• Happy holidays, folks! Thanks to each and every one of you for being part of the Tom's Hardware community!
Apr 21, 2022
20
1
15
To start with, I want to state that before doing any of this, my GPU was perfectly fine, with no odd behavior whatsoever.

I've recently bought a used graphics card (yes, I know) as my previous one had died and I obviously needed a replacement since I couldn't live with a gt 730. Despite it being used for 2 years straight (the previous owner told me he never shut off his PC and only used the card for VR and light gaming), the card was perfectly fine when I thoroughly tested it and kept on being fine until yesterday. That was the moment I decided to fully clean the card including re-pasting it (I had cleaned it twice before, however only dusting it). I was cautious and thorough in the process and honestly thought I did a decent job.

When I booted up my PC, everything was fine, idle temps were great actually (around 25 C), stress testing was fine as well, no stuttering, max temps of 70 or so. However, when I finished doing such tests and wanted to watch a YouTube video, not only did the video stutter, the audio was glitchy too. And not only did that happen to the video, but it was also happening to the whole PC. Scrolling on the browser had stutters, wallpaper engine stuttered, and anything I was doing on the PC stuttered. However, when I got into a game so as to troubleshoot this issue further, the game ran perfectly fine. Not only that, the audio was completely normal too. And while having the game running, all the tasks mentioned before were perfectly fine.

As soon as I quit the game and the load on the GPU went away, the stutters began once more. I have no idea what's causing this, all I did was carefully clean and replace the dried-out thermal paste. Ever since then, this issue has been present. Basically, whenever there is a decent load on the GPU (as in the clock speeds being higher and it being used), it works perfectly fine, but the moment the load drops and it's idling or when watching youtube videos and doing light tasks, it has this nasty visual and audio stutter.

Here's my PC specs for info:
MB: B450M-DS3H
CPU: Ryzen5 2600
GPU: Sapphire Nitro+ RX 580 4gb
RAM: 4x4gb Corsair Vengeance LPX 3000mhz
SSD: Kingston A2000 256gb
HDD: WD Blue 1tb
PSU: Seasonic 80+ bronze 520W
 
Very strange behavior.

My thought is that when gaming stops something else starts running or otherwise grabs up system resources.

Not really able to explain any specific whats and whys. However you may be able to take a closer look at what is happening.

Use Task Manager and Resource Monitor to observe system performance. Use both tools but only one tool at a time.

Leave the tool window open: game as usual, then quit the game and watch videos and do light tasks. Watch for what, if anything, changes with respect to system resources.

What resources are being used, to what extent (%), and what is using any given resource.

Process Explorer (Microsoft, free) may prove helpful as well. You will probably need to download Process Explorer.

https://docs.microsoft.com/en-us/sysinternals/downloads/process-explorer

Another line of thought.....

With the problems occurring when the GPU is "cooler" there could be a loose connection or component involved. Warms up, expands, and makes solid contact when gaming.

Then, cools and contracts when not being heavily used. Connection becomes intermittent and the problems begin.

See if you can find some pattern with respect to the load and problem(s).

Just be methodical and change only one thing at a time.
 
Very strange behavior.

My thought is that when gaming stops something else starts running or otherwise grabs up system resources.

Not really able to explain any specific whats and whys. However you may be able to take a closer look at what is happening.

Use Task Manager and Resource Monitor to observe system performance. Use both tools but only one tool at a time.

Leave the tool window open: game as usual, then quit the game and watch videos and do light tasks. Watch for what, if anything, changes with respect to system resources.

What resources are being used, to what extent (%), and what is using any given resource.

Process Explorer (Microsoft, free) may prove helpful as well. You will probably need to download Process Explorer.

https://docs.microsoft.com/en-us/sysinternals/downloads/process-explorer

Another line of thought.....

With the problems occurring when the GPU is "cooler" there could be a loose connection or component involved. Warms up, expands, and makes solid contact when gaming.

