Wednesday, November 15th 2017

NVIDIA Confirms Issues Affecting Early Production Run of GeForce RTX 2080 Ti Graphics Cards

NVIDIA, via a blog post on its forums, has confirmed widespread reports of failures affecting their flagship RTX 2080 Ti graphics card. The issues, which resulted in "crashes, black screens, blue screen of death issues, artifacts and cards that fail to work entirely," started cropping up throughout tech forums, before reaching a critical mass that warranted coverage - just in case this was exactly what it seemed, ie, a production issue.

It seems this was just so, and that the problem was luckily limited to some early manufacturing issues or QA controls. As NVIDIA themselves put it, "Limited test escapes from early boards caused the issues some customers have experienced with RTX 2080 Ti Founders Edition." The company then says that they stand ready to help customers who are experiencing problems - but nothing else was to be expected, really.
Source: NVIDIA
Add your own comment

60 Comments on NVIDIA Confirms Issues Affecting Early Production Run of GeForce RTX 2080 Ti Graphics Cards

#1
TheLostSwede
Well, I hope this is it, at least for the people that have spent a small fortune on these cards and then got duds.
Posted on Reply
#2
cucker tarlson
The new,fixed version of the card are shipped in black boxes and are based on Samsung's GDDR6. Old were grey-ish with Micron's memory.

[IMG]https://scontent-frt3-2.xx.fbcdn.net/v/t1.0-9/45658224_2114032368662184_6146846942224711680_n.jpg?_nc_cat=110&_nc_ht=scontent-frt3-2.xx&oh=62812450941ec4af0181a011487b7a51&oe=5C6D1B2B[/IMG]
[IMG]https://scontent-frt3-2.xx.fbcdn.net/v/t1.0-9/45642433_2114032585328829_4645990815573737472_o.jpg?_nc_cat=109&_nc_ht=scontent-frt3-2.xx&oh=c703fcee31c5ab6a0dc8e22cdbfb1779&oe=5C705149[/IMG]


TheLostSwede said:
Well, I hope this is it, at least for the people that have spent a small fortune on these cards and then got duds.
Nice to see people actually caring about other users rather than talking about how it was "justice" cause they can't afford a high end GPU themselves.
Posted on Reply
#3
mouacyk
"Inspired by gamers. Built by NVidia." (But not for gamers, only crazy rich kids.)

Is the box color confirmed? Source for this? And good luck telling Amazon or Newegg to ship you the black box one instead of the grey one.
Posted on Reply
#4
cucker tarlson
mouacyk said:
"Inspired by gamers. Built by NVidia." (But not for gamers, only crazy rich kids.)

Is the box color confirmed? Source for this? And good luck telling Amazon or Newegg to ship you the black box one instead of the grey one.
my private contact who runs an enthusiast tech channel.
Posted on Reply
#5
bug
Glad they got to the bottom of it. Because if you can't find the cause, you can't be sure you've addressed it.
Still major bummer for those affected, but at least we should be safe from now on.
Posted on Reply
#6
eidairaman1
The Exiled Airman
They forget to mention fires
Posted on Reply
#7
cucker tarlson
eidairaman1 said:
They forget to mention fires
it's good we've got you on guard.
Posted on Reply
#8
lexluthermiester
cucker tarlson said:
The new,fixed version of the card are shipped in black boxes and are based on Samsung's GDDR6. Old were grey-ish with Micron's memory.
Maybe I missed something, but where does it say that Micron's GDDR6 was the heart of the issues? Thinking it was something else.
Posted on Reply
#9
Vayra86
lexluthermiester said:
Maybe I missed something, but where does it say that Micro's GDDR6 was the heart of the issues? Thinking it was something else.
I don't think it was said, but rather alluded to here. We've seen problems with Micron memory before, on Pascal Nvidia issued a BIOS update for it.
Posted on Reply
#10
efikkan
cucker tarlson said:
The new,fixed version of the card are shipped in black boxes and are based on Samsung's GDDR6. Old were grey-ish with Micron's memory.
I believe there is no evidence so far supporting the claim that Micron memory is the culprit of anything. And to make one thing clear; having multiple sources for memory is common practise, and have nothing to do with the various claimed problems with RTX 2080 Ti.

