gpu Weirdest GPU crash

[Hobgoblin101]

(I am not this active here, so I am sorry in advance if I put this in the wrong section or something. However, the stock-exchange is where I normally go to for forum discussions and this is too personalized a problem to post there).

 

Very recently I had the weirdest crash screen I have ever seen, where a checkered grid section across both my screens of 10px*10px regions started displaying RGB noise. (Checker patter if you imagine white spaces being what was previously on my screens, and the black being the noise). This pattern appeared for 5 secs (not enough time to whip out my phone and take a photo), also my speakers did the buzz of which was usual for when my computer is off.

 

Prelude:

• According to Cinebench, my GPU performed the same as similarly spec-ed machines, however after I installed driver updates through AMD Radeon settings (23.20.15033.5003 aka 18.3.4) my performance when down to 10% of what it was. I continued using it for a day in case there were some hidden windows updates happening, shut it down overnight, and booted up the next day to the same performance. I then (within device manager) reverted back to my previous drivers (23.20.1507.3010) and received 70% the perf I was getting before. I then wiped all AMD software and drivers from my computer (through wizards, then hunting leftover files). Then reinstalled my drivers +AMD Radeon software, and received 90% performance.
• Well before this even I had an AMD r9 270x die while in this motherboard and got my current graphics card using the refund I got from warranty.
• Then well before that (1yr+) my USB 3.0 hub on my motherboard died with windows telling me it was shocked device or something like that (windows now doesn't detect it exists).

I am now not sure if this problem is my GPU or motherboard, and I fear it will persist and get worse.

Is it possible that my motherboard is dying and killing GPUs with it?

 

[Edit]

When I tried rebooting it would just display black (LCD monitors turned on, so they must have been receiving data). I had to leave it off for 10mins before booting successfully. Then it booted normally.

This seems like a weird error, and based on past experience I would expect it to continue until manually fixed.

 

Specs:

• Motherboard: Asrock b85m pro4
• CPU: Intel 4th Gen i7-4790 (3.6Ghz)
• RAM: 2x 4GB 1300hz DDR3 + 2x 4GB 1600hz DDR3
• GPU: AMD RX 580
• Storage: [Shouldn't matter]
• Monitors: Two HD 60hz monitors connected via DVI and HDMI

[Jurrunio]

probably the motherboard shorts itself, but check with another display cable first.

[Hobgoblin101]

Just now, Jurrunio said:

probably the motherboard shorts itself, but check with another display cable first.

Oh sorry, I forgot to say that whenever I rebooted it would display black (LCD monitors turned on so they were getting data).
I had to leave my PC off for 10mins then restart and it booted normally.

[Ace McPlane]

4 hours ago, Hobgoblin101 said:

Oh sorry, I forgot to say that whenever I rebooted it would display black (LCD monitors turned on so they were getting data).
I had to leave my PC off for 10mins then restart and it booted normally.

Still test it with another cable if it was only on one screen with the artifacting.

Whether or not it got a single does not mean that it's something that you still shouldn't test.

 

Anyways,

I'd reseat the graphics card next.

[Hobgoblin101]

13 hours ago, stateofpsychosis said:

Still test it with another cable if it was only on one screen with the artifacting.

Whether or not it got a single does not mean that it's something that you still shouldn't test.

 

Anyways,

I'd reseat the graphics card next.

I haven't tried different monitor cables yet (I don't have a spare HDMI). However, it is not really a high priority since the artefacts occurred across both screens which says to me it is an internal problem.

 

I reseated my GPU, as well as checking all power connectors on my motherboard and the 6pin GPU connector.

 

I have now run multiple Cinebench GPU tests, as well as reseated my GPU multiple times, however now I only get 50% the performance I have previously gotten.

Using Gigabyte's AORUS software I have switched between passive and active cooling with each reseat, however, in both modes, I still get 50% the score I normally do.

I presume it is not thermal throttling since it only reached 40c in both tests, and when the fan was active it only went to 33% speed.

(Note: I only tested passive once with each boot, and it was the first test I ran on each so it would start at a lower temp).

Also my GPU usage according to AORUS stays at 100% during the animatic part of the test.

 

When I first posted this thread my score was 122, the worst case I had gotten with this card was 30, after I reinstalled all drivers I got back to 110fps. Now I am sitting with a score of 68.46.

 

Am I correct in presuming that this is most likely a problem with the motherboard / PCI-E connector?

[Hobgoblin101]

14 hours ago, stateofpsychosis said:

Still test it with another cable if it was only on one screen with the artifacting.

Whether or not it got a single does not mean that it's something that you still shouldn't test.

