GTX 1080 ti crashes unless underclocked

[kamakashi]

GPU crashes randomly when playing games and running benchmarks.

Only way this card is stable and never crashes is if I set the:

Core clock -150mhz
Mem clock -150mhz

I hear the MOSFET whine when GPU is running at a higher fps but I read that's normal.

I've been able to play games for 5 months with no stability issues with msi afterburner underclock settings.

BIos is original GTX 1080ti gaming 11g and isn't flushed to some other bios.

i bought this card on eBay 6 months ago and noticed it started to have issues 1 month later at which I underclocked the card and now it's fine. It also doesn't overheat.

I use the the GPU in 5 other builds and same issue. Needs to be underclocked to be stable.

 

[eidairaman1]

Provide detailed system specs, that includes your power supply, no generic info

 

[kamakashi]

It does this with many builds. I have to install MSI afterburner and underclock by -150 and it works fine.

Current build Alienware r12
i5-11400f
Dell Aurora motherboard
Adata XPG SPECTRIX D41 16GB DDR4 3600MHZ
1000w dell gold power supply


Other builds
Am4 3600x
Seasonic 1000w platinum power supply
Adata XPG SPECTRIX D41 16GB DDR4 3600MHZ
Asus ROG strix gaming motherboard

Intel i7 4790k
MSI gaming motherboard
Gigabyte p750gm gold power supply
16gb ddr3 1600
700w OCZ power supply

Alienware r10
Am4 5600x
dell aurora gaming motherboard
Dell gold 1000w power supply
Adata XPG SPECTRIX D41 16GB DDR4 3600MHZ

 

[GerKNG]

just my own experience:
i have/had so far three Pascal Cards that degraded (without any overclocking) to a point where they weren't stable at stock clocks anymore (two of them were 1080 Tis One Strix OC, one Gaming OC White)

 

[kamakashi]

There other builds I could list but it's the same issue with all the builds. I use this GPU as my main daily and can't sell it due to instability unless underclocked by -150. It will crash if set core and mem to -100, -50. I could try other numbers but -150 has been stable.

It's just odd that with a -150 core and mem and heavy gaming it runs just fine with no issues.

 

[eidairaman1]

The card is broken, Time to upgrade

 

[P4-630]

It's a GPU from 2017, these things can happen and imo it's a sign to replace the GPU for a new one soon.

GL.

 

[eidairaman1]

It's a GPU from 2017, these things can happen and imo it's a sign to replace the GPU for a new one soon.

G

It's a GPU from 2017, these things can happen and imo it's a sign to replace the GPU for a new one soon.

GL.

Planned obsolescence

 

[kamakashi]

Well, looks like this is more of a solution for some people rather than a problem. I'll keep using it. Surprised it works flawlessly with a slight underclock.

 

[Marcus L]

Planned obsolescence

Planned obsolescence is 100%
GPU's like CPU's these days, for the top end anyway, are really pushed to the limits of the silicon and there is little room for overclocking, not only that, it increases their power limits to the max and the efficiency drops off a cliff for the highest end parts, of course they are going to degrade quicker.

TL;DR the GPU is on it's way out, it could very well last for another couple of years with the underclock/volt though how long is anyone's guess, could be a week or 5 years but probably time to retire it to a retro rig or box it at this point.

 

[izy]

Does it crash if you only underclock the vram or only the core clock ?
It could be only the VRM but who knows.

 

[kamakashi]

It will crash with vram set at -150 and core clock 0 or vice versa. Both amhave to be set -150.

I'm happy with the card working for a good solid 5 months and hopefully more. I only paid $70 for it.

 

[pavle]

That's alright. You could also try just lowering TDP to 75% and leave the clocks as they are; the card will adjust them lower.

 

[eidairaman1]

Not working as intended though

 

[Dr. Dro]

The GTX 1080 Ti is an elderly GPU at this point. They've been in service for over 6 years. If your card was previously used for mining - and given you bought it off eBay 6 months ago I'd say that it's a given, it is very likely that it has degraded to a point where it is no longer stable at stock. Even the cards that haven't been mistreated are approaching the age where they will begin to fail if they saw consistent load and little maintenance over the past half decade plus.

