RTX 2080 Ti FE Escapes Testing by Dying After 8 Hours @ [H]

[SilverSliver]

No wonder they wanted him to sign an NDA. They must have known what was coming with this release. Maybe not to this extent, but they had to have known.

 

[Joust]

No wonder they wanted him to sign an NDA. They must have known what was coming with this release. Maybe not to this extent, but they had to have known.

I have to believe this is true. The NDA might have been a blanket police change, but there's no way Nvidia didn't know they had QC/QA issues given the amount of issues we've seen.

Or.

They singled out all the [H] guys by stalking us all and have decided to try to end the Resistance by burning down our homes.

Hmmm.

 

[noko]

Wow! I am thinking also this is without the RT and Tensor cores being flexed or even used . . . What new issues may abound once those are put to use? If this is a heat issue then I cannot see them ever being used.

 

[DraginDime]

The pricing on 2000 series cards makes sense now. It's to cover all of the extra shipping charges for the multiple RMA's people are going to deal with.

 

[ablazhov]

Geforce RMA 2080 Ti - the way it was ment to be named.

 

[Geno750]

My bought today 2080TI XC Ultra lasted 2hrs of Diablo 3 gaming before black screen and dead card. At that rate, it was $625/hr to play D3 using a 2080ti.

 

[MrDeaf]

I am placing my bet on nVidia's core/mem boosting technology being overly aggressive with voltage vs. temp.
Either that, or there's a flaw or bug with nvidia's voltage regulation.

When I thought about it, 12nm FinFET works just fine on Ryzen+, even if it hits 90C under boost, because it's dumb and likes to dump 1.5V into Vcore.
Therefore, TSMC 12nm FinFET is proven to work under high voltages and heat, but with moderation.

of course, I am no engineer and this is pure speculation.

 

[sirmonkey1985]

My bought today 2080TI XC Ultra lasted 2hrs of Diablo 3 gaming before black screen and dead card. At that rate, it was $625/hr to play D3 using a 2080ti.

rip..

at this rate i wonder what law firms going to try and milk this fiasco for another class action lawsuit against nvidia..

 

[tunatime]

Could it be the solder? Odd the card worked then died after cooling off.
If it was some kind of heat issue you would think it would have died in the 8 hours of testing not after shurtung it down?

 

[ole-m]

I am placing my bet on nVidia's core/mem boosting technology being overly aggressive with voltage vs. temp.
Either that, or there's a flaw or bug with nvidia's voltage regulation.

When I thought about it, 12nm FinFET works just fine on Ryzen+, even if it hits 90C under boost, because it's dumb and likes to dump 1.5V into Vcore.
Therefore, TSMC 12nm FinFET is proven to work under high voltages and heat, but with moderation.

of course, I am no engineer and this is pure speculation.

I've never ever killed a card with mediocre stuff like clocks, they just crash and reboot.
Artifacts is usually because of bad memory meaning we should suspect GDDR6 is at fault from both samsung and micron however it works on RTX 2080\2070 with way better success meaning so we can exclude that.
we can then suspect the IMC in RTX2080 TI or QA not picking up defects in the IMC Or lastly the QA for the 2080 TI's being bad including memory module defects.

 

[Uvaman2]

A process shrink gone wrong?
This is something I worry with AMD going all in with 7nm... But might also be why the moved to chiplets?

 

[ltron]

This is why I refunded my faulty 2080Ti, I'm not going to take the risk on these cards. The fact that people are getting multiple faulty RMA units in a row, including Kyle, is practically unheard of. My refund came through yesterday and I was so happy.

 

[THRESHIN]

I think not signing that NDA was the best thing you guys ever did. I admit, I was unsure at first just due to the cost but wow. Funny how things work out sometimes.

 

[griff30]

This is about "3.5" times worse than the last time Nvidia screwed us.

Can't see how fanatics can defend this and the stock in Nvidia is dropping faster than cryptocurrency when SEC investigates

 

[nEo717]

I'm noticing around the web a new trend from both 2080 and 2080 Ti owners - many are requesting refunds, to which so far all of them I've seen have got their refund.

This surprises me that nVidia is so willingly giving refunds. Of course nVidia stock falling from $293 share down to $144 share in just a few days surprises me too... if it was just a Ray Tracing mix up (like it is), or just over heating, or just blue screens and games crashing, or just burning card (we really don't know if there's been more than 1 yet), or just so many owners posting failures, or just so many owners posting problems of long RMA waits and/or little to no contact - Any 1 of those happening would normally mean problems for nVidia, problem is, its not just 1 of these happening, its all of them at once.

My thoughts go to if these sorts of quality control issues are happening with nVidia's best gaming cards (most expensive), how can we believe lower, cheaper cards have better quality control (currently) and better parts? How can the quality control of any RTX cards be trusted at this point in time?

