AMD GPU crashing

[DjTeorio]

The problem in short:
I got a mining GPU, it kept crashing in a weird way. I flashed a new BIOS and nothing happened.

The problem in more detail:
When I upgraded my PC, I got a used mining card, a Radeon RX 580 (4GB) to be exact, but it was crashing in a weird way whenever it was under a certain load. The crashes would consist of my monitor losing signal and the fans in my PC ramping up. What was interesting about the crashes was that whenever it happened I could still talk to people in my discord call and I had to force shut down my PC to be able to use it again. I recently swapped it with my old card and the crashes were gone and it was better in some cases than the newer mining card even though it only had 2GB VRAM, so I believed it could be due to the card having a mining BIOS installed, so I spent many hours yesterday trying to flash the BIOS and it kept giving me errors until today when I finally did it using the DOS version. Nothing had changed the crashing kept happening and I even reinstalled the drivers to no avail. So I have returned to this forum with the hopes that you could help me fix my scuffed mining card.

Any help would be greatly appreciated.

 

[ebivan]

Probably a hardware error. If that card ran in a mining rig for 2 years or more, its lifetime is reached. Capacitors age. And they age much faster with every degree of warmth they are subjected to. So two or more years at full load in a hot mining rig at 60+ °C is much more heat over time than any average card endures in a gaming rig in five years.

First you could check the caps for visible damage, sometimes they kind of expand and get thicker and inflate/get a bulge, sometimes the chemical from the inside bursts out and they have a brownish coating on them. But most of the times they just dry out on the inside and you dont see anything from the outside.
If you can see damage, you can solder in new caps, which only cost a few cents, otherwise you can just switch them all and hope it was the caps. But maybe it was a completely different failure and I was totally wrong
I am sure there are people out there that have better knowledge of what exactly happens with mining cards and can give you more specific advice.

 

[DjTeorio]

Well all I can really say is that when i opened it nothing stood out as being broken and my soldering skills are pretty much non-existent anyway. :/

 

[mtcn77]

If you can, resetting the driver has the odd chance of fixing it. It happened to me whenever temperature caused an instability with specific overclocks.

 

[DjTeorio]

I mean I have reinstalled the drivers multiple times if that's what you mean.

 

[Punkenjoy]

I had similar issue with my rx 580 8 GB. Ended up RMA the card for a new one as it was hardware.

 

[Zyll Goliat]

There is also a possibility that the miners change&adjust the card bios to minimize power consumption and gain much profit as they can and that causing the card instability.....If that is the case you can find the original card bios and flash it with Atiflash utility......GL

 

[ebivan]

There is also a possibility that the miners change&adjust the card bios to minimize power consumption and gain much profit as they can and that causing the card instability.....If that is the case you can find the original card bios and flash it with Atiflash utility......GL

He already did that

 

[mtcn77]

I mean I have reinstalled the drivers multiple times if that's what you mean.

If you had a shortcut to reset the driver while in os, it could have helped. For some reason, the card can soft reset gpu errors, but not memory controller errors which is the cause you should suspect, imo. It could also be due to memory although the card wouldn't let you off without a hard reset.

 

[DeathtoGnomes]

what kind of temps were you seeing? cuz it sounds like something was overheating.

 

[DjTeorio]

what kind of temps were you seeing? cuz it sounds like something was overheating.

Nothing too high, it would max out around 70°c, but it would normally crash somewhere in that range.

 

[Zach_01]

Any monitoring done for speeds and voltages during gaming? For GPU and VRAM.

You can use GPU-Z sensors tab, or HWiNFO "sensors only" window. HWiNFO can be more detailed (more sensors).
Just load a game for a while and get out before crash happens, and you can see max values of each.

 

[VuurVOS]

The crashes would consist of my monitor losing signal and the fans in my PC ramping up. What was interesting about the crashes was that whenever it happened I could still talk to people in my discord call and I had to force shut down my PC to be able to use it again.

I recognize these symptoms. The VRM's seems to be the issue. You can try to put the fan on 100% to see if it takes more time before it crashes.

 

[mtcn77]

I recognize these symptoms. The VRM's seems to be the issue. You can try to put the fan on 100% to see if it takes more time before it crashes.

It happened to me. Afterwards I got the shifted color dots. I don't know what broke, but there are polka dots which is funny for a computer error. The card is working and the graphics aren't altered, so I think it busted 1 module, or its bga solders.
I suggest you use the card with vsync and chill turned on. You're soon to have the same joke of a card, otherwise.

 

[Hachi_Roku256563]

what do you mean joke of a card 580 is fine at 1080p