Then, cools and contracts when not being heavily used. Connection becomes intermittent and the problems begin.

See if you can find some pattern with respect to the load and problem(s).

Just be methodical and change only one thing at a time.

Alright, so I've done some testing after your recommendations and here are the results:

- This scenario is me doing some light browsing/youtube video watching:
View: https://i.imgur.com/00eN8rH.png

- This scenario is me opening up a game and stressing the GPU a bit (100%) usage:
View: https://i.imgur.com/RoNcuYI.png


I've looked both before and after these scenarios and haven't noticed any sort of program or app that mysteriously appeared and started messing things up. Don't think that would be the problem though since just before pulling the GPU out for cleaning my system was perfectly fine, and nothing could have changed in that time since I noticed the issue right as I finished cleaning it and put it back in the PC.

I've done one last interesting test, which is to check and see whether this issue is directly correlated to clock speeds or to the temperatures (or both). I set up a custom fan curve and set the fans to the lowest possible setting, ran furmark till temps reached 72 C or so, ended the stress test, and instantly switched to chrome so that I couldn't give the GPU enough time to cool down. Even at 55-60 C without any load whatsoever (clocks were at 300 for GPU and VRAM), chrome still stuttered like hell and videos still had auditory stutters.

I can guarantee that this issue could not have been caused by temperatures, because I could run GTA V steadily with temps as low as 49 but frequencies around the 1200-1400 MHz mark without any stuttering. So I've narrowed it down to the frequencies themselves. Once they rise above 1000 or so, regardless of temperatures/wattage, everything works perfectly. If they drop lower in idle stages, the GPU freaks out.
 
Nicely done.

Regarding frequencies - interesting:

"Once they rise above 1000 or so, regardless of temperatures/wattage, everything works perfectly. If they drop lower in idle stages, the GPU freaks out. "

Reading back, I would not expect that your cleaning and re-pasting caused the problem. Especially your having done so twice before without problems.

Unfortunately I cannot exclude the possibility that some coincidental failure occurred especially with an older, used GPU. AKA Mr. Murphy or the proverbial gremlins....

But one immediate thought is to test the GPU in another known working computer. Determine if the problem follows the GPU.

Otherwise I have no answer to the frequency related behavior so will need to defer to someone who has experience with such things.
 
Update:
I haven't managed to do any further testing on my GPU, but I did notice a couple of weird behaviors since.

First, whenever the PC hasn't been on for a while, booting it up takes a couple times in order to work. What I mean by that is, when I press the power button, the PC starts, but there's no image. I then shut it off, power it on again and it starts but freezes. On the third boot, it finally goes into the OS. Something of the sort also happens when restarting the PC, but I'm guessing since it's restarting from a power state, it only takes two boots instead of 3 to reach the OS. Weird behavior.

On the other hand, I've tried a couple driver reinstalls and something weird happens every time. Right after the driver reinstalls, the GPU works fine, light tasks are no problem, even stress tests like furmark are ok. But, for example, after one furmark run and a couple minutes passing, the stutters reappear somehow. It's like after every driver install it takes a little while for the problem to kick in. So unusual.
 
I don't believe it's been said yet, but worth trying...reseat EVERY physical connection in the case... reseat all ram, drives, cards, etc as well as all power connections.

I reseat every single possible connection I could think of, and although the PC booted up just fine and everything showed up, the issue still persists. I was hopeful at first, since I haven't tried that, but sadly it did not work. I was skeptical too since when I first cleaned it, all I did was remove the card and the power connections, and when I was finished cleaning it I put it back the exact same way.
 
Long shot, but maybe the cooler is on a little tighter than before and causing the card to bow/bend? Then heating it up makes it straighter. So cold boot, fails to start, gets a little warmer and starts working.

Even if the temperatures look fine, that doesn't mean a single part of the GPU or a single memory chip isn't losing contact and overheating.
 