 

[c3k]

How many “good” cards are going to fail as they age? Dust, etc., reducing cooling/increasing heating: how long will warranties be honored?

 

[Uvaman2]

This is why I refunded my faulty 2080Ti, I'm not going to take the risk on these cards. The fact that people are getting multiple faulty RMA units in a row, including Kyle, is practically unheard of. My refund came through yesterday and I was so happy.

I agree, as mentioned before it has to mean the failure rate is gigantic, i don't see how it could be statistically supposedly so low, and have multiple people with multiple failure, it doesn't make a lick of sense...
Nvidia should start extending warranties I can't imagine not feeling the burn if 'good ' cards start failing soon after warranty.. a component as expensive and important as these should last at least 3 years, and I bet most people expect 5.

Also yes, a cursory check of a few NDA signers websites seem to only tow Nvidias bullshit statement and not much else as far as i know.

 

[Uvaman2]

How many “good” cards are going to fail as they age? Dust, etc., reducing cooling/increasing heating: how long will warranties be honored?

Exactly.. being the honorable top of the line privileged company they should announce warranty extensions.

 

[Uvaman2]

https://seekingalpha-com.cdn.amppro...223459-nvidias-prospects-look-much-worse-hood

At least this person makes some argument that nVidia is in for more pain.. might just be true..

 

[Bobert]

 

[Chris_B]

Wow, the heatsink in this pic looks like its about to melt of the damn thing. Cool Pic.

They missed a trick by giving it the old 2900xt paint job

 

[Chris_B]

Wow that sucks. I think the bin for bad 2080ti gpu is far higher than nvidia leads on.

There's no way they will say if the issue is worse than it is, they'll just keep saying "a small amount" and leave it at that.

 

[Domingo]

Whatever the issue is - my EVGA 2080Ti seems immune from it. It passes AIDA64's GPU torture test sans issues and I've had no problems whatsoever.
For what it's worth, it's using Micron RAM, too.
Maybe I won the 2080 lottery, but I couldn't be happier with it.

 

[Aireoth]

Whatever the issue is - my EVGA 2080Ti seems immune from it. It passes AIDA64's GPU torture test sans issues and I've had no problems whatsoever.
For what it's worth, it's using Micron RAM, too.
Maybe I won the 2080 lottery, but I couldn't be happier with it.

So has mine, but I am quite concerned over the longevity of this card.

 

[N4CR]

tempted to snag a vega 64 for $399 the other day on newegg..but i feel that's a side grade and won't help that much .. i'll wait for AMD's next card now

A 1060 to a V64/1080 or most of the way to Ti if you WC and OC it is a sidegrade...?

 

[homernoy]

Whatever the issue is - my EVGA 2080Ti seems immune from it. It passes AIDA64's GPU torture test sans issues and I've had no problems whatsoever.
For what it's worth, it's using Micron RAM, too.
Maybe I won the 2080 lottery, but I couldn't be happier with it.

I've got the 2080ti Windforce OC, and I've played hours and hours with my Vive. Also ran Furmark for a few hours, and benchmarked all of my newer games at 4k and all is well. I've got Micron GDDR as well.

This is all well and good, but I've still got that feeling in the back of my head that this thing could simply die on me at any moment. I've knocked on wood, and crossed my fingers on both hands!!

 

[Aireoth]

I've got the 2080ti Windforce OC, and I've played hours and hours with my Vive. Also ran Furmark for a few hours, and benchmarked all of my newer games at 4k and all is well. I've got Micron GDDR as well.

This is all well and good, but I've still got that feeling in the back of my head that this thing could simply die on me at any moment. I've knocked on wood, and crossed my fingers on both hands!!

Exactly, I don't think its unreasonable to say this card should have a three year warranty at the price they are charging.

 

[alxlwson]

I've got the 2080ti Windforce OC, and I've played hours and hours with my Vive. Also ran Furmark for a few hours, and benchmarked all of my newer games at 4k and all is well. I've got Micron GDDR as well.

This is all well and good, but I've still got that feeling in the back of my head that this thing could simply die on me at any moment. I've knocked on wood, and crossed my fingers on both hands!!

Gotta cross cocks for it to actually work. Just sayin'

 

[N4CR]

Wow! I am thinking also this is without the RT and Tensor cores being flexed or even used . . . What new issues may abound once those are put to use? If this is a heat issue then I cannot see them ever being used.

I'm very curious now what the quadro/workstation card failure rates are...

 

[alxlwson]

I'm very curious now what the quadro/workstation card failure rates are...

I wonder if they are qualified before they get boxed?

 

[Bobert]

Gotta cross cocks for it to actually work. Just sayin'

 

[blandead]

It's the gtx280 all over again

Maybe the fixed product will be rtx 2085 ti

 

[piscian18]

I hate being in this awkward position with a classic GTX1080. Its feels like the only card in the heirarchy that benefits from an upgrade but is still difficult to justify. if i was running a gtx1070 I could totally talk myself into an RTX and if I had a gtx1080 Ti itd be easy to say no. Im trying to get "Hard" but while that RTX is sexy Im asking myself "is that a sore or is it herpes? "

 

[Boscoh]

Thanks Kyle for what you do. I just set up my patreon sub. I don't know why it's taken me so long to do it.

