How can I prevent artifacting on a RX580?

[Kaicer]

I'm stable while running the furry donut while overclocked to 1575Mhz on core but, it immediately becomes unstable and begins to artifact at 70C-72C and crashes soon after the temp reaches 76C. How can I fix this so I can clock to 1600Mhz+?

[LukeSavenije]

upped power and voltage?

[Levent]

Adding to what @LukeSavenijesaid, install a better cooler, you need to cool VRMs and VRAMs if you are thinking about pushing any further.

[Kaicer]

Just now, LukeSavenije said:

upped power and voltage?

Yes, I did. In fact, I had to in order to clock that high. I can't remember what my voltage was though.

[Zando_]

"I'm stable"

 

"it immediately becomes unstable"

 

These two statements disagree, your OC isn't stable at all. They can often seem stable but crash after 5 minutes, 20 minutes, or even an hour or two. You defo wanna max the power limit and push the voltage as far as you can without going over the safe max or getting temps too high though, if you don't your card doesn't have any more juice to push a higher clockspeed. 

 

[Kaicer]

Just now, Levent said:

Adding to what @LukeSavenijesaid, install a better cooler, you need to cool VRMs and VRAMs if you are thinking about pushing any further.

I'm currently using the stock cooler with the gpu

[Zando_]

Just now, Kaicer said:

I'm currently using the stock cooler with the gpu

What model? And what are you using to OC, wattman, afterburner, something else? 

[Rolling Potatoe]

Improve cooling or upp voltage and power.

TBH its not wort trying to get the last few MHZ when you have to deal with random crashes or artifacts. 

I usually even back down a little from what seems stable so I dont randomly run into issues.

[LukeSavenije]

2 minutes ago, Levent said:

Adding to what @LukeSavenijesaid, install a better cooler, you need to cool VRMs and VRAMs if you are thinking about pushing any further.

depends what's on it rn, if it's a armor, i agree

[jah118]

there has been a thread on this,  but they never overclock it that fare...
 

so maybe go with.

8 minutes ago, LukeSavenije said:

upped power and voltage?

[Kaicer]

1 minute ago, Zando Bob said:

What model? And what are you using to OC, wattman, afterburner, something else? 

XFX Rx580 GTS Black Edition. I'm currently using Wattman.

[Princess Luna]

5 minutes ago, Kaicer said:

'm stable while running the

 

6 minutes ago, Kaicer said:

it immediately becomes unstable and begins to artifact at 70C-72C and crashes soon after the temp reaches 76C.

So you're not stable at all.

 

6 minutes ago, Kaicer said:

furry donut

This is the worse and most useless software there is, it's workload is completely unrealistically and extremely more demanding than anything you'll ever use your GPU truly for, so it's pointless to test stability with it as it'll probably demand you to actually underclock your GPU to run stable.

 

Older cards like Maxwell and RX 400 series would actual break running FurMark for too long, so many lost their cards to this useless junk both nVidia and AMD had to develop a safer throttle mechanic specially for it to prevent permanent damage.

[Kaicer]

3 minutes ago, Princess Cadence said:

This is the worse and most useless software there is, it's workload is completely unrealistically and extremely more demanding than anything you'll ever use your GPU truly for, so it's pointless to test stability with it as it'll probably demand you to actually underclock your GPU to run stable.

What application would you recommend then?

 

[Princess Luna]

2 minutes ago, Kaicer said:

What application would you recommend then?

 

Heaven Benchmark or Superposition.

[Zando_]

1 minute ago, Kaicer said:

What application would you recommend then?

 

Unigine valley on a loop at 1080p or 1440p. That's what I used to OC my RX480, Vega FE, and for my RVII. Run that with afterburner open and tweak settings in realtime, as soon as you see a hint of artifacting, smack the reset button in afterburner, then put in the last stable settings or push more voltage/power target, hit apply, and see how it goes. If it artifacts for too long you get a blackscreen, sometimes it recovers, other times you have to restart the whole PC. 

[Kaicer]

9 minutes ago, jah118 said:

there has been a thread on this,  but they never overclock it that fare...
 

so maybe go with.

Wait If I'm stable at 1520-1530 I won the silicone lottery with my GPU or am I just being stupid?

[LukeSavenije]

1 minute ago, Princess Cadence said:

Heaven Benchmark or Superposition.

or valley, but it's a little older

[Kaicer]

2 minutes ago, Zando Bob said:

If it artifacts for too long you get a blackscreen, sometimes it recovers, other times you have to restart the whole PC. 

That's exactly what I've been experiencing.

[CharminUltraStrong]

So you've got a decent cooler, it's not an Armor card at least. Perhaps you ought to change the thermal pads on the VRMs, alter voltage and remove the backplate. I know on certain cards, the backplate is an after thought and negatively effects performance(EVGA learned this the hard way). The furry donut is a pretty awful benchmark to run on that GPU no offense. That'd be like running Prime 95 at full tilt without knowing what you're doing. When I want to tune my card I run the Heaven benchmark and up voltage/clockspeed in small increments until I reach the highest possible OC with good stability/temps.

[DJ46]

An OC becoming unstable when the temperatures rise doesn't mean the overclock is completely unstable, useless or whatever else everyone keeps saying.

Overclocking isn't just voltage and frequency. Otherwise liquid cooling and extreme cooling would have no impact.

 

I agree that it probably isn't worth it just to squeeze the last few MHz the card has to give, but you could upgrade the cooling to an aftermarket heatsink or some kind of waterblock (a universal one in a custom loop, an AIO with a bracket etc.), if you just enjoy tinkering with your rig.

I do that mainly to achieve low temperatures and low noise, but even I have noticed while playing around with my GPU that a previously stable overclock became unstable once I put the stock heatsink back on for a bit.

[Kaicer]

3 minutes ago, CharminUltraStrong said:

Perhaps you ought to change the thermal pads on the VRMs, alter voltage and remove the backplate. I know on certain cards, the backplate is an after thought and negatively effects performance(EVGA learned this the hard way).

3 minutes ago, DJ46 said:

I agree that it probably isn't worth it just to squeeze the last few MHz the card has to give, but you could upgrade the cooling to an aftermarket heatsink or some kind of waterblock (a universal one in a custom loop, an AIO with a bracket etc.), if you just enjoy tinkering with your rig.

I'll try doing all of the above and see what results I manage to get.

 

[fluxdeity]

XFX are lower tier cards compared to other brands such as ASUS Strix. If you're able to get 1575mhz then good for you but 98% of the people with a 580 can't hit that. They seem to top out at 1500-1550 and even then there's no guarantee. My 580 could only hit 1460 stable. Would crash at anything above that no matter what I did.

 

Just go for 1400-1450mhz and 1090-1100mv. That's an overclock and undervolt, which will help with the longevity of the card so you're not always pushing 1.20-1.3v through it.