 

Anyways,

I'd reseat the graphics card next.

I haven't tried different monitor cables yet (I don't have a spare HDMI). However, it is not really a high priority since the artefacts occurred across both screens which says to me it is an internal problem.

 

I reseated my GPU, as well as checking all power connectors on my motherboard and the 6pin GPU connector.

 

I have now run multiple Cinebench GPU tests, as well as reseated my GPU multiple times, however now I only get 50% the performance I have previously gotten.

Using Gigabyte's AORUS software I have switched between passive and active cooling with each reseat, however, in both modes, I still get 50% the score I normally do.

I presume it is not thermal throttling since it only reached 40c in both tests, and when the fan was active it only went to 33% speed.

(Note: I only tested passive once with each boot, and it was the first test I ran on each so it would start at a lower temp).

Also my GPU usage according to AORUS stays at 100% during the animatic part of the test.

 

When I first posted this thread my score was 122, the worst case I had gotten with this card was 30, after I reinstalled all drivers I got back to 110fps. Now I am sitting with a score of 68.46.

 

I rather forcefully turned my case onto its side to see if I would be able to move my GPU to the second PCI-E slot to see if it was just the slot, however, because of case restrictions it will not fit there. Interestingly I gently stood it back up, turned it on, and ran Cinebench and got my normal scores...

I think updated my drivers to the latest available, rebooted, and re-tested. Got the same score...

 

Am I correct in presuming that this is most likely a problem with the motherboard / PCI-E connector?

[Ace McPlane]

On 4/28/2018 at 12:22 AM, Hobgoblin101 said:

I haven't tried different monitor cables yet (I don't have a spare HDMI). However, it is not really a high priority since the artefacts occurred across both screens which says to me it is an internal problem.

 

I reseated my GPU, as well as checking all power connectors on my motherboard and the 6pin GPU connector.

 

I have now run multiple Cinebench GPU tests, as well as reseated my GPU multiple times, however now I only get 50% the performance I have previously gotten.

Using Gigabyte's AORUS software I have switched between passive and active cooling with each reseat, however, in both modes, I still get 50% the score I normally do.

I presume it is not thermal throttling since it only reached 40c in both tests, and when the fan was active it only went to 33% speed.

(Note: I only tested passive once with each boot, and it was the first test I ran on each so it would start at a lower temp).

Also my GPU usage according to AORUS stays at 100% during the animatic part of the test.

 

When I first posted this thread my score was 122, the worst case I had gotten with this card was 30, after I reinstalled all drivers I got back to 110fps. Now I am sitting with a score of 68.46.

 

I rather forcefully turned my case onto its side to see if I would be able to move my GPU to the second PCI-E slot to see if it was just the slot, however, because of case restrictions it will not fit there. Interestingly I gently stood it back up, turned it on, and ran Cinebench and got my normal scores...

I think updated my drivers to the latest available, rebooted, and re-tested. Got the same score...

 

Am I correct in presuming that this is most likely a problem with the motherboard / PCI-E connector?

I'm not sure.

Anytime a GPU is artifacting, it usually means the GPU itself is dying.

Can you post some screenshots of the actual artifacting so I can be sure that's what it is.

 

[Hobgoblin101]

On 30/04/2018 at 9:41 PM, stateofpsychosis said:

I'm not sure.

Anytime a GPU is artifacting, it usually means the GPU itself is dying.

Can you post some screenshots of the actual artifacting so I can be sure that's what it is.

 

3

I haven't had any GPU artifacting since then. Also last time it happened it displayed for such a short time before my PC rebooted that I couldn't get out my phone quick enough.

 

My GPU performance has gone back down to 50% again, and I isn't coming back so far.

 

When I searched through my Event log (windows) and couldn't find any unexpected shutdown entries or anything that seemed like a mention of the crash.

 

[Ace McPlane]

2 hours ago, Hobgoblin101 said:

I haven't had any GPU artifacting since then. Also last time it happened it displayed for such a short time before my PC rebooted that I couldn't get out my phone quick enough.

 

My GPU performance has gone back down to 50% again, and I isn't coming back so far.

 

When I searched through my Event log (windows) and couldn't find any unexpected shutdown entries or anything that seemed like a mention of the crash.

 

Okay, good to hear.

Let us know if the issue returns.

 

 

 

[Hobgoblin101]

13 hours ago, stateofpsychosis said:

Okay, good to hear.

Let us know if the issue returns.