Still chugging along with my RTX 2080ti FE. I hope it stays that way. I know I can count on you guys to report the reason for the failures when it's determined, and hopefully I'll know if I can expect a premature death from my RTX.

Lastly, thanks for not signing the NDA. This level of reporting is hard to find.

 

[Comixbooks]

That is some bad luck right there.....Nvidia doesn't know what the real problem is

 

[Chebsy]

Wow, I think I will stick with my 1070 card for the time being !!

 

[andrewaggb]

What's interesting to me is that if so many are failing so quickly, what sort of stress testing did nvidia do or not do before releasing them? It should be very interesting when the final cause is discovered.

As for the nvidia share price, I've been following it and I don't think this has much to do with it. All stocks dropped dramatically the last couple months and nvidia badly missed earnings projections. Nvidia's price grew so much the last few years because they've had high growth, helped by datacenter, amd's rehashes, a solid product lineup, and crypto mining. With the death of gpu crypto mining and high prices of the rtx combined with general stock market suckage have hurt them.

 

[R_Type]

Anybody else get a sneaking feeling that a respin is going to be needed?

With all these RMAs, would anyone be the wiser if Nvidia or the AiB's pried the GPUs off of the bad cards, plugged them into new cards, and sold them as new?

It would cost way more in labour than the cost of the chip being swapped.

someone mentioned in another thread and I'm sorry but the name/thread escapes me, that as these cards and densities get higher you need a specific range of transistors to not have traffic/use while others nearby are in use. That designers know this and should be designing cards to not have a specific range of neighboring transistors be active and any one point in time.

My 20+ years in the IT arena makes my gut say this is part of our problem. I think we are killing off transistors on these higher clocked cards because they are running in parallel physically and this is creating a condition where the transistors actually fail. Sometimes rather spectacularly.

That sounds like some pseudo science right there it really seems like something that just isn't/wouldn't/can't be remotely true. It would also mean that something like dram would have to be either dense or fast but never both as it would degrade at the drop of a hat.

I doubt its a design flaw in the GPU/IMC itself. I think the problem is heat. Look at the pick above that Kyle took. The middle area of the heatsink is approaching 90 C. There is no way any of the memory chips can handle this heat at longer stretches of gaming / benchmarking. It also explains why AIB cards even with a NVIDIA PCB last longer and the custom designed cards are fine.

But video cards being hot and wild is the norm. There's no data on 2080ti's in operation that shows anything that you would *really* call concerning. Not given how hot things get on ... Every >200w card ever.

hmm just thought of something.. nVIDIA MAY BE GOING THE WAY OF M S + WIN 10 .....GPU'S AS A SERVICE ?????
ie you need to keep replacing them ..smart move Ngreedia ??????????????????/ at $1200 bucks a card you get into more influx o cash for the end of the yr sales .. even if you screw up the drivers ..and cook cards ..
and RTX and so called ray ray traced games fail ..smooth move..lets jack prices ..

I like to huff paint fumes out of a plastic bag *after* I do my forum posts.

 

[SmokeRngs]

So has mine, but I am quite concerned over the longevity of this card.

I've got the 2080ti Windforce OC, and I've played hours and hours with my Vive. Also ran Furmark for a few hours, and benchmarked all of my newer games at 4k and all is well. I've got Micron GDDR as well.

This is all well and good, but I've still got that feeling in the back of my head that this thing could simply die on me at any moment. I've knocked on wood, and crossed my fingers on both hands!!

At this point I think these are the primary concerns. If I had one of these cards I'd be thinking the same thing every day. It's not about "what if" my card fails but "when" it fails.

It's likely even worse for many as I'm guessing quite a few of the people who jumped on these cards probably sold off their old card to help afford it which likely leaves them with no backup card if the new one goes tits up.

The longer nVidia takes to say what is actually going on the worse consumer confidence is going to be in the long run. While it appears the other cards in the lineup aren't suffering these issues I'm willing to bet the confidence in nVidia and these cards to not die is not high. This will harm sales no matter what but the longer it takes nVidia to admit there's an issue and what the real issue is the more it will damage confidence. It probably won't take much more damage before many people write off the whole 2xxx lineup which is going to be bad because nVidia is running out of 1xxx stock.

 

[Hagrid]

Whatever the issue is - my EVGA 2080Ti seems immune from it. It passes AIDA64's GPU torture test sans issues and I've had no problems whatsoever.
For what it's worth, it's using Micron RAM, too.
Maybe I won the 2080 lottery, but I couldn't be happier with it.

Yes, your card is immune. No need to worry.