RTX 2080 Ti Failing Too Early

[D13H4RD]

5 minutes ago, 79wjd said:

Oh look, someone is done fanboying.

 

  Hide contents

/s

 

or is just capable of thinking rationally and not jumping to conclusions before proper information is actually available

Haha, but yeah. When reports first came out, I figured that it just might be a slightly higher than usual number of defective units and that any company with a good RMA policy should be able to handle it.

 

But as time went on and more reports flooded in, I think there's something up with the cards, probably GDDR6 or something.

[Taf the Ghost]

6 hours ago, LukeSavenije said:

read about this already. Now I don't regret buying my 1070 ti 3 months ago

Good product, nothing to regret about it.

[Taf the Ghost]

 

Hardware Unboxed has an AIB card with reference design die and has been talking to their contacts. Seems like it might be mostly a FE card issue? We'll find out more, but it'll be interesting if there is a serious design issue with the reference card.

[Taf the Ghost]

1 hour ago, D13H4RD said:

Haha, but yeah. When reports first came out, I figured that it just might be a slightly higher than usual number of defective units and that any company with a good RMA policy should be able to handle it.

 

But as time went on and more reports flooded in, I think there's something up with the cards, probably GDDR6 or something.

If HU's contacts about the FE design being the problem, it might be worth looking more closely at Buildzoid's breakdowns to see if there can be picked out the issue. The most direct thought is that it's the GDDR6, but if the AIB cards aren't failing at a higher rate then it's something about the PCB design. That would explain the RMA problems, as supply was already tight and now they're rushing to fix the PCB.

[Gullerback]

A friends FE died this week while playing AC:O.   Said he got some flashes of artifacts then just black screen.

[asus killer]

i guess they don't RTX but they sure as shit RMA. 

 

 

yes i now personally know one crazy idiot who bought one at insane prices and now it's dead

[Princess Luna]

25 minutes ago, Taf the Ghost said:

.

Did you even watch the video? Their FE is fine, what they had die on them was a Gigabyte card, albeit it used the reference PCB.

 

Mine's working fine as well, I still think people are taking this a bit out of proportion, for the youtubers it's great news, any news that gets them audience.

[Taf the Ghost]

Just now, VegetableStu said:

does this apply to AIB partners who use the same PCB? o_o do they buy PCBs that are already populated? o_o

Some, I believe, however they might use a different PCB producer. Which could point to it being an issue of the Reference Design production company.

[Taf the Ghost]

2 minutes ago, Princess Cadence said:

Did you watch the video? Their FE is fine, what they had die on them was a Gigabyte card, albeit it used the reference PCB.

 

Mine's working fine as well, I still think people are taking this a bit out of proportion, for the youtubers it's great news, any news that gets them audience.

Yes, I did. I edited the post to fix the mistake.

 

As for the cards, as Tim even said, one of the system builders had to RMA a bunch of cards, again with the reference design. Which is why it might either be a reference PCB issue in design or maybe the first run had some bad materials. 

[RollTime]

RTXbox 360

[jones177]

Since 2007 I have had 2 out of 16 Nvidia cards fail. The failures were 8 years apart and the cards were RMAed.

 

With these ods I keep older cards as spares to use while I am waiting for the RMA. 

 

I bought a RTX 2080 ti. If it dies I will use a GTX 1080 ti during the RMA.

 

I am not happy with the situation but that is how it goes with new high performance anything.

. 

 

 

 

 

 

 

 

 

 

 

 

 

  

[Nowak]

Read somewhere last night that the failure rate for the 2080 Ti is estimated to be around 10%. That is, 1 out of every 10 2080 Tis sold is a $1000+ lemon.

 

That's really bad.

[jasonvp]

9 minutes ago, Nowak said:

Read somewhere last night that the failure rate for the 2080 Ti is estimated to be around 10%. That is, 1 out of every 10 2080 Tis sold is a $1000+ lemon.

 

Don't believe everything you read on the Internet.  The failure rate is nowhere near that high.

[Nowak]

3 minutes ago, jasonvp said:

Don't believe everything you read on the Internet.  The failure rate is nowhere near that high.