Long shot, but maybe the cooler is on a little tighter than before and causing the card to bow/bend? Then heating it up makes it straighter. So cold boot, fails to start, gets a little warmer and starts working.

Even if the temperatures look fine, that doesn't mean a single part of the GPU or a single memory chip isn't losing contact and overheating.

Might have been a long shot, but after several turns of slightly un-screwing the GPU mounting screws and slowly lifting pressure, IT FINALLY WORKS! There's no issue both at idle and under load. Your theory was completely right! Thanks!
 
  • Like
Reactions: dudio
I return yet again filled with despair to this forum. After loosening the mounting screws and finally fixing the issue, I was so happy that the GPU finally worked. It kept working for a good several hours. Then I shut off my PC and headed off to bed. However, I could not manage to sleep and booted up my PC again, only to find that the problem came back. I did NOTHING in the time it was off. Frantically, I opened the side panel and started unscrewing some more and check it again. That did nothing. Once again I took it out and unscrewed it some more. Nothing again. I tried unscrewing the backplate screws as some were connected to the heatsink. Nothing changed. At one point, I unscrewed it so much that it straight up started crashing, even making the OS not recognize the card itself, forcing me to reinstall the drivers. At that point I had to completely remove the heatsink and try to reseat it again. I think I removed and placed my GPU in the PC at least 20 times. Finally, I simply tightened everything up reasonably well and managed to bring the GPU to a bootable state, however the issue still remains. It pains me to return to this post as the issue seemed to finally be fixed, but here I am, desperate for an answer.
 
Could you clarify what you did differently? Like, what you used to clean the card?

I used some q-tips and a bit of perfume because I didn't have another source of alcohol, for the dust i just used pressurized air. I did drop a tiny bit of thermal paste (arctic mx-4) on the little chiplets around the die itself, and i did my best to clean that out but it shouldn't be thermally conductive i believe. There was a bit of dust and grime around the thermal pads, but they were mostly fine, one or two had little tears on the edges, but i simply pushed them back and they were intact. I reassembled the card exactly the way it was before, and tightened the screws. That's it.
 
Perfume? Au de toilet would have been better. Perfume has low amounts of alcohol and higher amounts of oils to maintain ether strength. Au de toilet has higher amounts of alcohol and less oils, is slightly stronger at first but fades fast and is cheaper to make. Both have oils and other crap which don't do well with trying to get paste to stick where it needs to, leaves deposits.

Don't use perfume. Period. Isopropyl or nothing.

I'd pull the gpu apart, again, and make sure you clean the gpu off correctly. Use isopropyl (rubbing) alcohol, preferably higher than 70%. Get it at any drug or dollar store for @ $1-$2 a bottle.

Mx-4 is good gpu paste. But with gpus, they require a perfect application. Don't worry about using too much, just make sure the entire surface of the silicon is covered, leave no gaps or corners bare.
 
  • Like
Reactions: Phaaze88
Perfume? Au de toilet would have been better. Perfume has low amounts of alcohol and higher amounts of oils to maintain ether strength. Au de toilet has higher amounts of alcohol and less oils, is slightly stronger at first but fades fast and is cheaper to make. Both have oils and other crap which don't do well with trying to get paste to stick where it needs to, leaves deposits.

Don't use perfume. Period. Isopropyl or nothing.

I'd pull the gpu apart, again, and make sure you clean the gpu off correctly. Use isopropyl (rubbing) alcohol, preferably higher than 70%. Get it at any drug or dollar store for @ $1-$2 a bottle.

Mx-4 is good gpu paste. But with gpus, they require a perfect application. Don't worry about using too much, just make sure the entire surface of the silicon is covered, leave no gaps or corners bare.

Alright, I ordered 1 liter of 99.9% isopropyl alcohol, I'll make sure to properly clean and re-paste the GPU again once it arrives. Will provide photos and updates when it gets here.