Amazon's mmo New World is bricking 3090 gpus

[Bombastinator]

5 minutes ago, wkdpaul said:

Seems to be higher tier cards from both sides, in the threads I linked (there are others) there's people that have the issue with their 2080ti & 2080 super (though the cards didn't die, only got the 100% fans & black screen bug that was fixed with a hard reset).

Implies there was/is also a code problem But the bricking though? That’s more serious than  just a software bug.  If bricking is possible there’s a big problem.

[joaopt]

20 minutes ago, wkdpaul said:

This is completely false.

 

There are plenty of reports of OTHER cards breaking, even on New Worlds own forums ;

https://forums.newworld.com/t/issues-with-evga-rtx-3090-ftw3-ultra/112757/8

https://forums.newworld.com/t/known-issue-nvidia-rtx-3090-series-100-gpu-usage/126068

 

Please make sure you're aware of the issue before throwing blanket statements like this.

that's also not true, only the ftw3 3090's broke, none of the other cards did. They just shut down or crashed, we don't know the condition of those cards but it's normal for demanding games to bring badly mantained cards or with inferior cooler to it's knees.

[Bombastinator]

6 minutes ago, joaopt said:

that's also not true, only the ftw3 3090's broke, none of the other cards did. They just shut down or crashed, we don't know the condition of those cards but it's normal for demanding games to bring badly mantained cards or with inferior cooler to it's knees.

I see breaking as the big issue.  Bad code happens.  It’s possible, even easy to make code that attacks a system and forces the system to defend itself. Shutting down is not nearly as bad as breaking. That’s defending itself vs failing to defend itself.  A card that merely shuts down can simply be used elsewhere or restarted when the code is fixed.  A bricked card can’t. 

[WkdPaul]

28 minutes ago, joaopt said:

that's also not true, only the ftw3 3090's broke, none of the other cards did. They just shut down or crashed, we don't know the condition of those cards but it's normal for demanding games to bring badly mantained cards or with inferior cooler to it's knees.

No, people have reported that other 3090 broke, the link I supplied proves so.

 

Again, if you're saying that people are lying, while they actually give proof they aren't, that's extremely disingenuous.

 

Here's a direct link to someone that has a fried 3090 that IS NOT a FTW3 ;

https://forums.newworld.com/t/issues-with-evga-rtx-3090-ftw3-ultra/112757/9

 

[Bombastinator]

3 minutes ago, wkdpaul said:

No, people have reported that other 3090 broke, the link I supplied proves so.

 

Again, if you're saying that people are lying, while they actually give proof they aren't, that's extremely disingenuous.

 

Here's a direct link to someone that has a fried 3090 that IS NOT a FTW3 ;

https://forums.newworld.com/t/issues-with-evga-rtx-3090-ftw3-ultra/112757/9

 

So the actual bricking is happening only with 3090s but with multiple iterations of that general type of card only.  Implies the problem is with Nvidia which created the 3090 and it’s spec.  Might just be that 3090s are MORE vulnerable but a whole series of things are somewhat vulnerable.  This is still bad. 

[WkdPaul]

Just now, Bombastinator said:

So the actual bricking is happening only with 3090s but with multiple iterations of that general type of card only.  Implies the problem is with Nvidia which created the 3090 and it’s spec.  Might just be that 3090s are MORE vulnerable but a whole series of things are somewhat vulnerable.  This is still bad. 

For now, the threads on New Worlds forum are showing 3090s are getting bricked, and other GPUs have the bug but aren't breaking, though some are reporting bricked GPUs other than the 3090 on Reddit, I'm on my phone right now so I can't easily link to the conversation I saw about it, will do later when I have time.

[Bombastinator]

7 minutes ago, wkdpaul said:

For now, the threads on New Worlds forum are showing 3090s are getting bricked, and other GPUs have the bug but aren't breaking, though some are reporting bricked GPUs other than the 3090 on Reddit, I'm on my phone right now so I can't easily link to the conversation I saw about it, will do later when I have time.

I hope this isn’t one of those just found but there for a long time flaws like specter/meltdown.   They can be out of control before they are discovered.

[WkdPaul]

1 hour ago, Bombastinator said:

*snip*

Like I said, other GPUs than the FTW3 are getting bricked, found the Reddit thread I talked about ;

 

 

And there's more (in that thread and others), but with that said, it's primarily FTW3s, the other GPUs could possibly be defective ones that would've died anyway, we have no way of knowing, but saying people are lying about this is pretty disingenuous (not referring to you BTW).

[Bombastinator]

6 minutes ago, wkdpaul said:

Like I said, other GPUs than the FTW3 are getting bricked, found the Reddit thread I talked about ;

 

 

And there's more (in that thread and others), but with that said, it's primarily FTW3s, the other GPUs could possibly be defective ones that would've died anyway, we have no way of knowing, but saying people are lying about this is pretty disingenuous (not referring to you BTW).

I’m not disbelieving you. I don’t know what this is in reply to so it’s hard for me to comment. It could be an old post or referencing a specific statement by someone else. This is why snips are difficult.  There’s no way to backtrack what was being discussed sometimes.  There was apparently conflicting statements by different people at one point earlier about what was actually being affected. 

[Vishera]

22 minutes ago, wkdpaul said:

Like I said, other GPUs than the FTW3 are getting bricked, found the Reddit thread I talked about ;

 

 

And there's more (in that thread and others), but with that said, it's primarily FTW3s, the other GPUs could possibly be defective ones that would've died anyway, we have no way of knowing, but saying people are lying about this is pretty disingenuous (not referring to you BTW).