I do like the investigation that Gamers Nexus is conducting. According to them, most/all BSOD issues have been resolved by driver updates.
The video shows artifacts on at least four cards, e.g. 31:24. By the looks of it, this certainly look like a memory corruption issue. Whether defective cards like these are part of a systemic problem comes down to the volume of faulty cards. RMA rates of 1-2% within a few months and up to 4-5% throughout the warranty period is normal for graphics cards, so far the claimed RMA rates for RTX 2080 Ti is <0.1%, only time will tell.
Posted on Reply
#11
lexluthermiester
Vayra86 said:
I don't think it was said, but rather alluded to here. We've seen problems with Micron memory before, on Pascal Nvidia issued a BIOS update for it.
Did we ever get conformation on that? It's unfair to lay blame without evidence.
efikkan said:
By the looks of it, this certainly look like a memory corruption issue.
That also happened with the soldering issues NVidia GPU's had last decade. This set of problems might be similar to that issue.
Posted on Reply
#12
efikkan
lexluthermiester said:

That also happened with the soldering issues NVidia GPU's had last decade. This set of problems might be similar to that issue.
I've seen this kind of thing of various cards over the years. It's normal that every product have a certain failure rate, and it certainly do vary from product to product. The failure rate of RTX 2080 Ti is so far below average, but time will tell if this increases.

For anyone having artifacts while operating within specs the answer is clear; obviously RMA it at once.
Posted on Reply
#13
John Naylor
How'd we get from "not a broad issue"' to "widespread" ? From what I have seen on the site, 1 batch number is particularly relevant to the problem and another one to a lesser extent. And nVidia, to all evidence, is addressing the issue to customer's satisfaction. If ya wanna be on the bleeding edge and buy 1st stepping products, this is a hazard associated with being an early adopter.

As to the fires... yeah that pops up every now and then ... but again avoidable.... stay away from reference cards and stay away from early steppings. EVGA 970 SC (1/3 heat sink missed card, no backplate), EVGA 1060 - 1080 (no thermal pads), MSI tape adhesive over fans that was so string some folks broke the fans taking it off. We always recommend waiting 3 months after release ... a) Supply has caught up with demand and so lower costs, b) ya have all the data from reviews to make the best decision, c) you avoid the "birthing pains" which sometimes occur with a new product d) production line improvements change your % likelyhood or getting a better OC in the silcon lottery and e0 all the driver and game coding issues are ironed out.
Posted on Reply
#14
SIGSEGV
efikkan said:
The failure rate of RTX 2080 Ti is so far below average, but time will tell if this increases.
how do you know? are you soothsayer or something? :confused:
or perhaps Nvidia rep? :eek:
Posted on Reply
#15
EarthDog
cucker tarlson said:
The new,fixed version of the card are shipped in black boxes and are based on Samsung's GDDR6. Old were grey-ish with Micron's memory.
Can you link where memory was the issue or is this simply a supply thing on the Micron side? We knew from the get go that GDDR6 will find wide adaptation from both AMD and NVIDIA so there were going to be multiple options eventually.

Just trying to paint the right picture. :)


Edit: maybe I should have kept reading... lol. Regardless of a correlation (past unrelated issues mentioned) where did anyone say that as a fact?

Remember these sammy ics are already supposed to be in the RTX Quadros...
Posted on Reply
#16
cucker tarlson
Can you please read and comprehend ? I said a lot of micron cards are being replaced with samsung. No one officially stated that it's the reason,neither did I. I guy tha runs extreme hardware channel had 4 2080ti's on micron, all dead, replaced by four on micron, two dead, replaced by samsung, all fine and dandy. That's all I'm saying. If you can't read the signs and need an official confirmation then you can disqualify it.
RMA rates you are quoting are for AIB cards,while the problem was always with 2080ti FE mainly. Is there a problem with all micron's gddr6 ? - no. But does that mean there can't be a problem with a large percentage of micron's gddr6 on FE cards, caused either by the chips themselves or nvidia's error ?
Posted on Reply
#17
CheapMeat
Paying $$$ for the privilege of bug & defect testing.
Posted on Reply
#18
bug
CheapMeat said:
Paying $$$ for the privilege of bug & defect testing.
There's always a risk with first batches of anything really. But hey, it still sucks big time if there's less than 1% change of getting a lemon and you do. This being an over $1k product just make things even "better".
Posted on Reply
#19
EarthDog
cucker tarlson said:
Can you please read and comprehend ? I said a lot of micron cards are being replaced with samsung. No one officially stated that it's the reason,neither did I. I guy tha runs extreme hardware channel had 4 2080ti's on micron, all dead, replaced by four on micron, two dead, replaced by samsung, all fine and dandy. That's all I'm saying. If you can't read the signs and need an official confirmation then you can disqualify it.
RMA rates you are quoting are for AIB cards,while the problem was always with 2080ti FE mainly. Is there a problem with all micron's gddr6 ? - no. But does that mean there can't be a problem with a large percentage of micron's gddr6 on FE cards, caused either by the chips themselves or nvidia's error ?
Just asking as the words used, to me, allude to that being the problem in particular with the context of the thread. I can see the writing on the wall, but nothing has been confirmed yet that was the issue. It certainly felt that was what you were trying to say, if I was mistaking, I was.... but there is no need to throw out insults. WTF is your problem??????

