[SOLVED] Is my 1080Ti fixable? Crashes/artifacts but runs better underclocked

[TheRandomGuy7]

Hi, so I picked up an Asus Strix 1080Ti sold as faulty. The card hadn't been opened before. I thoroughly cleaned it (very dusty) and replaced the thermal paste (which was in poor condition). But this didn't fix the issues whatsoever.

Fans work perfectly, it boots into Windows perfectly, it does everything fine until put under load. Then it artifacts (colourful mostly blue star shapes) and crashes the program, or occasionally the whole PC.

I've played around extensively with Afterburner, and it is far more stable with the maximum -500Mhz memory clock and -400Mhz core clock. This allows 10-15 mins of use before a crash (but no artifacts).

So my best guess here is faulty VRAM. Is there any combination of undervolting/overvolting/underclocking that could fix this other than I've already tried? Any ideas on if this can be fixed?

EDIT: forgot to add that temps are absolutely optimal. Also the thermal pads on the VRAM and VRMs are in good condition.

 

[Phaaze88]

Usually faulty Vram or an unstable, or no longer stable, Vram overclock.
The artifacting is something that just happens to some cards.
Downclocking the Vram frequency is a workaround, not an actual fix, but there's probably nothing else you could do after that...

I'm sure it's not overheating; GDDR5X doesn't run particularly hot like R6 and R6X, and can be easily cooled with fans only, even without thermal pads and heatsinks.

How does it crash during the times it's not artifacting?

I trust you have a good psu paired with this card?

 

[TheRandomGuy7]

Usually faulty Vram or an unstable, or no longer stable, Vram overclock.
The artifacting is something that just happens to some cards.
Downclocking the Vram frequency is a workaround, not an actual fix, but there's probably nothing else you could do after that...

I'm sure it's not overheating; GDDR5X doesn't run particularly hot like R6 and R6X, and can be easily cooled with fans only, even without thermal pads and heatsinks.

How does it crash during the times it's not artifacting?

I trust you have a good psu paired with this card?

Yeah I'm almost certain it's faulty VRAM, the only strange part is that the core clock also needs to be reduced along with memory clock to get more stability?

When underclocked it will run a game/benchmark for 10 mins or so. Then application suddenly freezes and closes itself. No artifacting.

On Valley benchmark I also sometimes get a Windows message 'Application blocked from accessing graphics hardware'.

PSU is solid of course, EVGA G3 750W (overkill). I've tested dozens of GPUs in this rig so it's definitely the 1080Ti.

If I had to guess, VRAM chips have been overworked/overheated for too long by previous owner and are now failing. Large chunks of dust came out the card so the VRAM could've been running red hot without him realising (no sensors). If the VRAM has failed there is no real way to fix this card then?

 

[Phaaze88]

Applications freezing and closing themselves is usually drivers or a bad overclock - I've done the latter; triggered d3d.exe crashes, for example.

"Application blocked from accessing graphics hardware", sounds like drivers. This wasn't a problem with the previous gpu, was it?
Was the previous card AMD? Did you run Display Driver Uninstaller in Safe Mode to remove the leftover AMD drivers?

If I had to guess, VRAM chips have been overworked/overheated for too long by previous owner and are now failing. Large chunks of dust came out the card so the VRAM could've been running red hot without him realising (no sensors). If the VRAM has failed there is no real way to fix this card then?

Now that was just them being lazy...
GDDR5X doesn't run particularly hot, unlike R6 and R6X. I've got a Kraken G12 kit on mine, with no thermal pads and heatsinks on the Vram and it's been just fine with the air from the single fan. But if it were caked with dust, well...

There's no real way to fix the card that I know of.

 

[TheRandomGuy7]

Applications freezing and closing themselves is usually drivers or a bad overclock - I've done the latter; triggered d3d.exe crashes, for example.

"Application blocked from accessing graphics hardware", sounds like drivers. This wasn't a problem with the previous gpu, was it?
Was the previous card AMD? Did you run Display Driver Uninstaller in Safe Mode to remove the leftover AMD drivers?

The software error only occurs rarely, usually just a crash or causing the entire PC to hang unresponsive forcing a hard reset. Yeah drivers properly installed and DDU used etc, even tried reflashing the BIOS and using the non-OC Strix bios with lower clocks.

I think we both know this card is doomed and I'd wager money at this point it's definitely dying VRAM. It's a shame the card wasn't treated well enough, waste of a perfectly good 1080Ti chip which itself seems to be working fine. Probably looking at a textbook example of VRAM failure here and if I were to keep using the card I'm sure it would eventually die completely.

 

[Phaaze88]

Well, I was trying to be positive...

 

[Karadjgne]

When you repasted, in what condition were the thermal pads on the VRM's/vram. If they overheat, you'll get artifacts etc.

 

[TheRandomGuy7]

Well, I was trying to be positive...

Me too but at this point I think I'm just wasting my time

When you repasted, in what condition were the thermal pads on the VRM's/vram. If they overheat, you'll get artifacts etc.

They were a bit grubby but I cleaned the dirt/dust off the surface with cotton buds/q-tips and a little isopropyl. I have Thermal Grizzly Minus pads ready to use but the pads on this card were definitely not in a bad enough condition to warrant changing them out.

In fact speaking of these.. you wouldn't think a bit of dirt on the pads would impact the VRAM so much they'd cause this critical failure and artifacting? I'd imagine it might make them slightly less effective but surely considering the pads are intact and in reasonable condition they would still be cooling the VRAM properly

 

[Karadjgne]

Hard to say. It only takes 1 VRM to go south and the card is next to useless under loads. The VRM's are well capable of dumping out heat in excess of 90°C under a decent load, so adequate cooling is somewhat critical.

VRM's are not cpus though, they are solid chunks of silicon on top and apart from passage to the ground plane heat is going to go out the top. I can't see a tiny chunk of dirt being an issue unless it's realistically not a tiny chunk of dirt, but something large enough to prevent good pad contact between the VRM and heatsink.