Curious what people think (Possibly dying 3080 Ti)

[BetteBalterZen]

Last night, about 15 minutes after I turning on my PC, I noticed around 10 small blocks appear on my screen, I was only browser the web while having a few programs open. These blocks reminded me of dying VRAM. There were only about 10-ish of them and very small (forgot to take a picture). 

By default, I have a undervolt profile loaded with Windows, around 850mV (can't remember exact, at work, can't check atm) on the core, at 1930MHz with a +500 on the memory. I set my 3080 Ti back to stock and the small blocks disappeared. I then tried to start FH5 as a test, because it is very GPU intensive, but FH5 crashed while loading. 

I then at this point obviously thought one of my VRAM chips was dead, since loading them up with a game causes said game to crash... 

I then went for a reboot with stock GPU settings, and after reboot there were no blocks on the screen. I tried starting FH5 and it started perfectly. After driving around ingame with no issues, I started enabling my MSI Afterburner profiles. First I enabled the undervolt which I mentioned in the start of the post and this worked perfectly. And then I enabled the profile with my maximum possible OC (stable OC), with +75 on the core which results in 2130MHz on the core, with +500 on the memory aswell. Card pulls 430 watt using this profile. No issues at all. 

But now I'm obviously thinking about this. Was this just a hardcore, weird graphics bug, or is my graphics card starting to die? Was it the start of the end I just saw happening?

I still have 2 years of warranty on the 3080 Ti. If this happens again, I will start RMA case ASAP, and yes I will remember to take a photo for the RMA case. 

But, any thoughts?

[DoctorNick]

23 minutes ago, BetteBalterZen said:

Last night, about 15 minutes after I turning on my PC, I noticed around 10 small blocks appear on my screen, I was only browser the web while having a few programs open. These blocks reminded me of dying VRAM. There were only about 10-ish of them and very small (forgot to take a picture). 

By default, I have a undervolt profile loaded with Windows, around 850mV (can't remember exact, at work, can't check atm) on the core, at 1930MHz with a +500 on the memory. I set my 3080 Ti back to stock and the small blocks disappeared. I then tried to start FH5 as a test, because it is very GPU intensive, but FH5 crashed while loading. 

I then at this point obviously thought one of my VRAM chips was dead, since loading them up with a game causes said game to crash... 

I then went for a reboot with stock GPU settings, and after reboot there were no blocks on the screen. I tried starting FH5 and it started perfectly. After driving around ingame with no issues, I started enabling my MSI Afterburner profiles. First I enabled the undervolt which I mentioned in the start of the post and this worked perfectly. And then I enabled the profile with my maximum possible OC (stable OC), with +75 on the core which results in 2130MHz on the core, with +500 on the memory aswell. Card pulls 430 watt using this profile. No issues at all. 

But now I'm obviously thinking about this. Was this just a hardcore, weird graphics bug, or is my graphics card starting to die? Was it the start of the end I just saw happening?

I still have 2 years of warranty on the 3080 Ti. If this happens again, I will start RMA case ASAP, and yes I will remember to take a photo for the RMA case. 

But, any thoughts?

My thoughts, card was never 100% stable and something triggered it. After you loaded the stock profile, the driver was still affected by this, until you rebooted the machine. I have experienced similar behavior with a handful of overclocked cards. I never bother to oc my GPU anymore because no matter how stable it gets, it dosen't seem to hold up when the profile is not flashed directly to the card. Back in the old days 600 series and earlier, you could just flash the OC profile and everything was fine.

No need to RMA. I would simply, if you still want to run the OC, just bump up the voltage slightly and call it a day. Most likely just memory not being stable at +500 at that voltage.

[BetteBalterZen]

1 minute ago, DoctorNick said:

My thoughts, card was never 100% stable and something triggered it. After you loaded the stock profile, the driver was still affected by this, until you rebooted the machine. I have experienced similar behavior with a handful of overclocked cards. I never bother to oc my GPU anymore because no matter how stable it gets, it dosen't seem to hold up when the profile is not flashed directly to the card. Back in the old days 600 series and earlier, you could just flash the OC profile and everything was fine.

No need to RMA. I would simply, if you still want to run the OC, just bump up the voltage slightly and call it a day. Most likely just memory not being stable at +500 at that voltage.

I love your respons and I seriously hope you are correct!

I will go home and bump up the voltage a tat more and see how things go. 

Thank you for the reply. 

[DoctorNick]

Just now, BetteBalterZen said:

I love your respons and I seriously hope you are correct!

I will go home and bump up the voltage a tat more and see how things go. 

Thank you for the reply. 

 Forgot you where in Denmark. Its hot atm, thats why. You generally need more voltage when ambient temps are higher to keep everything stable. Unless you see the green squares at stock settings don't worry, everything is fine.

[BetteBalterZen]

Just now, DoctorNick said:

 Forgot you where in Denmark. Its hot atm, thats why. You generally need more voltage when ambient temps are higher to keep everything stable. Unless you see the green squares at stock settings don't worry, everything is fine.

Roger that, and yup - it's hotter than normal here these days. 

Thanks again

[xg32]

if it's just on the oc, no need to rma, it's just...vram being vram.

 

I had a space invader card, underclocked the memory by 500 below stock and used it for years til the fan died. I don't minmax vram oc anymore, i just match the clocks to the oc'd models and it usually works.

 

Whether or not i should have RMA'd is another issue though

[BetteBalterZen]

12 minutes ago, xg32 said:

if it's just on the oc, no need to rma, it's just...vram being vram.