As I said, between NV and AMD, GDDR6 is going to be popular and made by more than one fab. It doesn't surprise me in the least that there is Samsung on them.
Posted on Reply
#20
cucker tarlson
What are you insulted by ? me telling you to understand what I'm saying correctly ? if that was your trigger,then you are too easily triggered. I'm sorry anyway.

Guy from extremehardware disassembled one of his micron-based 2080ti FE's that got back from RMA,says they were definitely tinkering around 3 memory chips

[IMG]https://scontent-frt3-1.xx.fbcdn.net/v/t1.0-9/46469266_2123356797729741_3281477944697946112_o.jpg?_nc_cat=104&_nc_eui2=AeHd7k2bcFBicOyFhUuj1DvXJn_5E-4dECI-KtOMHZ8NMY-HoB3wAPzyXDXdPvJ_NKbUf8Wsfs979F9RmN94kNwKeyq4TUBGCMGFM9rnwD-m4A&_nc_ht=scontent-frt3-1.xx&oh=a16f597a444a561d6ce0402ad6983a4e&oe=5C6E9A24[/IMG]

[IMG]https://scontent-frt3-1.xx.fbcdn.net/v/t1.15752-9/46330667_542485269557809_2189414077976018944_n.jpg?_nc_cat=109&_nc_ht=scontent-frt3-1.xx&oh=d16e5104d8d88469e30a088ad44e2a5d&oe=5C7B517C[/IMG]
Posted on Reply
#21
efikkan
SIGSEGV said:
how do you know? are you soothsayer or something? :confused:
or perhaps Nvidia rep? :eek:
Steve from Gamers Nexus claims to have talked with several AIBs, he has mentioned it in the videos about this problem, probably also the one I linked.
Posted on Reply
#22
lexluthermiester
efikkan said:
Steve from Gamers Nexus claims
While I'm not willing to call Steve a lair, one has to question the why NVidia themselves has not given details themselves. It's not like they have much of anything to lose.
Posted on Reply
#23
EarthDog
cucker tarlson said:
What are you insulted by ? me telling you to understand what I'm saying correctly ? if that was your trigger,then you are too easily triggered. I'm sorry anyway.

Guy from extremehardware disassembled one of his micron-based 2080ti FE's that got back from RMA,says they were definitely tinkering around 3 memory chips
Telling someone to read and comprehend what you are saying is pretty rude is all. Certainly there are better ways to get your point across than that if someone misunderstands something, yes (I also wasn't the only one who took it that way...was it really me??)? But hey, its TPU and is par for the course here... I should know better. Same shit, different day!


As far as the rest... Correlation /= Causation and all.... so nothing concrete from either side. :)

If this really is a memory problem, I wonder why the card partners cards are seemingly not affected? They use the same memory and all at the same speed (the couple I have reviewed). I wonder what would be the deal ONLY on the FE cards as opposed to everyone else... or why one(ish) batch was affected as opposed to the others?
Posted on Reply
#24
efikkan
EarthDog said:

If this really is a memory problem, I wonder why the card partners cards are seemingly not affected? They use the same memory and all at the same speed (the couple I have reviewed). I wonder what would be the deal ONLY on the FE cards as opposed to everyone else... or why one(ish) batch was affected as opposed to the others?
Just for context, the "RTX 2080 Ti problem" consists of at least two distinct types of driver problems causing BSODs, and at least once distinct problem manifesting itself as artifacts. The two known BSOD problems are related to 1) multi-monitor and 2) certain G-Sync displays, which should be reproducible on all Turing cards and supposedly resolved in newer drivers, but these are not hardware issues.

Then, we have so far one distinct hardware defect, which you can see demonstrated here. Steve demonstrated this on several cards, including ones from EVGA. The problem manifests itself as a distinct type of memory corruption, but I want to stress that this is the symptom and not necessarily the cause. There are many possible causes for this, including defective memory, memory controller, memory power delivery or the PCB itself. The problem is limited in scope, as many reviewers have stressed their cards with no issues, but we still don't know if the problems have a common denominator, or if it's just random defective cards. But we do know it doesn't affect all Micron memory.

Unfortunately there are too much misinformation around, partly due to anyone with any problem with RTX-cards throwing their hat into the ring. A good portion of reported problems is highly likely unrelated to graphics cards at all, and could be unstable system memory, bad PSUs etc.
Posted on Reply
#25
bug
lexluthermiester said:
While I'm not willing to call Steve a lair, one has to question the why NVidia themselves has not given details themselves. It's not like they have much of anything to lose.
Probably because it doesn't really matter?
If was some faulty component in there. It's like if we were 100% sure it was the VRAM, we'd all start cancelling the millions of GB we have ordered from Micron.
The parties that need to know are probably informed already. And we will know when we compare the original cards with future batches. But no, some of us need to know here and now otherwise the sky is falling. As usual.
Posted on Reply
Add your own comment