[Digital Trends] Nvidia RTX 2080 Ti Graphics Cards Are Dying in Alarming Numbers

[Ottonomous]

The price of innovation and early adoption
https://www.digitaltrends.com/computing/nvidia-rtx-2080-ti-graphics-cards-dying/

 

[n0x1ous]

oh boy.....so far so good for me, but glad I decided to run them on air for a while before getting blocks.

Seems to be an issue with Micron GDDR6 from what I've read and seen today. Most people are having issues after only a few days of use so I guess I'll know soon enough

 

[moonbogg]

This could prove to be a fantastic opportunity for RTX haters to gloat and rub the failure right in Nvidia's face. Sucks for the consumers though. I feel for them. I would not be feeling very confident if I had a 2080ti card in my rig right now. It could be fun though. Think of it as a GPU Battle Royale or something; last card standing wins.

 

[Despoiler]

Definitely looks like bad VRAM is the culprit based on the artifacts. This won't blow over quickly if the memory is the issue. It's the worst as a consumer. They send you a new card and it's bad. Over and over again. This happened to me with a Gainward 4400Ti. 5 RMAs until I got a good card.

 

[4K_shmoorK]

Been pretty let down by the state of PC gaming lately. Too many console ports. New cards too expensive. Tired of playing games from 3-5 years ago.

I'm seeing pretty low benefit to even owning a 1080 Ti right now. Nickel and dime'd on hardware, skeleton crews porting over console games for software. Feel like I've got an expensive web browsing and DAW machine lately more than anything else

The fact that the going rate for a Ti one month ago was $1200 and already EVGA puts out a card for $999 tells me NV doesn't care much about people who bought and have continued to buy their hardware over the years. They know 20xx cards are overpriced and are comfortable milking their most committed customer base for more profitable enterprises (AI, auto, etc.)

Shouldn't it be the other way around?

 

[Fallen Kell]

The fact that the going rate for a Ti one month ago was $1200 and already EVGA puts out a card for $999 tells me NV doesn't care much about people who bought and have continued to buy their hardware over the years. They know 20xx cards are overpriced and are comfortable milking their most committed customer base for more profitable enterprises (AI, auto, etc.)

Shouldn't it be the other way around?

There will always be price gouging on a product in which there is scarcity. That is the way supply and demand works. $999 was the original MSRP, everything above that was price gouging on a scarce resource. EVGA obviously feels that the demand has dropped enough to price a card at MSRP and not feel like they should milk the cow for higher prices anymore. Be glad someone finally put out a MSRP card, as it will force others to do the same.

 

[Mopetar]

Are there any concrete numbers on this? Otherwise it seems like any new product launch where a lot of people buy a product and everyone is talking about it so an average number of failures suddenly seems like an outlier. Unless someone has measured forum posts about failures for previous launches, I don't know how useful this data actually is. Perhaps it's true, but right now there's no solid evidence.

 

[Kenmitch]

Guess it explains the lack of availability.

 

[Kenmitch]

Are there any concrete numbers on this? Otherwise it seems like any new product launch where a lot of people buy a product and everyone is talking about it so an average number of failures suddenly seems like an outlier. Unless someone has measured forum posts about failures for previous launches, I don't know how useful this data actually is. Perhaps it's true, but right now there's no solid evidence.

Really?

I'll start your research project with the following link.

https://forums.geforce.com/default/topic/1078162/geforce-rtx-20-series/rtx-2080ti-massively-die-/

I'm sure you'll find many more if your truly wanting to get to the bottom of it.

Poor soles didn't even get a chance to experience the new features yet.

 

[Mopetar]

Really?

I'll start your research project with the following link.

https://forums.geforce.com/default/topic/1078162/geforce-rtx-20-series/rtx-2080ti-massively-die-/

That's useless by itself. Unless you have a baseline to compare it to, it's just a useless statistic. If you go to an individual city in Nebraska and ask people there if they have cancer and come up with some number, can you tell me anything of real use? Of course not. There's nothing to compare it to.

Now, even assuming someone did have some similar comparison with a previous generation of cards, that's not necessarily useful either. If you go to one city in Nebraska and ask people if they have cancer and then I go to a city in Texas and ask people if they have cancer do you have anything useful. Almost, but only if you know the population since we want a rate, not a raw number.

