GTX 1080 Ti crashes for no apparent reason

[acoustic egret]

Hi guys, i am having severe problems with my GTX 1080 ti graphics card, the card is around 5 years old i got it 3 years ago but I already bought it used. The GPU was watercooled from the start and still is. temperatures never reaching over 55° Celsius.

But for a few weeks now the card has been crashing mid game for no apparent reason, also the PC dosen’t turn off with it, and I can still hear sounds and voice chat like discord for a few seconds.

Sometimes the PC restarts on its own with the GPU working again and other times I have to turn the PC off and on manually.

It mostly crashes during animation sequences in game, like entering wraiths portal in apex legends, dying on the hook in DBD, and similar things.

the Card dosen‘t crash during FurMark but during 3D mark runs

OC scanner from msi afterburner and asus GPU tweak both crash on about 80%

Similar Posts I found where solved by uncapping the Voltage and replacing the PSU.

Uncapping the Voltage dose nothing and the PSU in my PC is rated for 800 Watts while the GPU and CPU only pull 500 Watts at 100%

[slim perch]

Using overclocking tools and automatic overclocks

You probably want to uninstall all of them, and perform a driver reinstall with DDU, to rule out lingering OC profiles first

Then we can determine if the video card is legit dying

[acoustic egret]

Thank you for helping, I uninstalled named MSI and ASUS software and did an uninstall using DDU. Tried a 3Dmark run afterwards but still crashes

3DMark run after a Driver reinstall obviously

[slim perch]

OK. This time, install only MSI Afterburner. When it opens, instead of overclocking, try underclocking (use the sliders and push them as left as they can be. do not touch the voltags or the voltage/freq curve)

we're not undervolting, just using a negative offset so that a given voltage has less clocks

Run the same 3DMark: does it still crash? At least we've ruled out software causes

FurMark isn't a good GPU test as it uses too much power for a given clock, so most GPUs downclock automatically

[acoustic egret]

Yeah, i lowered the clocks as far as i could, and passed the Time spy test

do you have an idea why every software reports a different GPU memory clock? GPU-Z = 1200, HwMonitor 5003 and Asus GPU Tweak sad 11050ish

also GPU-Z reports a ✅ at the Ray Tracing

[slim perch]

Real clocks, pumped clocks, actual delivered MT/s respectively

[slim perch]

GTX 1060 6GB and up, as well as the GTX 1660, have a software DXR implementation. This is much slower than even the basic RT implementation found on AMD RDNA2 as there is no hardware assistance. Many games will not allow you to enable RT on these GTX GPUs for this reason.

It still exists for testing purposes, comparison purposes vs the RTX 2060/3050 (which the latter two RTX GPUs should always be faster), and in very light games

[slim perch]

Anyway, your 1080 Ti has become faulty (cannot maintain stock clocks). Time to start going shopping for a new card.

[acoustic egret]

I am so brock that i run 4*4gb on my 3 jeahr old 1920x 😭

But thanks for your help

[acoustic egret]

Why is that though, i mean the card never had temperature issues or something similar, and it was never been used for mining.

Is there rely nothing i can do?

[acoustic egret]

With the GPU Downclocked it sometimes freezes and sometimes crashes

[slim perch]

That model is at the right age since its production to start failing

[slim perch]

It still crashes? How frequently?

[acoustic egret]

I would say les often but at the same situations

[slim perch]

have you only reduced the clock offset or did you also reduce the VRAM speed?

[acoustic egret]

Bouth, -400 clock -500 memm

[slim perch]

Hmm, what CPU do you have?

