I recently upgraded my R9 280x to RX 590 and I started having overheating issues ( I guess ? ) I tried to check my PS and CPU as well just to be sure but it all started with this new GPU. As a side note a friend of mine upgraded to 5600XT and is having the same problem. If I play GPU heavy games, my screen turns black, fans start spinning at maximum speed for couple of seconds and I can hear distorted audio loop so I have to manually reset it with a button. It doesn't happen right away as I need to be gaming for couple of hours but some games ( like Subnautica, ARK ) can trigger it almost instantly. I tried monitoring my GPU fan speed and temperature but it never exceeded 85 ℃ (185 ℉). The highest value I saw was 84 ℃ while my fans were running at 87% speed. I was able to extend my gaming period without crash by disabling ZERO RPM mode to get constant air flow but the problem still persists.
First off IMHO they pushed Polaris too far and many users complain about this especially with the 590. In addition to what I am going to tell you, some users have to lower the clock speed to get stability which negates having bought a 590 vs a 580.
You can return that card if it is still new or rma it if you don't end up with results you are happy with.
Now set a manual fan curve in Wattman. Have it kick the fans on to 100% at about 45-50 degrees. This will help the throttling which is half your issue. Then set the power limit slider to +50 and this raised the power ceiling which is really set too low by defaults and causes the black screens.
If those alone don't help then also lower the clock speed, if that then helps you have to decide if you are good or you want to pursue getting another card.
Thanks, I posted my problem on many forums and this is the most useful answer so far! I bought the card few weeks ago so I can technically return it. I downloaded newest version of adrenaline software ( I had to do that manually because it said it is up to date even tho it was May release and I downloaded July one ) I'm gonna test what you suggested and post results.
If you still have issues with the newest driver, don't write that off as what I told you doesn't work. Many are reporting issues with the 2020 drivers on Polaris.
Run DDU then return to a late 2019 driver and then apply what I said if the latest does not work for you.
Also if you are using the auto upgrade, IMHO it doesn't work right, at least never for me. You are better to get the latest download from the download page on AMD's site. Then install it yourself.
So, I tried to download last 2019 release of adrenaline driver and it totally bricked my computer. I had to reinstall my operating system :X ( I was lucky that I had a backup, so no data lost I just needed to install all my games again ) I tried different version not the latest one of 2019 and it worked. So far my temperature is lower by 2-3 ℃ and I was able to play GPU heavy game like Subnautica for 2 hours straight before it crashed. This is basically my new record. Even tho my temperature never exceeded 80℃ it still crashed...
Also I read something about undervolting it from 1150 to 1100 and manually change frequencies in stage 1-7 to the 1500MHz ( while 1560Mhz is currently the highest )
Honestly once you have to start also decreasing speed to get stable then you have a bad card. I do know that many resort to this. All I can say is that if I could return the card at that point I would. If that is not an option and I could RMA the card I would. Only you though can decide what you are willing to tolerate. To me these cards should all work perfectly fine at default settings. The fact that moving away from default settings helps even if it does not alleviate the issue proves they are defective.
I do wish you the best of luck.
Thank you for your helpful comments, I really appreciate your help. I just checked and my GPU is only 1 month old so I'm can definitely make use of customer rights and turn in back, but the fact that so many people even a friend of mine with 5700XT is having this issue is really worrying me. I'll try to fiddle with it a bit more maybe take it apart and re-apply thermal paste.
FYI don't check the thermal paste with most OEMs that VOIDS the warranty. By all means if you are past the warranty period that is a good thing to try, but you are not. You might also call your retailer and ask for a manager, explain you have been having issues since day one and have been working with tech support and maybe they would let you exchange it, instead of having to go through RMA or maybe even give you store credit. Worth asking, all they can do is say yes or no.
also if you do try to RMA it don't mention any of the changes you have attempted with changing anything other than defaults. That too, they will tell you voids your warranty. Just tell them it hasn't worked and you have tried different drivers and eliminated other components being issue.
Gonna RMA it tommorow. I was testing gpu with OCCT GPU memory test and it crashed my driver multiple times with black screen and then finally just digital noise on screen and then I tried 3D Gpu test and I got black screen almost instantly. So that might be the issue
I prefer Furmark which I like to call the fuzzy donut
20.8.1 is the latest driver
i recommend windows 10 v 2004 x64