 

I had a space invader card, underclocked the memory by 500 below stock and used it for years til the fan died. I don't minmax vram oc anymore, i just match the clocks to the oc'd models and it usually works.

 

Whether or not i should have RMA'd is another issue though

Roger that, awesome to hear your opinion as well! 

What exact card did you have?

[xg32]

42 minutes ago, BetteBalterZen said:

Roger that, awesome to hear your opinion as well! 

What exact card did you have?

my 970, kept a mem oc, stable for awhile, then artifacts appeared even at stock, underclocked the ram and didnt have a problem for years til fans gave out

[BetteBalterZen]

1 hour ago, xg32 said:

my 970, kept a mem oc, stable for awhile, then artifacts appeared even at stock, underclocked the ram and didnt have a problem for years til fans gave out

Damn, so the OC actaully may have harmed the VRAM? Kinda crazy but also so cool how you kept it alive

[BetteBalterZen]

Update: 

Last night it happened again, right after booting up loading Windows and a couple of programs. And this was with no memory overclock applied. I did however still run with the undervolt (1930MHz@875mV). After a reboot, all went away again. Now I have increased voltage with 10/15mV, will see how this goes. Might have flown too close to the sun with 875mV. My PC actually also BSOD'd with a Video Scheduler Internal Error but I managed to get a photo of the squares I mentioned just before the BSOD. 

Not expecting anyone to reply to this, just simply updating my thread here. More information is good information. Some might be in a similar situation in the future. 

[Xkillerpn]

RMA that card. just dont wait. it will get worst 

[BetteBalterZen]

1 hour ago, Xkillerpn said:

RMA that card. just dont wait. it will get worst 

Yeah I have this feeling too... 

I will ask Komplett what they can do for me. I still have 2.5 years of warranty. 

[clutchmaster]

19 minutes ago, BetteBalterZen said:

Yeah I have this feeling too... 

I will ask Komplett what they can do for me. I still have 2.5 years of warranty. 

this could be do to excessive undershoot in voltage regulation have u capped the fps in nvidia control panel to ur monitors refresh rate

[clutchmaster]

21 minutes ago, BetteBalterZen said:

Yeah I have this feeling too... 

I will ask Komplett what they can do for me. I still have 2.5 years of warranty. 

and also is this happening after or before the msi AB profile loads?

[BetteBalterZen]

4 minutes ago, clutchmaster said:

this could be do to excessive undershoot in voltage regulation have u capped the fps in nvidia control panel to ur monitors refresh rate

Actually, I have always run with 120FPS cap because my main monitor is 120Hz but I just got a 240Hz and because of that I sat the cap to 240FPS. And it is after this it started happening. 

I am in talk with Komplett now and I will see what they can do. If they're able to switch it out with a 3090 Ti possibly because my GPU was bought back when prices were insane, I'd happily do the RMA lol 

[BetteBalterZen]

5 minutes ago, clutchmaster said:

and also is this happening after or before the msi AB profile loads?

After hehe

[clutchmaster]

8 minutes ago, clutchmaster said:

this could be do to excessive undershoot in voltage regulation have u capped the fps in nvidia control panel to ur monitors refresh rate

 

1 minute ago, BetteBalterZen said:

Actually, I have always run with 120FPS cap because my main monitor is 120Hz but I just got a 240Hz and because of that I sat the cap to 240FPS. And it is after this it started happening. 

I am in talk with Komplett now and I will see what they can do. If they're able to switch it out with a 3090 Ti possibly because my GPU was bought back when prices were insane, I'd happily do the RMA lol 

yep thats voltage sag but those artifacts should fix themselves in the next frame refresh if not then either theres a hardware issue or the driver glitched out or not enough voltage while under load
if it happens again while @0.950v then possibly a hardware fault is the issue

[clutchmaster]

cause at stock clocks that should be enough voltage 

[BetteBalterZen]

1 minute ago, clutchmaster said:

 

yep thats voltage sag but those artifacts should fix themselves in the next frame refresh if not then either theres a hardware issue or the driver glitched out or not enough voltage while under load
if it happens again while @0.950v then possibly a hardware fault is the issue

Thank you clutchmaster, this is good info. 

As said though, currently waiting for Kompletts response. If they are willing to upgrade me to a 3090 Ti, I'll happily do it hehe. 

[BetteBalterZen]

Right so Komplett did say they would replace my card if they were able to recreate the errors seen in my photos but they most likely will not see those because it is kinda a periodic error, or may be cause of the undervolt. They will not just sent a replacement card upfront, which make total sense of course. So we agreed that if things got worse and became a permanent thing, like, booting up Windows without MSI Afterburner and there's artifacts all over, then we will do a replacement upfront. 

@clutchmasterI hope what you said is what is causing it. Currently running with a tat more voltage than before. If artifacts comes back, I will revert back to default voltage and see how that goes. 

[BetteBalterZen]

Update: 

After increasing undervolt voltage with about 10mV, I have not since last comments experienced the same problem.

I most likely just did fly too close too the sun. 

Not expecting anyone to reply to this, just simply giving an update, in hopes of this may be a fix for someone in the future. I'm gonna give it a few weeks before marking @clutchmaster comment as the solution but it most likely is the solution. 

[BetteBalterZen]

@clutchmasterMarked your comment as solution. Ever since I bumped up voltage a bit, the artifacts have not shown again. Thank you.