Just in case

[acoustic egret]

a Thredripper 1920x Watercoold

*watercooled

[slim perch]

Not a CPU problem then

[acoustic egret]

i took the GPU apart to look for obvious flaws, reapllied the liquid mattel and changed the Thermalpads. i just run a 3DMark run on stock clocks and it worked. I think the problem was overheating of somethin that does not have a thermal probe. Will check this now

[acoustic egret]

It seams like that did not fix the problem, maybe it was a coincidence that the benchmark worked

[slim perch]

If it were a thermal issue the video card would simply shut down

[acoustic egret]

Yeah, but i could be a shutdown or? i mean the card shuts down but the PC continues to function without registering the "missing" GPU

[slim perch]

exactly. video card shut downs (disconnects) but the rest of the PC should still function in that case. if your computer had either an iGPU, or you can remote desktop it, the computer will be usable

if it were not a thermal shutdown you'd know

[acoustic egret]

yeah, that would be the reason for Discord continuing

[acoustic egret]

i could test this with my old 1050, i plug it in and wenn my 1080ti crashes i plug the monitor in the 1050 and check that

[acoustic egret]

But if the problem ould be thermal related would FurMark not be the one to crash it the fastest? i mean if the GPU pulls es much power as it can it would overheat faster? Exapt it is the VRAM because FurMark only uses like 600mb of it

[slim perch]

The problem being that your problem is likely not thermal related

You even monitored the temps in the first place

on the contrary, furmark is terrible for checking stability at normal gaming clocks

since furmark due to its nature will always throttle the gpu

[acoustic egret]

never throttled mine, clocks where pretty stable, the power consumption was jumping a bit but clocks where within 50-80mhz

[slim perch]

You should be taking a look at the actual achieved clock speed in FurMark vs anywhere else

not the fluctuations

[acoustic egret]

i run FurMark and my clocks are not fluctuating at all HW and Asus monitor report solid clocks

but my power limit seams to be 250w not 300 as it should be

and the GPU is currently on normal not downclocked

[slim perch]

do not refer to how the numbers change. please provide exact values.

[acoustic egret]

It seams not like a Overheating problem to me, the GPU clock hops from 1582 over 1907 to 1620 and back

the only thing that could be overheating now would be the VRAM but i looked at the thermal pads and they made clear contact.

[slim perch]

I already told you. Do not use FurMark or anything based on it (telltale sign: fuzzy models)

The clock speeds running these tools are meaningless for stability testing.

The fact that the GPU dropped from 1.9GHz briefly to 1.5GHz means that the tool isn't testing stability at normal game speeds.

[acoustic egret]

i did not ment it as stability test, i just wanted the GPU to pull as much Watts as it can to potentially overheat the core and/or the voltage converters with did not happened, so i think it is a problem wit the VRAM, dos the VRAM usage spike during animations? The GPU only crashed during these. I think i will try wo bake it to rule a bad solder joint out

[slim perch]

The fact that this thing maxes out the power and it stayed on already told you that the power delivery and thermals aren't problems

Core and VRAM instability can cause problems in everything, doesn't have to be anything specific

[acoustic egret]

I just ordert new liquid metal, gonna bake the GPU and pray that this will help. Thanks a lot so far for your help ❤️

[acoustic egret]

After Baking the Card the Problem continues, so yeah it seems like a not curable problem.

[slim perch]

baking only helps soldering issues

if the chip itself went, not much you can help

[acoustic egret]

yeah🥲 , i just hoped maybe there is a bad soder joint which cuts contact by heating up. so i will just run with the 1080ti till i can afford a new one in about 3 months.

[acoustic egret]

I noticed that there is this static noise coming from the GPU if there is a load and that the GPU seemingly crashes when this sound gets to loud, i know how this sound is called in German but not in English

[slim perch]

audible coil whine at high load

Hmm... can you try reducing the power limit slider?

[acoustic egret]

all as low as it gets

can i replace these?

which physical part does that?

Ohh i noticed that my BIOS reports just 10.5v on the 12v rail could that be the Problem? i mean the crashes. In the Internet it says that coil whine is just annoying but doesn't matter

[acoustic egret]

i might have fount the problem

[acoustic egret]

it seams like that fixed it, i just played for like 2h and had no crashes on normal clocks

[slim perch]

Your computer shouldn't even be starting up if that actually happened, so your motherboard, unfortunately, has an inaccurate voltage reading

But...

[slim perch]

Did you reinsert the cable?

[acoustic egret]

no, it is just a extension

threw that thing out and my GPU is stable on OC again

i would not say it is inaccurate, i mean it sat 10,5v and now 12,3v

[slim perch]

ah

avoid using unnecessary extensions, yeah