 

 

 

I feel like you have miss understood, my 500AUD graphics card is currently only capable of the performance equivalent of a 100AUD card.
Also, I would really like it if I could solve the problem without going through warranty since last time I did that I had to wait six months for my card to finally go through the warranty process to end up with them posting me back a dead card saying "fixed". After a further 2mo of calls, I convinced them to take it back again and give me a refund. (With that money plus some savings I got my current card).

 

I was hoping it might be a software issue or something I could fix myself it a relatively low risk and cost factor. However, it seems I will need to spend a period of time without my graphics card again.

 

Does anyone know of the best way to attempt to trigger any existing faults with a PCI-E slot? Because if I can do that while using my back up graphics card (r7 250) then I can confirm if the problem is with my motherboard since I have already had a GPU die in this PCI-E slot before.

[Hobgoblin101]

I checked the behaviour of what happens when I attempt to boot my PC when the GPU's 8pin power connector is unplugged. Resulting in the computer seeming to boot, however, both screens remained black. I then plugged the connector back in, rebooted had the same Cinebench scores, then rebooted again and have the same performance as it should have.

Note: Sometimes my computer boots and has the correct/expected performance and sometimes it is 90% what it should be, and other times 50%. This would say to me that it is a loose connector. However, between these reboots, I don't even touch the case, let alone any cables inside the case and still get differing results.

 

Also the fact that last time I had to deal with their warranty system and they sent me back a dead card that wouldn't let my motherboard even get to BIOS tells me that they aren't going to pick up on this erratic problem and they will just send it back to me, wasting my money and time.

[Ace McPlane]

On 5/4/2018 at 10:18 PM, Hobgoblin101 said:

I checked the behaviour of what happens when I attempt to boot my PC when the GPU's 8pin power connector is unplugged. Resulting in the computer seeming to boot, however, both screens remained black. I then plugged the connector back in, rebooted had the same Cinebench scores, then rebooted again and have the same performance as it should have.

Note: Sometimes my computer boots and has the correct/expected performance and sometimes it is 90% what it should be, and other times 50%. This would say to me that it is a loose connector. However, between these reboots, I don't even touch the case, let alone any cables inside the case and still get differing results.

 

Also the fact that last time I had to deal with their warranty system and they sent me back a dead card that wouldn't let my motherboard even get to BIOS tells me that they aren't going to pick up on this erratic problem and they will just send it back to me, wasting my money and time.

Try uninstalling your GPU driver in safe mode using DDU

following this guide: https://windowsreport.com/display-driver-uninstaller-windows-10/

then download/install your GPU driver again and if that doesn't improve things, try doing the same thing but rolling back to a slightly older driver.

[Hobgoblin101]

On 07/05/2018 at 9:39 PM, stateofpsychosis said:

Try uninstalling your GPU driver in safe mode using DDU

following this guide: https://windowsreport.com/display-driver-uninstaller-windows-10/

then download/install your GPU driver again and if that doesn't improve things, try doing the same thing but rolling back to a slightly older driver.

Thank you for your help so far,

I did as you said and reinstalled my GPU drivers from the disc that came with it (17.10.1030-170306a-311841...) and it had normal performance (not this may have just been luck due to the inconsistency of the problem, though I did the test with two different boots).

 

I updated to 18.2.1 ran one test and got a normal result.
I have not updated to 18.4.1 since it is an "optional" update.

 

I suspect that this has not fixed the problem due to the inconsistency of the issue, however, if the problem does not arise in 7days I will mark it as complete.

[Ace McPlane]

On 5/13/2018 at 1:00 AM, Hobgoblin101 said:

Thank you for your help so far,

I did as you said and reinstalled my GPU drivers from the disc that came with it (17.10.1030-170306a-311841...) and it had normal performance (not this may have just been luck due to the inconsistency of the problem, though I did the test with two different boots).

 

I updated to 18.2.1 ran one test and got a normal result.
I have not updated to 18.4.1 since it is an "optional" update.

 

I suspect that this has not fixed the problem due to the inconsistency of the issue, however, if the problem does not arise in 7days I will mark it as complete.

No, never use the disc.

The driver on the disc is super old and no good.

We never use discs for drivers anymore.

 

You want to go to AMD's website and download the newest driver from there.

[Hobgoblin101]

On 15/05/2018 at 11:39 PM, stateofpsychosis said:

No, never use the disc.

The driver on the disc is super old and no good.

We never use discs for drivers anymore.

 

You want to go to AMD's website and download the newest driver from there.

Well, I still used the disc ones because I presumed they would be stable.
Then, as I stated, upgraded to the latest non-optional drivers for AMD.
I am now consistently getting 89% the performance I should be getting, which is acceptable.

