Corsair CX450), and worse yet is the fact that I only bought this card a mere year ago. Here's some related specs: Asus M5a97 motherboard Powercolor Radeon RX 480 Red Dragon Corsair CX450 Any help would be sincerely appreciated.

Do you have any other GPU handy to rule out everything else on your system? Or another system to test it with?

I do. I have an old Sapphire Radeon HD 6870 that I could use. I suppose I'll have to stress test it and rule out any other components causing the issue, but I doubt it isn't the graphics card -- right now I'm just hoping it's a software related issue.

If you're still within the 2 year Powercolor warranty period I would get that RMA process started ASAP. Better safe than sorry. Even if it's not the GPU.

I think you shouldn't rule out other components. That "no card detected" error you described could simply be a driver issue caused by the hard crash of another component, or a motherboard error. Have you already checked the event log? Additionally, have you tried some FurMark stress testing?

I have, but it wasn't very helpful. I came back to my computer one morning and saw that it was on the Windows 10 login screen, so I took a look at the event viewer and saw a slew of errors that Windows categorized as "Hardware error". All of them would read LiveKernelEvent 141, which, according to Microsoft, is a video adapter / graphics related error. Moreover, I was playing Payday 2 last night and my card's signal went out twice -- both of the times it happened, Windows reported the LiveKernelEvent in my reliability history. As for FurMark stress testing, yes I have, but I don't know how long I should be running it. I know you run memtest86 for 8 or more hours, and Prime95 around the same time, but I'm not sure if it'd be best to push my card like that.

I've also have similar video adapter BSODs in the past. I was like fuck it and did a clean install. Not that it matters very much but do you have your CPU overclocked? When my overclock was becoming unstable it showed video driver crash errors when gaming. Raising my voltage a bit fixed the issue entirely

Not necessarily overclocked, but my Phenom II X4 980 runs at a clock speed of about 3.7GHz. Come to think of it, I remember waking up after running Prime95 over night and being greeted by the Windows 10 login screen again, indicating that it, of course, crashed. That might be a step in the right direction. I'd say reinstalling Windows is a last resort since I've done it so many times within the last year and a half, but I suppose I could try running Prime95 again and see where that leads me. This processor is pretty old -- at about 5 or so years I'd have expected it to outright quit, but here we are.

Run prime95 for about an hour and see if it brings up errors.

As for FurMark stress testing, yes I have, but I don't know how long I should be running it. 15 mins should be enough. A quick search showed that you can also just run this benchmark multiple times. Also, another idea: Put the card into the other PCIe slot (the black one), just to make sure there isn't some problem with the connector.

You have a 450W PSU, the rated power requirement is 500W minimum. The HD 6870 also has a requirement of 500W. Of course the actual usage might be lower, but this is based on overall consumption with other components. The reason for having a 500W PSU or higher generally stems down to efficiency and cleaner power as well as stability. Now I'm not saying the PSU could be to blame, but it's more than possible.

I let Prime95 run for about 14 hours while I was out. According to results.txt, it managed to get to the 4096K test before it kicked the bucket. No errors, just came back to a freshly rebooted computer with no indication on what went wrong. Rule out the CPU, or...? I'm not so sure about that personally. Before Corsair replaced the CX430 with the 450, I ran with a 430 and a 6870 for years before it quit. Exact same setup, too -- processor and all. Generally speaking, I'd expect this power supply to have failed after 3-4 months of system crashes, but it's still chugging along. The rails are in spec as well; no more than a .1v difference on each. That sounds like a good idea. I'll try that and the benchmark as well.

Swapped the PCI-E slot on my old motherboard and replaced the connector from the power supply with a molex to PCI-E. Made sure it was extra snug in the slot and the crashing stopped completely. On a new build now, and my drivers haven't crashed or hanged at all -- been about three days since I put it together. Thanks to anyone who provided their input, I really appreciate it.