GamersNexus Reproduced Multiple Modes of Failure on NVIDIA RTX 20-Series

cageymaru

cageymaru

Fully [H]
Joined
Apr 10, 2003
Messages
22,331
The guys over at GamersNexus asked their viewers to send their broken NVIDIA RTX 20-series cards to them for testing. This was done to see if they could reproduce the failures. Today they announced via Twitter that they have been able to reproduce 2 modes of failure in the RTX 20-series. Our readers have discussed the hardware failures that they have experienced. Hopefully this leads to NVIDIA fixing the issues so that customers will be happy. In the end that's all everyone wants.

As of today, we have successfully reproduced 2 modes of failure on the RTX 20-series cards that were sent in.
 
These guys are leading the charge on actual technical discussion of parts. The fearless teardowns are great. So refreshing.

And that doesn't detract from HardOCP at all, it compliments it really. [H] spends more time on actual use and value and balances it with informing on hardware.

These GamerNexus guys just tear stuff down, look at part numbers and put temp probes everywhere to see what going on at a deeper level.
 
I'm glad this kind of work is being done. I wouldn't be surprised to find HardOCP in the mix on videos like this and doing collaborative work with GamersNexus. I could even see them actually doing this and just no publishing the relationship because they don't want GamersNexus to be black balled by Nvidia due to the NDA they signed.

It is a solid piece by GamersNexus. Lets see if Nvidia actually takes notice and takes action. The tell will be in how they play their hand.
 
Grimlaking said:
I'm glad this kind of work is being done. I wouldn't be surprised to find HardOCP in the mix on videos like this and doing collaborative work with GamersNexus. I could even see them actually doing this and just no publishing the relationship because they don't want GamersNexus to be black balled by Nvidia due to the NDA they signed.

It is a solid piece by GamersNexus. Lets see if Nvidia actually takes notice and takes action. The tell will be in how they play their hand.
Click to expand...

Nvidia taking notice and action might equal putting GamersNexus on the naughty list. That's cheaper than fixing cards.
 
I installed my waterblock on Monday and still haven't had problems, but I'm definitely nervous when I wake my computer from sleep.

*sweaty fingers crossed
 
The risk that this kind of thing will happen is part-and-parcel of being on the bleeding edge of technology, and always will be.
GPUs are too complicated for every possible scenario to be tested, until, of course, they are deployed en masse to the users.

Hopefully, NVidia will make it right without nerfing the performance of feature set of the affected cards.
 
So they reproduced failures on cards that were known already to be broken? This doesn't indicate a trend, just that they have verified that the users cards were broke dick.... I have a Zotac 2080 ti, and enjoying the shit out of it...so neener neener....
 
phawkins633 said:
So they reproduced failures on cards that were known already to be broken? This doesn't indicate a trend, just that they have verified that the users cards were broke dick.... I have a Zotac 2080 ti, and enjoying the shit out of it...so neener neener....
Click to expand...

So you don't think its relavent for the consumer to find out why they are failing?

I've been on the fence about a 2080 to for months, I would like to know.
 
7w8uh58adxv11.png
 
I saw during their fan mail segment they were unboxing 2080's that had failed.

I'm really confused why someone would send them a several hundred dollar GPU rather than RMA it. I hope Steve is sending them back.
 
bigdogchris said:
I saw during their fan mail segment they were unboxing 2080's that had failed.

I'm really confused why someone would send them a several hundred dollar GPU rather than RMA it. I hope Steve is sending them back.
Click to expand...
Yep, he is sending them back when they are done.
 
bigdogchris said:
I saw during their fan mail segment they were unboxing 2080's that had failed.

I'm really confused why someone would send them a several hundred dollar GPU rather than RMA it. I hope Steve is sending them back.
Click to expand...
"email us at [email protected] with information on your card, what it’s doing, and what country you live in. This will help us bring in more defects. We will pay for shipping both ways and will send some GN merch back with the card once we’re done testing."
 
So can someone spill what faults a 2080ti has which cause them to die?
 
phawkins633 said:
So they reproduced failures on cards that were known already to be broken? This doesn't indicate a trend, just that they have verified that the users cards were broke dick.... I have a Zotac 2080 ti, and enjoying the shit out of it...so neener neener....
Click to expand...

