Do YOU have Problems with YOUR RTX card?

Do YOU have Problems with YOUR RTX card?

  • My RTX 2080 Died

    Votes: 2 0.9%
  • My RTX 2080 Ti Died

    Votes: 30 13.6%
  • My RTX 2080 has BSOD Issues

    Votes: 3 1.4%
  • My RTX 2080 Ti has BSOD Issues

    Votes: 13 5.9%
  • My RTX 2080 has Other Issues

    Votes: 7 3.2%
  • My RTX 2080 Ti has Other Issues

    Votes: 23 10.5%
  • I have RMA'd my RTX 2080

    Votes: 2 0.9%
  • I have RMA'd my RTX 2080 Ti

    Votes: 26 11.8%
  • My RTX is a Founders Edition

    Votes: 62 28.2%
  • I have NO ISSUES with my RTX card

    Votes: 156 70.9%

  • Total voters
    220
So, based on the data so far, out of 61 total cards, 6 died and an additional 4 have "issues" such as BSODs

The "it died" rate (admittedly from a small sample range) seems to be hovering around 10% and, if you add in the BSOD and other issues, there's an ~16% of people who are experiencing issues.

Seems a bit higher than the ~3% one would expect on launch of a new generation.

Also seems, based on some of the responses I've seen online, that nVidia is aware there is an issue of some sort and is working hard to remedy the situation...nVidia isn't pulling their normal bullshit of deleting threads on the forum (yet) so I do hand it to them for handling the issue with a modicum of dignity.

That said, some of the AIBs need to remove their head from their ass. When the P67 SATA bug occurred, every mobo mfgr I dealt with (7 different mfgrs) cross-shipped new boards with RMA Labels for the old ones.

Zotac in particular in this case is doing themselves and the community a disservice by being assholes about RMAs for a known issue.

EDITED TO ADD: Thanks for BETA Testing for the rest of us. HAHAHA :LOL:
 
I've had one 2080 Ti on stock settings and stock cooling running in a machine for a few weeks. Everything is fine, but I noticed last night that the card was throttling due to Vrel even though it was running at only about 70% TDP. The PSU is a Corsair HX1200i and this is the only GPU in the system, so there should be an abundance of power available and all outputs should be stable. I hope this isn't an early sign of impending failure.
 
My EVGA 2080TI XC, out of the box would only work in VGA mode and was not recognized by the system or NVIDIA's Geforce Experience in three different boxes. RMAed and the replacement is working fine.
 
My EVGA 2080TI XC, out of the box would only work in VGA mode and was not recognized by the system or NVIDIA's Geforce Experience


MY MSI did that at first, until i realized i hadnt installed the latest drivers. downloaded and installed.. good to go.
 
Have a gaming z; aside from it being huge, heavy and comes with a bracket thats mostly just for decoration, no issues here... Only had the card for a few days though.
 
So, based on the data so far, out of 61 total cards, 6 died and an additional 4 have "issues" such as BSODs

The "it died" rate (admittedly from a small sample range) seems to be hovering around 10% and, if you add in the BSOD and other issues, there's an ~16% of people who are experiencing issues.

Seems a bit higher than the ~3% one would expect on launch of a new generation.

Also seems, based on some of the responses I've seen online, that nVidia is aware there is an issue of some sort and is working hard to remedy the situation...nVidia isn't pulling their normal bullshit of deleting threads on the forum (yet) so I do hand it to them for handling the issue with a modicum of dignity.

That said, some of the AIBs need to remove their head from their ass. When the P67 SATA bug occurred, every mobo mfgr I dealt with (7 different mfgrs) cross-shipped new boards with RMA Labels for the old ones.

Zotac in particular in this case is doing themselves and the community a disservice by being assholes about RMAs for a known issue.

EDITED TO ADD: Thanks for BETA Testing for the rest of us. HAHAHA :LOL:


Well, if we model this as a binomial distribution we can do some inferential statistics to try to estimate the population failure rates.

Code:
Test and CI for One Proportion
Sample  X   N  Sample p         95% CI
1       6  61  0.098361  (0.036960, 0.201896)

So, based on this sampling with 95% confidence anywhere from 3.7% to 20.2% of these cards are outright failing.


Code:
Test and CI for One Proportion
Sample  X   N  Sample p         95% CI
1       4  61  0.065574  (0.018154, 0.159469)

And anywhere from 1.8% to 15.9% of these cards have other issues.


Code:
Test and CI for One Proportion
Sample   X   N  Sample p         95% CI
1       10  61  0.163934  (0.081517, 0.280885)

This leaves us with a total outright failure or issue percentage of somewhere beterrn 8.2% and 28.1%


That doesn't paint a pretty picture for Nvidia. Even at the low end as a manufacturer I wouldn't want 8% of my parts sold failing.
 
Last edited:
11 percent failure rate on [H]

that's not great is it?

Again, it's a relatively small sample size, so the true population failure rate is somewhere in a range around it.

See my stats calculations above.