Certainly seems higher than the normal failure rate. A $1000+ product should not die in less than a month, especially not in such numbers.

[jasonvp]

3 minutes ago, Nowak said:

Certainly seems higher than the normal failure rate.

Are you completely sure about that?  People bitch when they're unhappy.  Rarely do they rave when they're happy.  They just enjoy using the product.

 

So do we have hard numbers when it comes to failure rates?

 

(I'm kinda baiting you, because the answer is: no, we don't.)

[Nowak]

Just now, jasonvp said:

Are you completely sure about that?  People bitch when they're unhappy.  Rarely do they rave when they're happy.  They just enjoy using the product.

 

So do we have hard numbers when it comes to failure rates?

 

(I'm kinda baiting you, because the answer is: no, we don't.)

As I said, it's an estimation. Not an exact figure.

[pas008]

3 minutes ago, Nowak said:

As I said, it's an estimation. Not an exact figure.

how can you even estimate?

you know how many cards were produced?

you know how many failed?

[Nowak]

Just now, pas008 said:

how can you even estimate?

you know how many cards were produced?

you know how many failed?

not my estimation. it was from a german-language source.

[jasonvp]

5 minutes ago, Nowak said:

not my estimation. it was from a german-language source.

Then the same questions apply to them.  I'll bet you a paycheck they pulled "10%" out of thin air.

[D13H4RD]

10% is way higher than usual. I don't think it's that high unless there's calculated evidence that supports that theory. 

 

What I'm interested in is what exactly is causing them to fail. 

[mr moose]

8 hours ago, Hellion said:

Clearly my metaphor went above your head... Not going to bother breaking it down for you. Think a little harder.

 

Your poor attempt

 

I read this far, I don't need to read any further because logical and well reasoned discussion does not require such derision.

 

Also it appears a few others have already explained to you why you are wrong.

[mr moose]

7 hours ago, D13H4RD said:

10% is way higher than usual. I don't think it's that high unless there's calculated evidence that supports that theory. 

 

What I'm interested in is what exactly is causing them to fail. 

AMD had 13% at one stage,  I'd agree I don't like seeing it that high, but it is a fairly complex product with tight tolerances (they have been for a while now).  I imagine even a heatsink restraint at the wrong torque spec might cause issues with either heating or physical damage.

 

This is from 2014 but it gives an indication of where failures can occur and how common they might be.

 

https://www.pugetsystems.com/labs/articles/Video-Card-Failure-Rates-by-Generation-563/

[Maticks]

Back when i watched Gamers Nexus video with an Nvidia rep discussing overclocking on the RTX series cards.

I heard something alarming but thought surely that is wrong.

 

Summarising the video, the RTX cards are way more locked for Voltage over Pascal which they expect to last 5 Years overclocked.

The RTX cards are locked down a bit more so they don't blow up.

 

Fast Forward to 6:08 into the video "we're going from five years reliability to 1 year reliability."

 

These cards clearly are very close to blowing themselves up out of the box if they are "factory overclocked" who knows what the result will be but 3 increments 0.15mV's at best. and even less at worse. thats not a lot of room for error.

 

 

[leadeater]

1 hour ago, Maticks said:

Fast Forward to 6:08 into the video "we're going from five years reliability to 1 year reliability."

Nice memory on that one, watched that video and didn't really take much note of that. Rather interesting in light of this.

[mr moose]

2 hours ago, Maticks said:

Back when i watched Gamers Nexus video with an Nvidia rep discussing overclocking on the RTX series cards.

I heard something alarming but thought surely that is wrong.

 

Summarising the video, the RTX cards are way more locked for Voltage over Pascal which they expect to last 5 Years overclocked.

The RTX cards are locked down a bit more so they don't blow up.

 

Fast Forward to 6:08 into the video "we're going from five years reliability to 1 year reliability."

 

These cards clearly are very close to blowing themselves up out of the box if they are "factory overclocked" who knows what the result will be but 3 increments 0.15mV's at best. and even less at worse. thats not a lot of room for error.

 

 

 

1 hour ago, leadeater said:

Nice memory on that one, watched that video and didn't really take much note of that. Rather interesting in light of this.

It just goes further to how close to tolerances they might be.