They stated that there were issues with people who had high refresh monitors with gsync as well as multi-monitors. Not everyone has those, so they might believe their card works fine. Others, such as me, have both. In fact, I'd say that the people who are willing to spend $1200+ on at a 2080TI (or two) are probably the same group of people who buy those high refresh gsync monitors, so it's not surprising that quite a lot of people are seeing issues.

Now, the fact that I have two new cards which don't blue screen constantly makes me think that the driver fix is more of a band aid for a hardware error.
 
Nytegard said:
They stated that there were issues with people who had high refresh monitors with gsync as well as multi-monitors. Not everyone has those, so they might believe their card works fine. Others, such as me, have both. In fact, I'd say that the people who are willing to spend $1200+ on at a 2080TI (or two) are probably the same group of people who buy those high refresh gsync monitors, so it's not surprising that quite a lot of people are seeing issues.

Now, the fact that I have two new cards which don't blue screen constantly makes me think that the driver fix is more of a band aid for a hardware error.
Click to expand...

I have a 2080ti FE and an Acer X27.

No issues.
 
while not 4k 60, I'm trying to keep a 1440p 120fps threshold. I gamed last night for a few hours at a max oc 2100/15500 which had the fans running 85% to keep at 80C. This boosted my 3dmark timescore from stock 13100 to 15330. Did not notice any artifacting playing FC5 and only stopped because 85% fans were enough to distract from the game. FC5's minimum bumped up a bit from the OC, but not enough to warrant running at a max oc.

I'd like to know more about this discovery to see if any particular combination of usage is what breaks it.
 
Stimpy88 said:
Oh well, I guess everyone else is dreaming it then? Or do I just say, good for you? Or do I say, wow I’m jealous, he has so much money...
Click to expand...

Alot of hate there fella. You should steer clear of the dark side of the force.
 
phawkins633 said:
So they reproduced failures on cards that were known already to be broken? This doesn't indicate a trend, just that they have verified that the users cards were broke dick.... I have a Zotac 2080 ti, and enjoying the shit out of it...so neener neener....
Click to expand...

We still don't have valid statistical data to back up high rates of failure and I'm sure NVIDIA will keep those cards close to their chest unless a subpoena from a class action pops up. So there might be a problem, there might not be. We don't have data to say one way or another.

That said,

I think what is of greater interest here is the common mode of failure. If for example, memory chips ARE overheating because of a poorly routed power bus line overheating, then that would indicate a design flaw that would affect many MANY card holders eventually. This is akin to the RROD on XBOX. You knew if was coming because it constantly pushed the edge of heat limits. Before you yell AMD is guilty of the same crimes due to RROD, remember Microsoft designed both the cooling and PCB. However in this case NVIDIA designed the reference PCB.

If there was an indicative trend it would be cause for redesign and lots of warranty claims. Remember the motherboard solder fiasco and how NVIDIA wormed their way out of paying for those fixes? It's one of the reasons I hesitate to buy NVIDIA. You would have thought NVIDIA would have learned by now if it's a heat issue. At least Microsoft learned and made sure all future consoles were properly cooled.

It might be time to short the stock given how many are out in the field, and how long it takes to validate a revision pcb and predicted failure rate. (Don't take that as financial advice. Without hard statistical numbers of failure analysis, it can't be predicted how much this would cost. This is why actuaries are worth their weight in gold.)

P.S.: I wonder if NVIDIA has contacted gamers nexus yet? I'm sure that would be a fun phone call to listen in on.
 
Last edited by a moderator:
Latest nvidia driver release 416.81 lists this fix -

[GeForce RTX 2080 Ti][G-Sync]: Blue-screen crash may occur when exiting games
when using a G-Sync monitor with a non-G-Sync monitor. [2431628]

This would fix the BSOD described by gamers nexus.
 
Must be a slow news day for GamersNexus that he's doing work for nvidia. ;) "Bugs in nvidia products, driver patches needed, some unnecessary RMAs have been filed, water wet."
 
velusip said:
Must be a slow news day for GamersNexus that he's doing work for nvidia. ;) "Bugs in nvidia products, driver patches needed, some unnecessary RMAs have been filed, water wet."
Click to expand...
Actually kinda good news in that some of this stuff will be fixed. Now the other issues how they turn out will be very interesting.
 
Okay, at least it's good to know that the blue screen i got with my 2080 is probably not a hardware issue. Looks like that IRQL thing will just be a driver fix.
 
You must log in or register to reply here.
Back
Top