I've been invaded by aliens. RTX 2080 Ti *Updated 5/6*

[IEC]

He's dead, Jim. Not even 5 months old:

eVGA RTX 2080 Ti "Black Edition"
Run for most of its short life at 80% power limit
Always running nice and cool with 7 case fans including 2 bottom intakes blowing cold air directly on it.

Subsequent reboots yielded black screen/crash in Windows as soon as it hits 3D clocks.

Press 'F' to pay respects.

*****

4/29/2019 update: Per eVGA support advice, I tried the following:

Please try the following.
Open up Nvidia Control Panel. Click on Manage 3D settings.
Find the Power Management settings and then set it to Prefer Maximum Performance.
That has been a fix for a lot of people with the space invader pattern.
See if that helps you as well.

Because of the black screening issue, it took several tries to implement the suggested "fix" (really, a less-than-ideal bandaid). I'm happy to report that the card actually functions, with some serious drawbacks.
1) The card runs at the base clock as its minimum clock
2) The card draws ~70W GPU *core only* at "idle" as a result (over 100W for sure with GDDR6/VRM losses included).
3) If I were to ever try to resell the card, the fact that you have to use this bandaid (and initially struggle with black screens/freezing to implement it) would probably seriously tank its resale value.

So, the card works and I'm stress-testing it now. But I'm likely not returning it to duty in my main rig due to idle power draw.

The question remains as to the cause of this issue, and why it only affects certain cards. Is it a design defect, an overly aggressive power management profile, a hardware issue, or some combination of the above?

Another question: Will whatever caused this issue lead to a reduced lifespan for the card? I still have over 2.5 years of warranty left, and the card WILL get beat on, so stay tuned on that front.

*****

5/6/2019 update:
After my polite but firmly worded response requesting an RMA, I got a response back from a different tech suggesting the following:
Run DDU and remove/clean drivers
Install latest nVidia driver (430.39 WHQL as of this writing) - I used express install
This has the effect of returning all settings to default (including the power management kludge previously suggested)

I did as instructed and rebooted. I promptly got blue and black screens for 20-30s at a time every 1-2 minutes of desktop usage. Yup, not the driver. I await eVGA's next response.

 

[Ranulf]

RIP

 

[.vodka]

F

Let's hope the replacement lasts more.

 

[mopardude87]

Very sad people have to deal with this on such a high end card. Not like its a one time thing its a very common issue from what i hear. Its pushing me away from even wanting one at this point. Even though by the time i save up for this the next generation may very well be here.

Sorry for your lost. Good luck on the next one.

 

[Elfear]

That sucks man.

 

[VirtualLarry]

Wow, sorry to hear that, and quite frankly, wholly disappointed by Nvidia for allowing another "Bumpgate"-esque mfg gaffe.

These aren't even 7nm, right?

 

[alcoholbob]

Well as long as you bought new its covered under warranty. But if u bought it used you can try underclocking the memory.

 

[Indus]

Last time that happened to me, it was a GTX 560 that I had bought 2nd hand. I was angry for 2 weeks straight. Infact it still irks me and I refuse to buy 2nd hand electronics now.

Hopefully you have the warranty and have a better road ahead than I did.

 

[Shmee]

Pity, RIP. Good thing EVGA has a good warranty service generally. Good luck on the replacement.

I am not sure if this is true, but I heard that this issue is only found on reference PCB 2080Ti cards. For those considering a 2080Ti, you may want to take that into account and get one with a fancy non reference PCB. I would be interested if anyone has done further analysis into this.

 

[DrMrLordX]

F

Hope the warranty process goes well for you. Also, I want to know why @Innokentij found it necessary to downvote you.

These aren't even 7nm, right?

No, they aren't.

@Shmee is OP's EVGA "Black Edition" a reference PCB card?

 

[Shmee]

According to this list of 2080 Ti card models, it is. https://www.overclock.net/forum/69-...0-ti-owner-s-club.html#/topics/1706276?page=1

 

[coercitiv]

F , even the Titans fall sometimes.

 

[Mnpctech]

bummer, at least EVGA is pretty good in handling RMAs

 

[ozzy702]

Pity, RIP. Good thing EVGA has a good warranty service generally. Good luck on the replacement.

I am not sure if this is true, but I heard that this issue is only found on reference PCB 2080Ti cards. For those considering a 2080Ti, you may want to take that into account and get one with a fancy non reference PCB. I would be interested if anyone has done further analysis into this.

For EVGA I believe the FTW3 and up cards have custom PCBs but everything below is still reference. I have an EVGA 2080TI XC Ultra and have my fingers crossed that it doesn't die like this on me. EVGA's customer service is the only reason I took the risk on a 2080TI and so far so good...