Two things to do: run it undervolted, or write off and replace. It cannot be fixed.

 

[docnorth]

It seems you know the problem and the solution (for now). If I may suggest, keep it as long as it works, 1080 ti still performs acceptable. Just be prepared, like other members already said, for a replacement.

 

[maxfly]

Have you pulled the heatsink and checked the tim and thermal pads? If not, you may as well check them out and see what it looks like under the hood.

 

[kamakashi]

Have you pulled the heatsink and checked the tim and thermal pads? If not, you may as well check them out and see what it looks like under the hood.

Yes, thermal pads are fine and cleaned/reapplied thermal paste. Like everyone said it's just tired but still usable.

Yes, thermal pads are fine and cleaned/reapplied thermal paste. Like everyone said it's just tired but still usable.

No overheating issues.

 

[Lew Zealand]

This sounds like exactly the same behavior I see in my aging 2015 Intel NUC5i7. It worked at the rated speed, gaming at 35W (over the 28W nominal power limit) for years and when I transitioned to newer computers I kept it on as a video server. Start doing some QuickSync encoding to get more vids on the server and I see it crashing out after a while. Run more tests like Fire Strike which it easily passed in previous years and crash out every time. By now if you push it more than 20W or about 75°C, it'll crash out. It's clearly degrading like your 1080 Ti.

Don't expect your -150 to last for too long but on the other hand, $70 for a 1080 Ti that only needs a -150 underclock seems like a good deal even if you only get a year or so of use out of it.

Edit: almost forgot! I repasted it and zero difference. I was kinda hoping but not realistically expecting that to help. Bonus was taking it all apart for the first time so that was cool.

 

[Deleted member 182555]

could be faulty caps...
you need an esr meter to check them.

 

[Dr. Dro]

could be faulty caps...
you need an esr meter to check them.

It's very unlikely it's any caps. This symptom is consistent with spent GPU ASICs. I've got a GTX 780 that does the exact same thing.

Heat and voltage over the years will damage the silicon and it will lose its capacity to operate at higher speeds due to electromigration.

This phenomenon will become more pronounced with smaller lithography chips as time goes on.

 

[Shrek]

Planned obsolescence

Not that I necessarily disagree, but I am still running a Core 2 CPU with a GT 1030 GPU and it still seems quite usable. In over 2 years Windows 10 support stops, and then I will have to finally replace the machine.

 

[Marcus L]

Not that I necessarily disagree, but I am still running a Core 2 CPU with a GT 1030 GPU and it still seems quite usable.

But higher end silicon is usually pushed to it's thermal and clock limits with modern GPU's and CPU's with little/no room for much overclocking and of course more voltage to run at those clocks which will have an effect over time, I'm sure if you OC the snot out of your GT 1030 it would start to go that way also

 

[Macro Device]

Summing it up, your GPU went into age-caused dementia. Either sell it for whatever money and get a newer (<4 y.o.) card or just stay underclocked till it dies or pisses you off completely.

I had a thought about your video card VRM possibly playing tricks but it's not so likely. In case you can fix it for funny money (<$30) just do it and see if the problem persists. 95 percent it will persist and you got a dying die.

150 MHz is in overall not a biggie, just 8 percent loss. Your 1080 Ti is still of plenty room for gaming. I'd start saving money for a new GPU if I was you but I wouldn't rush it. Most likely your card will go to -300/-300 in 2024 or later and not sooner. I had a GTX 670 which had the same issue and it's still working (downclocked to hilarious 650 MHz [stock is about 1 GHz]) despite initially running in stability trouble 5 years ago.

 

[Dr. Dro]

Not that I necessarily disagree, but I am still running a Core 2 CPU with a GT 1030 GPU and it still seems quite usable. In over 2 years Windows 10 support stops, and then I will have to finally replace the machine.

Eh, 11 will run fine on a Core 2. Just grab a E8600 or a Q/QX9650. It's quite usable, as long as you have a SSD and don't mean to run a lot of heavy games.