It doesn't appear as though anyone has done the kind of substantive analysis that would allow anyone to conclude whether or not this is an actual problem or to what extent the problem actually exists. It reminds me of iPhone launches where there has to be some kind of -gate issue. Maybe there is an actual problem, but without the full data, it's mostly hearsay and trying to drum up a story. If it later turns out to be correct, anyone who reported it early will trumpet this fact, but if it doesn't pan out, it will quickly be forgotten in place of the next bit of potential sensationalism.

 

[GodisanAtheist]

These stories always tend to carry a wisp of schadenfreude, people like it when misfortune befalls people that splurge on luxuries, like the 2080ti. That's really what gives these things legs.

Regardless of bad press, I expect the price on the 2080Ti to steadily lower back to the "expected" launch price of $799-$650 amd the rest of the stack to adjust back as well. Otherwise there just isn't enough of a market to support the existing prices. Might as well extract what you can when you can.

 

[Ottonomous]

That's useless by itself. Unless you have a baseline to compare it to, it's just a useless statistic. If you go to an individual city in Nebraska and ask people there if they have cancer and come up with some number, can you tell me anything of real use? Of course not. There's nothing to compare it to.

Now, even assuming someone did have some similar comparison with a previous generation of cards, that's not necessarily useful either. If you go to one city in Nebraska and ask people if they have cancer and then I go to a city in Texas and ask people if they have cancer do you have anything useful. Almost, but only if you know the population since we want a rate, not a raw number.

It doesn't appear as though anyone has done the kind of substantive analysis that would allow anyone to conclude whether or not this is an actual problem or to what extent the problem actually exists. It reminds me of iPhone launches where there has to be some kind of -gate issue. Maybe there is an actual problem, but without the full data, it's mostly hearsay and trying to drum up a story. If it later turns out to be correct, anyone who reported it early will trumpet this fact, but if it doesn't pan out, it will quickly be forgotten in place of the next bit of potential sensationalism.

Nvidia denies any pervasive issues but we'll see whether its the standard failure rate within 1-2 months

In response to a request for comment, Nvidia told Digital Trends that it was, “working with users individually but we are not seeing any broader issues.”

 

[darkswordsman17]

Ouch that would add insult to injury after spending that much money. Might not be more than normal, but for that price, I'd be angry if my card died, so definitely would be vocal about it.

Nvidia denies any pervasive issues but we'll see whether its the standard failure rate within 1-2 months

That's pretty much standard response from any company, and Nvidia in particular has a history of trying to claim there aren't issues when there are. I remember one of the focus group members on here being very vocal in defending Nvidia when the...think it was dual GPU Fermi cards (590?) were dying because part of the card was getting so hot it was burning things off of it (people posted videos of like VRMs melting off), after some driver botched the power management or something. The ridiculousness over that is one of the reasons why I can't believe they let the focus group people back here. At one point the person was going "There's not a driver killing cards! But if you have one of these cards, make sure to use this updated driver to make sure it doesn't happen to your card." And then of course bumpgate (which wasn't entirely Nvidia's fault since it was due to changing to lead free solder in the industry at the time; its the same thing that hit the Xbox 360 because Microsoft cheaped out and didn't compensate for the change to the solder, same with Nvidia), which I think is a big reason why Apple has refused to work with Nvidia again as they simply refused to accept blame that it was their chips that was the problem (so Apple and other OEMs had to shoulder the blame until court cases worked through). I recall people putting their 8000 series cards in the oven to try and reflow the solder.

 

[amenx]

I remember the GTX 590 problem. It involved a premature driver release that had no over-volting protection which resulted in a couple dead cards at a review site. There were no mass failures since that required someone to over-volt the card for it to fail. The driver release missing the over-current protection was quickly withdrawn. The issue was massively overblown in forums as a widespread problem when it only involved one reviewer and 2 dead cards.

A bigger problem that occurred which nobody seems to remember involved 7900gt cards back in 2006. I know this because I owned one and was affected. Some batches of these cards had faulty vram and closely resembles the issue now affecting RTX cards. Nvidia never acknowledged the problem with the 7900gt, they just withdrew the card from their lineup and replaced it with the 7950gt after which the issues ceased.

With the 2080ti, seems some batches are affected while others are not, again very similar to the 7900gt problem.