 

[ericlp]

I've read reviews on Amazon, people having the same problem, even earlier. You aren't alone. I wonder if these chips just can't handle the heat, or maybe it's the ram that is failing. Hope you get a new one and it lasts longer than that. I guess.... They don't build em like they use too.

 

[Stuka87]

Bummer, sucks to spend that kind of money and have it die. I am sure EVGA will get you a new one in quick order. I will be curious to see if its the same exact model, or an updated one. Despite what nVidia claims, this is obviously a real issue.

Kind of a shame the OP got a negative reaction for his post from Innokentij.

 

[IEC]

I've read reviews on Amazon, people having the same problem, even earlier. You aren't alone. I wonder if these chips just can't handle the heat, or maybe it's the ram that is failing. Hope you get a new one and it lasts longer than that. I guess.... They don't build em like they use too.

See update from eVGA/my testing notes. It's still unclear what the actual cause is because nVidia has not elaborated on the cause beyond a terse statement about "limited test escapes".

My best guess is that part of the issue is an overly aggressive power management scheme combined with some sort of hardware or design defect in some RTX cards.

 

[Red Squirrel]

Seeing that gave me hardware troubleshooting PTSD. Sorry to hear!

I had the same issue with a 560 TI in Linux. It was fine in Windows but in Linux it would randomly start making squares like that and also cause weird GUI lockups. Took me almost a year to troubleshoot that, was a freaking pain. I had two cards and tried both but at the time I was not sure what part of the system was at fault, was just trying everything, different ram etc. I think it ended up being the video driver, but with no real way around that I ended up just buying a different make/model card so that it does not use the same driver. Recently sold both on Ebay for a few hundred so there's that. (I disclosed that they don't work well in Linux because of the driver)

 

[Ketchup]

They actually call that a fix? LOL! They must be getting a lot of returns for this. Unless they have some "magic" firmware update coming, a "so what?" would be my response to them.

 

[IEC]

They actually call that a fix? LOL! They must be getting a lot of returns for this. Unless they have some "magic" firmware update coming, a "so what?" would be my response to them.

I'm pretty easy-going, but even I was a bit taken aback by them saying it's a fix.

It's an ugly bandaid at best. Unless they can implement a real fix that doesn't disable power management in firmware/vBIOS or via driver, I'm disinclined to buy any further Turing products (I have 3). For such an expensive card it's really an unacceptable defect.

 

[Red Squirrel]

With a fix like that it seems the only thing this card is really good for is mining. Assuming it actually works for that and does not crash lol.

 

[thigobr]

Probably the voltage/frequency curve stored on the BIOS is not stable for some of these GPUs... Bad binning by nVidia? Who knows...

 

[Stuka87]

So from the sounds of it nVidia may have gone a bit overboard on trying to get power usage down? 70+ watts at idle with the performance setting is kind of nuts, as thats most likely high enough to cause the fans to run (Which would drive me nuts as I have gotten use to fans shutting off when at idle). Tests show they use around 13-20W at idle when using the default power mode.

But like you said, the card is now basically unsellable :/

 

[amenx]

I wonder if the issues affecting RTX cards had something to do with bad batches of 12nm wafers at TSMC.
https://www.extremetech.com/computi...stroys-tens-of-thousands-of-nvidia-gpu-wafers

Although the destroyed wafers were supposedly isolated, were other batches 'partially' contaminated and got through to manufacturing or retail? When reports of dead or dying RTX cards first surfaced several months ago, Nvidia described the problem as "limited test escapes" involving their FE cards. Did other 'test escapes' reach their board partners? Remember, these are massive dies and thus more difficult to get good yields out of.

 

[DrMrLordX]

I'm pretty easy-going, but even I was a bit taken aback by them saying it's a fix.

I'm genuinely fascinated by their proposed fix. We're finally beginning to peel back the layer of secrecy surrounding the space invader Turing problem.

FOR SCIENCE!!!!!!

That being said, I'm disappointed that they didn't just hustle your new card to you right away thank you sir have a nice day but imagine the kind of financial trouble they'll be in if they have to do this with any significant number of cards. We have no reliable data on exactly how many cards have been returned due to this problem.

It's an ugly bandaid at best. Unless they can implement a real fix that doesn't disable power management in firmware/vBIOS or via driver, I'm disinclined to buy any further Turing products (I have 3). For such an expensive card it's really an unacceptable defect.

Well let's face it, the first reports of space invaders plus nVidia's (as you put it) terse response was not particularly inspiring. I'd put a "caveat emptor" on all their 2080Tis (at least) until something more substantial comes out about the problem.

Kind of makes me wonder if all that it would take to fix the problem would be a new BIOS for the card. It looks like the card is space invadering from something malfunctioning in a low power state.