RX 580 crashes at 75c

[iBSOD]

Hi there,

I've had this ex-miner 580 for like 2 years now. The default fan curve on this is very aggressive and pushes temps down to mid 60s at the expense of a lot of fan noise. This never bothered me until a few weeks ago when I got open back headphones (HD600!)

Tried undervolting the card, but any attempt at changing voltage, core clock or memory clock causes crashes. I set a more conservative fan profile but the card crashes very quickly once it hits 75c.

Since this is an ex-miner I tried flashing the stock bios, but it may have been the wrong one as it will no longer boot. Luckily the gpu has a vbios switch!

The card I have is a Powercolor red devil 580 1380mhz. This is the vbios I tried to flash https://www.techpowerup.com/vgabios/199230/powercolor-rx580-8192-171123

I also experienced some artifacting recently, although it only happened once when the temps were around 75c. I was able to screenshot the visual mess meaning its the vram?

Any way I can let this card run over 75c or should I start saving for a new card?

 

[Meddling-Monk]

That's the default normal OC BIOS, and it's been set if I remember correctly to always have that default fan going on max nearly all the time.

You likely want the "Silent" BIOS option instead, as that one had the fans power down, go at a slower rate and also had a lower power profile setting that enabled it to undervolt more easily (unsure if the Drivers can override these settings). The normal OC BIOS I remember trying to undervolt mine, and it was REALLY finicky and decided to crash very easily instead. So I stuck to the Silent BIOS, which limits it to 1355 Core instead of 1380 (as you can see, the difference is really small in performance) and undervolted, etc from there. You might want to try the same with yours.

For extra reference, as I'm after silence, and I game and do stuff without any kind of headphones, and use 2.0 speakers instead, that are set to really low volumes, any fans are easily audible here. So I've got the (same) RX580 running at a REALLY aggressive fan curve, where its temps just about peak at 86C with all fans on silent (25% case fans, 45% GPU fans, but fans on GPU were modded by removing stock and replaced with Noctua A12x25s). So you should be safe if the temps being reported there are accurate and letting the GPU head a bit more northwards. The only danger is if the reporting of the temps is actually wrong, but even then it should in theory detect the higher than normal temps and throttle anyway (90C is where it throttles automatically).

As for the messy screen, it does sound like the VRAM has taken a hit. You could try and reduce the VRAM clock and up the voltage a tiny bit and see if that helps.

Also, one final thing, if it's post-mining stock, you might want to make sure that the cooling is functioninig properly too, unscrew the 4 to 6 screws on the backplate and gently remove the plastic shroud with fans from the card. Have some paste ready to repaste if you've not don't this previously. Clean up any old paste on the GPU core, apply new paste to it. And examine the pads on the VRAM and power delivery system on the left of the card, near the display connectors. If any of them are worn out or entirely gone, you should grab some replacements and then add them back in to see if that'll fix the problem with them; hopefully it's just normal overheating issues and that fixes them. This should be one of the cheaper options to you short of going for a new card.

 

[iBSOD]

With the card set to 'silent', it is anything but silent with fans ramping up to 3000rpm within a minute of gaming. Also it only runs around 1150mhz on this mode unless I set the power limit to max... Which fixed the fans! I have no idea why this worked but now furmark tops ar 80c with fans at 43% 1355mhz.

Currently trying out the undervolting!

Thank you monk

edit

So while undervolt stress testing the card crashed, nothing unusual, but upon restarting the fans are now ramping up really high again and not letting the card run higher than 75c, regardless of the power limit. and now it will only boost up to 1310mhz. Can't get my head around this card, feel like throwing it at a wall

 

[Meddling-Monk]

That's great to hear.

Yeah, the BIOS were a little weird really. I've had the fans do funny stuff on the silent one too. But got it under control now, as do you.

 

[Meddling-Monk]

Have you tried reflashing the Silent BIOS from a stock one like you did for the default OC BIOS? Just incase the Silent BIOS was messed up?

In drivers, push Power to +50. And try to use manual control of the fans, try setting what you want there. Then slowly go through all diagnosis for the other issues to find out what's gone off.

 

[Vince]

Have you tried reflashing the Silent BIOS from a stock one like you did for the default OC BIOS? Just incase the Silent BIOS was messed up?

In drivers, push Power to +50. And try to use manual control of the fans, try setting what you want there. Then slowly go through all diagnosis for the other issues to find out what's gone off.

I believe that only one of the Bios are flashable on RX cards with a bios switch. Haven't come across a card where you can flash both yet. You certainly can't on my vega cards and nitro cards.

 

[Meddling-Monk]

I believe that only one of the Bios are flashable on RX cards with a bios switch. Haven't come across a card where you can flash both yet.

Not sure as I've not had to on mine before, but I have noted that there is the Silent BIOS (appears to be stock) on the same location that was linked in OP. So wondered if that was an option or not.

 

[Vince]

Not sure as I've not had to on mine before, but I have noted that there is the Silent BIOS (appears to be stock) on the same location that was linked in OP. So wondered if that was an option or not.

You can pull the bios off easily, just one of the bios chips is read only Also if the card was flashed for mining it would be super likely that the memory timings would have been edited, as soon as you edit those it will throw the drivers off causing them not to work unless you run a patch after driver install. Assuming you don't patch the drivers you can also assume then that the bios on the card is correct or at least correct enough that the hardware ID's marry up.

First port of call on this for me would be stripping the cooler and re-pasting the card. Of course make sure that the memory has all its heat pads etc and that they aren't all dried out and doing nothing. Random question to the OP, are the fans on the card the stock fans? What you are experiencing I have also experienced when I replaced some fans on an RX570, the fans i put on there couldn't reach the RPM of the stock fans and that caused that card to go haywire. It's also the reason I tend to sell my cards with a new replacement fan.

 

[Meddling-Monk]

Nuts. Didn't know about one of them being Read only. Or was it read only from certain types of editing only?

But if that's the case, can OP try writing the Silent BIOS on the OC BIOS? See if that helps as a new base to work from? The Silent BIOS should never have been anything less than 1350 or 1355Mhz core anyway, so something was amiss there as well from their testing.

 

[Vince]

Nuts. Didn't know about one of them being Read only. Or was it read only from certain types of editing only?

But if that's the case, can OP try writing the Silent BIOS on the OC BIOS? See if that helps as a new base to work from? The Silent BIOS should never have been anything less than 1350 or 1355Mhz core anyway, so something was amiss there as well from their testing.

You can do that for sure! you can even pull the silent bios off of the card in GPU-z. My understanding is one of the bios cannot be written to at all for any kind of write operations.

 

[Meddling-Monk]

Then I'd recommend OP @iBSOD try and grab the stock Silent 1350/1355Mhz core BIOS (which also limits power) and try and flash it onto the OC BIOS.

But, also, as suggested, remove the Shroud and look at the VRAM, see which versions are in use, because if it's a Micron VRAM version, then the Silent BIOS version there might not do the job, as it lists the other variants (Samsung and Hynix), but not Micron. Otherwise, give that a try and see if that's stable. If so, try the undervolt and fan control from there (as it should be defaults then). Right now, we're not entirely sure where it's deciding to give you problems exactly.