As the sample size grows, that range shrinks and you get closer to finding where the true failure rate lies.
 
Well, if we model this as a binomial distribution we can do some inferential statistics to try to estimate the population failure rates.

Code:
Test and CI for One Proportion
Sample  X   N  Sample p         95% CI
1       6  61  0.098361  (0.036960, 0.201896)

So, based on this sampling with 95% confidence anywhere from 3.7% to 20.2% of these cards are ourtright failing.


Code:
Test and CI for One Proportion
Sample  X   N  Sample p         95% CI
1       4  61  0.065574  (0.018154, 0.159469)

And anywhere from 1.8% to 15.9% of these cards have other issues.


Code:
Test and CI for One Proportion
Sample   X   N  Sample p         95% CI
1       10  61  0.163934  (0.081517, 0.280885)

This leaves us with a total outright failure or issue percentage of somewhere beterrn 8.2% and 28.1%


That doesn't paint a pretty picture for Nvidia. Even at the low end as a manufacturer I wouldn't want 8% of my parts sold failing.



This is what I love about H.

Somebody asks a question/poll.

Instead of an answer, we get code instead.

H Gold award goes to to you dude,
 
Again, it's a relatively small sample size, so the true population failure rate is somewhere in a range around it.

See my stats calculations above.

As the sample size grows, that range shrinks and you get closer to finding where the true failure rate lies.

that's why i specified on [H]

it's still not great.
 
I selected other issues as my 2080 Ti FE is just totally unstable, the driver is crashing randomly in games and I'm getting graphics exception and warp errors in the event viewer. Also, the LED logo is randomly turning red. I am trying to return it for a refund, but the call centre people keep saying wait 48 hours for the shipping label email which never comes! Disgraceful service, how hard is it to arrange a courier to come and collect the card?

My GTX 1080 is totally fine by the way.
 
I selected other issues as my 2080 Ti is just totally unstable, the driver is crashing randomly in games and I'm getting graphics exception and warp errors in the event viewer. Also, the LED logo is randomly turning red. I am trying to return it for a refund, but the call centre people keep saying wait 48 hours for the shipping label email which never comes! Disgraceful service, how hard is it to arrange a courier to come and collect the card?

My GTX 1080 is totally fine by the way.

you should have also selected dead card. it's dead.
 
that's why i specified on [H]

it's still not great.


Agreed.

It is still possible based on these numbers that the outright failure rate is as low as 3.7%, but even that is a bit high (remember, six sigma manufacturers like to talk about failures in PPM, parts per million, not percentages.)

Based on how much noise I'm hearing - however - I'm guessing that the true failure rate is much closer to the high end of the range (20.2%) than it is to the low end of the range.
 
Last edited:
Well, if we model this as a binomial distribution we can do some inferential statistics to try to estimate the population failure rates.

Code:
Test and CI for One Proportion
Sample  X   N  Sample p         95% CI
1       6  61  0.098361  (0.036960, 0.201896)

So, based on this sampling with 95% confidence anywhere from 3.7% to 20.2% of these cards are outright failing.


Code:
Test and CI for One Proportion
Sample  X   N  Sample p         95% CI
1       4  61  0.065574  (0.018154, 0.159469)

And anywhere from 1.8% to 15.9% of these cards have other issues.


Code:
Test and CI for One Proportion
Sample   X   N  Sample p         95% CI
1       10  61  0.163934  (0.081517, 0.280885)

This leaves us with a total outright failure or issue percentage of somewhere beterrn 8.2% and 28.1%


That doesn't paint a pretty picture for Nvidia. Even at the low end as a manufacturer I wouldn't want 8% of my parts sold failing.

Wasn’t the overall failure rate on the 290x like 8-9%? I think nVidia is generally much lower. I don’t think I’ve seen failure rates since that generation.

A 10% failure rate wouldn’t surprise me horribly for LRIP on such a complex card... unforunately.
 
So, around 20% of cards are dead or have issues? That's an insanely high rate.
 
I think my only real weird glitch was when I first installed the card. I could not get it to display on my primary screen. It took 2 reboots for it to decide accept my primary screen was my primary screen.
 
I wouldn't really say it's 20%. Rather, I'd take the number of people who voted to RMA against the number of people who own a card. That seems to be closer to about 10-12%.
 
The rate isn't as high as the poll might lead you to believe.
1) There's a sampling bias inherent in the poll since unhappy customers make noise whereas most happy ones do not.
2) The poll doesn't account for multiple cards (both concurrent ownership and sequential via RMA)

Also, for attribute data, the minimum sample size for 95% CI has not yet been met.
 
So, around 20% of cards are dead or have issues? That's an insanely high rate.

20% is within the realm of possibility, but on the high side of the range.

The thing is, we are trying to infer what the failure rate of the population (all cards sold) is based on our sample size here.

With 95% confidence we can say - based our small sample - that this true population failure rate is somewhere between 3.7% and 20.2%. That's a rather large range because our sample size is small. The more the sample size grows, the more that range will shrink and the closer we will get to the true occurrence rate in the field.
 