[Hobgoblin101]

35 minutes ago, Hobgoblin101 said:

Well, I still used the disc ones because I presumed they would be stable.
Then, as I stated, upgraded to the latest non-optional drivers for AMD.
I am now consistently getting 89% the performance I should be getting, which is acceptable.

I am now getting 120fps scores on Cinebench, of which is what my graphics card normally gets

I am going to attribute the solution to this problem being the use of DDU, as it removed a lot of old ATI/AMD drivers as well as some nVidia drivers from old graphics cards. However, I cannot be sure if this was actually the solution to the problem as it is so inconsistent that it may just be away for a while.

 

It seems odd to me though that switching from 18.2.1 -> 18.5.1 made my score go from 90fps -> 120fps :\
I mean I know better drivers can boost performance, but I didn't expect a 25% increase.

[Hobgoblin101]

Back down to 50% performance again.

I haven't uninstalled or installed anything but used my computer casually.

 

Using one of Linus' previous videos as a reference if some of the individual lanes within my x16 slot died they shouldn't have much of an effect right?

Therefore we can rule out the motherboard as being the issue, and I need to try and go through with the warranty of my graphics card.

[Hobgoblin101]

@stateofpsychosis I finally got an image, and some new info.

So I was about to lodge the warranty claim when I ran another test to be absolutely sure about everything, and that there was nothing I could do.
Then I noticed that Gigabyte's AORUS software stated that my GPU was clocked to 300MHz, and it never increased during workload. I thought that was odd, so I tried changing settings and it did nothing. Then I updated AORUS, restarted, still nothing. I then installed the 'optional' AMD Radeon update available, and after a reboot, I noticed, that when I opened things, my GPU clock went up from 300MHz to 304MHz then back down.

 

I excitedly ran a Cinebench test, thinking I had fixed it, and saw the GPU clock rate increase to the correct 1340MHz. Then I got the crash. The same one as before.
I was only able to get one crappy image because I didn't realize my flash was on so it took forever. But right after the photo got taken both monitors claimed to of lost signal.
After waiting it didn't come back, so I hard reset.

I then repeated the process.

Opened AORUS;
  GPU: 300MHz
 VRAM: 1200MHz
Opened Cinebench;
 It ran normally, getting perfect results...
 I am pretty sure that my GPU is probably in an unstable state now, and I expect to see more random crashes and will keep you posted if they happen.

 

It is also interesting to note, that the effect occurred on the first screen (cinebench) before it appeared on the second screen about half a second later.
Also while the glitch effect was displayed the 'background' content stayed static.
Also, it may be hard to tell from the two images I have, but the white-ish squares are actually what looks like RGB (not grey-scale) random noise.

I also included the results from the Cinebench test after the reboot.

https://imgur.com/a/cgUWC5r

[Ace McPlane]

On 7/4/2018 at 11:55 PM, Hobgoblin101 said:

@stateofpsychosis I finally got an image, and some new info.

So I was about to lodge the warranty claim when I ran another test to be absolutely sure about everything, and that there was nothing I could do.
Then I noticed that Gigabyte's AORUS software stated that my GPU was clocked to 300MHz, and it never increased during workload. I thought that was odd, so I tried changing settings and it did nothing. Then I updated AORUS, restarted, still nothing. I then installed the 'optional' AMD Radeon update available, and after a reboot, I noticed, that when I opened things, my GPU clock went up from 300MHz to 304MHz then back down.

 

I excitedly ran a Cinebench test, thinking I had fixed it, and saw the GPU clock rate increase to the correct 1340MHz. Then I got the crash. The same one as before.
I was only able to get one crappy image because I didn't realize my flash was on so it took forever. But right after the photo got taken both monitors claimed to of lost signal.
After waiting it didn't come back, so I hard reset.

I then repeated the process.

Opened AORUS;
  GPU: 300MHz
 VRAM: 1200MHz
Opened Cinebench;
 It ran normally, getting perfect results...
 I am pretty sure that my GPU is probably in an unstable state now, and I expect to see more random crashes and will keep you posted if they happen.

 

It is also interesting to note, that the effect occurred on the first screen (cinebench) before it appeared on the second screen about half a second later.
Also while the glitch effect was displayed the 'background' content stayed static.
Also, it may be hard to tell from the two images I have, but the white-ish squares are actually what looks like RGB (not grey-scale) random noise.

I also included the results from the Cinebench test after the reboot.

https://imgur.com/a/cgUWC5r

Send those photos to Gigabyte and claim that warranty.

Pretty much any time I've seen those weird pixely boxes as artifacts, it has meant a dying GPU.