 

[PrincessFrosty]

The price of innovation and early adoption
https://www.digitaltrends.com/computing/nvidia-rtx-2080-ti-graphics-cards-dying/

They quote "alarming numbers" but then don't quantify precisely how many, so how do we know it's alarming? All hardware has fault tolerances and some failure rate, so we'd need to know what the numbers are and what is common of previous generations before jumping to conclusions about what is and is not "alarming". I very much doubt it's as bad as it sounds.

 

[crisium]

One of you mining zealots should buy 50 RTX 2080 Ti's and run them non-stop and let us know the failure rate. For science.

 

[Hitman928]

https://www.youtube.com/watch?v=P18d6uLZ4qk

UFD Tech reporting that their 2080 Ti gives immediate BSOD when trying to install drivers on one of their motherboards. 1080 Ti works just fine in the system and the 2080 Ti works in another board.

https://www.youtube.com/watch?v=z75wDmwfIJY

Hardware Unboxed said that their 2080 Ti has been fine but their 2080 showed line artifacts and had to be returned, still waiting on new card. Their 2070 card would also only work in x4 PCIe mode. Both were AiB cards. They mention that AiB partners they talked to said they haven't seen an unexpected about of RMA cards at this point but that a system builder they talked to said that 3 out of 9 of their 2080 Ti cards they purchased for builds had already failed.

As others have said, we have no idea how statistically significant the issues are. Due to supply issues and obviously Nvidia not wanting to come clean unless they have to, we probably won't know for a little while either.

 

[EXCellR8]

Bummer... this is what warranties are for though. Can't say for sure how long the turnaround would be but the second my card showed signs of faulty memory that thing would be out the door at that price. I wonder what the actual percentage of launch cards are really cooked though...

 

[Ottonomous]

They quote "alarming numbers" but then don't quantify precisely how many, so how do we know it's alarming? All hardware has fault tolerances and some failure rate, so we'd need to know what the numbers are and what is common of previous generations before jumping to conclusions about what is and is not "alarming". I very much doubt it's as bad as it sounds.

Like Hitman pointed out, even reviewers have received defective units, which would skew that representative sample somewhat. Let's wait and see if the issue is reported by other users or receives further acknowledgement from nvidia

 

[Elfear]

Like Hitman pointed out, even reviewers have received defective units, which would skew that representative sample somewhat. Let's wait and see if the issue is reported by other users or receives further acknowledgement from nvidia

Quantifiable numbers would be the best but seeing so many failure posts on a low-volume card is troubling.

 

[VirtualLarry]

Hardware Unboxed said that their 2080 Ti has been fine but their 2080 showed line artifacts and had to be returned, still waiting on new card. Their 2070 card would also only work in x4 PCIe mode. Both were AiB cards. They mention that AiB partners they talked to said they haven't seen an unexpected about of RMA cards at this point but that a system builder they talked to said that 3 out of 9 of their 2080 Ti cards they purchased for builds had already failed.

You know, that's interesting. Were the lines horizontal green lines, when using HDMI output (to a 4K UHD screen)?

I recently purchased two EVGA GTX 1050 3GB cards, just the vanilla I think, no SC, no SSC. Anyways, I put them into a pair of Ryzen 3 1200 rigs, that were overclocked. I also updated the UEFI. Both of them, upon reboot, started displaying some horizontal green lines, but they went away after another reboot.

 

[RichUK]

https://www.reddit.com/r/nvidia/comments/9t1fwc/failing_rtx_2080_ti_source_discovered_its_the/

Possible GDDR6 issues which could be resolved by down clocking the memory.

 

[alcoholbob]

Kind of a bummer seeing as how 2080 Ti is memory bound compared to 2070/2080, lol. Have to gimp it even more to make sure it works.

 

[Omegaboost]

IMO this is caused by a faulty memory VRM in the FE PCB. No reported issues with AIB cards using a custom PCBs so the GDDR6 chips themselves are probably not to blame. FE PCB could be sending voltage spikes to the vram causing damage over time.

 

[sd5500]

I just almost pulled the trigger on the Asus ROG 2080 ti for my new build, but it would seem stupid to not wait a bit to see whats going on here. Plus I didn't like that it was a no refund policy with newegg even if it was unopened.