The rate isn't as high as the poll might lead you to believe.
1) There's a sampling bias inherent in the poll since unhappy customers make noise whereas most happy ones do not.
2) The poll doesn't account for multiple cards (both concurrent ownership and sequential via RMA)

Also, for attribute data, the minimum sample size for 95% CI has not yet been met.

Partially agree. The sample is not perfect. We know that, but it is what we have to work with.

There is - however - no minimum sample size for binomial distributions. You can go as small as you'd like, it's just that your confidence interval will be unworkable large, as it is with this data.

Somewhere between 3.7% and 20.2% at 95% confidence isn't incredibly useful.
 
Strangely, I had one black screen at boot the day after installing my 2080ti, but since then I've had zero issues with the card. At this point I've point considerable hours on the card @ 84c with AssCreed and World of Boatships
 
We also don't know if people are responding to the questions consistently, nor do we know if they have a driver mess, a bad PSU, a splitter from the PSU to their card rather than separate power feeds, etc, etc.
 
For those of you that have had your fail, when does it typically happen? Out of box? Within a week? After a couple of weeks?
 
If you own a failed FE, can you post the first 4 digits of the serial number? Trying to collect data, I think their early batch might be bad.
 
For those of you that have had your fail, when does it typically happen? Out of box? Within a week? After a couple of weeks?

My first one didn't even boot out of the box. The third one which had artifacts had them pretty much within an hour. (I ran time spy and time spy extreme once).

If you own a failed FE, can you post the first 4 digits of the serial number? Trying to collect data, I think their early batch might be bad.

Both ones which failed started with 03238.
 
i have a gigabyte gaming oc 2080ti only issue i have is horrible coil whine. but once the card heats up and the fans start spinning i cant hear it over the fans.


the aorus engine software is horrible. cant game while its running random major stuttering happens.

I got the 2080 version and I uninstalled the aorus software, it gave an error every hour in event logger, also crashed games sometimes, I mainly installed it to change the colour of the LED and that seems to stick, it also took over the lighting options of my CM mouse for some reason.

I do wonder, recently I have seen some image corruption in YouTube video's, it usually only lasts a couple sec if that, not on all video's, but not sure if that's my machine or YT itself (looks like an image that's hugely pixelated) e.g. in this one


yterror.jpg
 
Nice. What kind of temps/OC are you getting on that EVGA board?

EK waterblock / HWLabs 360GTS / Noctua NF-F12's load temps
44C with the fans at 60% (somewhat audible)
58C with the fans at 30% (not audible from 3ft)

When I first put it under water, I ran some benchmarks and it was boosting to 2100 with +700 on the memory. My daily settings are boosting between 2030 and 2070 depending on temps with +500 on the memory. The 130% power limit bios was a significant boost in clock & memory speed for me.

I do have one weird issue that seems to be specific to Chrome - if I access certain websites like the Spotify Web player while in a game, it will cause wicked stuttering. Turning off hardware acceleration in Chrome's settings didn't seem to help. That is really the only issue I have dealt with so far and using Firefox for those websites instead seems to solve the problem. I am assuming that this will be fixed with either a driver update or Chrome update.
 
MSI Gaming X Trio RTX 2080 Ti here.

Zero issues with my card.

However, I did have one issue I should describe. For some reason, after buying this card ... my bank account was missing a lot of money. I am mostly working with myself in hopes of figuring out that problem.
 
MSI Gaming X Trio RTX 2080 Ti here.

Zero issues with my card.

However, I did have one issue I should describe. For some reason, after buying this card ... my bank account was missing a lot of money. I am mostly working with myself in hopes of figuring out that problem.

Try uninstalling then re-installing your bank account with information you looked up from a data breach posted on the dark web. You should then be back in business!
 
The thread title and OP says "RTX Cards", but poll options only include 2080 and above, and last 3 choices say only "RTX". Are there any reported issues with the RTX 2070's?

I haven't seen any noteworthy news about anything but the 2080ti. And honestly, it's mostly just the FE version.
 
I got the 2080 version and I uninstalled the aorus software, it gave an error every hour in event logger, also crashed games sometimes, I mainly installed it to change the colour of the LED and that seems to stick, it also took over the lighting options of my CM mouse for some reason.

I do wonder, recently I have seen some image corruption in YouTube video's, it usually only lasts a couple sec if that, not on all video's, but not sure if that's my machine or YT itself (looks like an image that's hugely pixelated) e.g. in this one


View attachment 116721

that cat vid is a IRL ray tracing demo
 
For those of you that have had your fail, when does it typically happen? Out of box? Within a week? After a couple of weeks?
It was crashing the display driver from the start randomly, particularly in Quake Champions for some reason which is not very demanding. This got worse over time until I got warp errors and graphics exception errors in the event viewer. The LED also randomly turns red. For some reason 416.34 was less stable than the older 411.70 until that became totally unstable over time too.

My serial number starts with 03238 too.
 
Back
Top