Question GPU Artifacts Please HELPPPP

[ItsBash2022]

Hey, i have an XFX RX 570 4GB GDDR5 Graphics thats work fine on games/furmark until yesterday, from yesterday to now, when i start a game and playing for like 5 minutes, or running furmark for 30-40 seconds, gpu starts artifacting ( objects get into objects )

( this is happening on stock clock speeds of memory ( 1750 ) )

if i underclock memory to 1650 or 1700, the artifacts are gone but i know thats not a permanet fix
so is this memory problem rapairable ? ( there is some gpu repair centers in my city )

i tried to clean install the new drivers too, problem still exist

i think the problem is memory chips overheating, because this take like 5 minutes on gta 5 ultra settings to artifacts start and the backplate of gpu is hot ( before yesterday the backplate is cold on games )

gpu core temps are fine, after repasting the thermal paste with DeepCool Z5, temps are 30-35c on idle and 50-55c on load ( on furmark max temp is 71c )

please help me i dont have money to buy new gpu

sorry for my bad english

 

[A_Goat]

Have you monitored the temperatures under load? If not, do so. MSI afterburner is a free program that can accomplish this.

When was the last time you physically cleaned the GPU hardware itself? I.E removing the GPU and spraying the GPU with compressed air, as well as cleaning out the PCI-E Slot to remove dust particles

Does your GPU sag at all in the slot? Can you wiggle it up and down when you touch it? See picture.

Edit: I must add, it's not a good sign that underclocking the GPU returns stability, but the troubleshooting steps above may just extend it's life a bit more.

 

[ItsBash2022]

yes i monitored the temps, the gpu temp itself is under load stable at 55c on games, and 61c on furmark (now)

 

[A_Goat]

yes i monitored the temps, the gpu temp itself is under load stable at 55c on games, and 61c on furmark (now)

And the other points I mentioned?

 

[ItsBash2022]

last time is 3 days ago, i think problems starting from that date ( i replaced thermal paste with deepcool z5 )

NOTE: gpu is fine when i downclock the vram to -30mhz or higher
problem is with stock memory speed and higher

gpu position is fine, i tested it on another pc, same thing

is that repairable ?

 

[Karadjgne]

No. Not repairable. The vram has overheated, burned up a solder joint, degraded, or more likely the pads are shot etc. Those cards have multiple heat producing components but only the 'gpu' (the processor itself) has a temp reader. There's no temp listing for the VRM's or vram, so they can hit 100°C + and the gpu only report 55°C.

Of those options, only replacing the thermal pads is repairable, if that's the problem allowing the vram to overheat.

 

[ItsBash2022]

in my city we have some gpu technecial repair centers that replacing/reballing vrams, but the problem is im not sure this is a vram or gpu problem?
if problems are a burned solder joint, can it fix with reballing the memory chips ?

 

[Karadjgne]

Reballing is very hit or miss, there's no guarantee at all of any fix. And that's including knowing if the reball is necessary or if it's the silicon that's trashed. What it generally costs to do a reball is often more than the gpu is worth, especially for the low-end models. A 3080/ti/3090 would be worth it, but a Rx570?

Cheaper for you to replace than diagnose and/or fix. You gotta pay someone to test the validity of the chip first, diagnose the gpu to ascertain exactly what went wrong, could be as simple as an smd failed, but unless that's burned and obvious, someone has to go find it. And that could take hours of testing, at hourly shop rates.

That's why I said No. Not that it's not feasibly fixable, but the cost is generally far higher than the card is worth to fix and that's after paying for diagnosis fees.

 

[ItsBash2022]

i think it better to underclock vrams to -30mhz until the problems get bigger and gpu/vrams die, because reballing cost in my country is 15-30% of this gpu price in there
for example for a 150$ gpu, they reball it with 30$

and another question, if i downclock memory for example from 1750 to 1380mhz and dont let vram chips to get hot, gpu lifespan is going down or normally work ?

 

[ohio_buckeye]

It may work a while by downclocking but you may have to downclock more later. Downclock for now and save your pennies for a new gpu.

 

[A_Goat]

i think it better to underclock vrams to -30mhz until the problems get bigger and gpu/vrams die, because reballing cost in my country is 15-30% of this gpu price in there
for example for a 150$ gpu, they reball it with 30$

and another question, if i downclock memory for example from 1750 to 1380mhz and dont let vram chips to get hot, gpu lifespan is going down or normally work ?

It's going to continue to degrade no matter what you do to it, but yes downclocking will let you use it longer. If you really want to try and squeeze as much life as possible out of it, I'd give the card a refurbish. Good cleaning, new paste, and new pads. If it's still unstable after that, I'd undervolt the card entirely and slightly reduce both clock and memory speed to reduce heat output as much as possible.

 

[Karadjgne]

Lifespan is a total unknown. I still have a working 3dfx Voodoo3 3000 16Mb AGP x8 gpu on a Asus P2B 440 motherboard with a Genuine Pentium II 350MHz that's overclocked to 400MHz and has been that way for the last 22± years. There are others who have working Commodore Vic20 pc's that are 42 years old.

So whether your particular silicon Wil degrade further with underclocking is also unknown. The silicon might still be perfectly good, but 1 ball has come loose, so no degradation at all. You might be perfectly fine for as many years as you own the card, at an underclocked state. Or could be a bad batch of silicon, and one small part of the vram has died completely, but has no affect on the rest.

Only thing you can do is re-pad, re-paste, clean it out, check the fan curve vrs rpm, and see how long it goes. If it gets really bad, there is always the last ditch effort, the oven.

 

[ItsBash2022]

fan going to 4200rpm under load, and this is normal for my gpu

and my gpu is dont have any pads on vrams, do you think its necessary to add pads to vram's ? i think the xfx company dont put any on this cart

thanks

 

[Karadjgne]

It's not going to hurt to have vram actually have a way to dissipate heat, just be sure to get the right thickness, if it's too small it does nothing, if it's too thick it can stress the pcb.

 

[ItsBash2022]

you think if i do that ( buy thermal pads and use for vrams ) it cloud be fixed ? because it does not have thermal pads on vram's when i bought this card

 

[A_Goat]

you think if i do that ( buy thermal pads and use for vrams ) it cloud be fixed ? because it does not have thermal pads on vram's when i bought this card

Possibly, but impossible to know.
Unless you can afford to lose the card, like I mentioned earlier I'd just undervolt it and underclock it until it dies for good.