Buildzoid speculated from the way those cards died,that the cards died from design flaws:

1.Overcurrent Protection isn't configured properly so it allows more current than the card can handle

2.VRMS - The amount of phases for MSVDD or memory power is not enough,so fuses or even the VRMs themselves blow.

[WkdPaul]

27 minutes ago, Bombastinator said:

I’m not disbelieving you. I don’t know what this is in reply to so it’s hard for me to comment. It could be an old post or referencing a specific statement by someone else. This is why snips are difficult.  There’s no way to backtrack what was being discussed sometimes.  There was apparently conflicting statements by different people at one point earlier about what was actually being affected. 

Was quoting you because I saw that you marked informative, the reply from the person that said others card are dying is a lie.

 

Saying that people are lying about other cards dying is, IMO, disingenuous (Not saying that you were).

 

The issue is a lot more complex than it seems, and I personally didn't like jayz2cents video (for many reasons) but saying he's lying about other cards dying is simply not true, plain and simple (I supplied many links that proves this).

 

Yes it's impacting mostly the FTW3, but other cards have died, and multiple other cards have had the bug but not died.

 

 

16 minutes ago, Vishera said:

Buildzoid speculated from the way those cards died,that the cards died from design flaws:

1.Overcurrent Protection isn't configured properly so it allows more current than the card can handle

2.VRMS - The amount of phases for MSVDD or memory power is not enough,so fuses or even the VRMs themselves blow.

Haven't seen his video yet, but overcurrent protection seems to be a valid hypothesis that I've seen explain a few times elsewhere.

[xAcid9]

7 hours ago, Video Beagle said:

Is that a good thing to do?

AFAIK, yes. 

 

+ Less power consumption

+ Less heat

+ No screen tearing

+ Smoother motion because no drop/skip frames

 

- Higher input latency

- Sometime can introduce judder when FPS drop below monitor refresh rate

[Moonzy]

7 hours ago, Video Beagle said:

Is that a good thing to do?

for this game? absolutely

for other games that doesnt kill your hardware? eeh, pros and cons

[Vishera]

3 hours ago, xAcid9 said:

- Higher input latency

That is one thing i am sensitive for,

I can really feel it.

[joaopt]

22 hours ago, wkdpaul said:

No, people have reported that other 3090 broke, the link I supplied proves so.

 

Again, if you're saying that people are lying, while they actually give proof they aren't, that's extremely disingenuous.

 

Here's a direct link to someone that has a fried 3090 that IS NOT a FTW3 ;

https://forums.newworld.com/t/issues-with-evga-rtx-3090-ftw3-ultra/112757/9

 

read carefully

 

https://www.igorslab.de/en/evga-geforce-rtx-3080-rtx-3090-and-not-only-new-world-when-the-graphics-card-goes-amok-because-of-design-failures/

 

i have nothing more to add

[grg994]

Here is a picture of a fried EVGA 3090 (https://www.guru3d.com/news-story/mmo-new-world-(closed-beta)-is-killing-geforce-rtx-3090-graphics-cards

 

The visible damage is in the where I marked on this pcb layout. That component should be a fuse on one of the 12V lines, connected directly after the rightmost 8pin.

 

Looks like Nvida/EVGA is redefining the term "a blown fuse"...

[Forbidden Wafer]

46 minutes ago, grg994 said:

Here is a picture of a fried EVGA 3090 (https://www.guru3d.com/news-story/mmo-new-world-(closed-beta)-is-killing-geforce-rtx-3090-graphics-cards

 

Looks like Nvida/EVGA is redefining the term "a blown fuse"...

Buildzoid said they at least protected the rest of the components by doing that. Other OEMs don't include the fuse and blow up the VRM/GPU die.

[Spotty]

49 minutes ago, grg994 said:

Here is a picture of a fried EVGA 3090 (https://www.guru3d.com/news-story/mmo-new-world-(closed-beta)-is-killing-geforce-rtx-3090-graphics-cards

-snip-
The visible damage is in the where I marked on this pcb layout. That component should be a fuse on one of the 12V lines, connected directly after the rightmost 8pin.

Yeah, that'll do it.

 

2 minutes ago, Forbidden Wafer said:

Buildzoid said they at least protected the rest of the components by doing that. Other OEMs don't include the fuse and blow up the VRM/GPU die.

Looking at that picture the board is dead regardless.

[Forbidden Wafer]

6 minutes ago, Spotty said:

Looking at that picture the board is dead regardless.

Not necessarily. Louis have shown several macbook mobos with gigantic holes blown up by Thunderbolt working. e.g.

 

[Moonzy]

12 minutes ago, Forbidden Wafer said:

Buildzoid said they at least protected the rest of the components by doing that.

True, but the PCB is probably a goner, and manufacturers don't really unsolder parts afaik

Maybe the GPU core? Idk, I'd wager that they'd just junk it after analysing it

[grg994]

17 minutes ago, Forbidden Wafer said:

Buildzoid said they at least protected the rest of the components by doing that. Other OEMs don't include the fuse and blow up the VRM/GPU die.

This is true, but I'm confused by seeing that:

 

Is this really how a popped fuse of that type should look like on a circuit board?

[Forbidden Wafer]

6 minutes ago, grg994 said:

Is this really how a popped fuse of that type should look like on a circuit board?

I'm not really sure, but I do think something went wrong other than the fuse. Maybe it didn't break the connection completely and created an arc that melted everything, or it got hot enough to melt the resin letting traces short (e.g. fuse with wrong rating). 

[Spotty]

20 minutes ago, grg994 said:

Is this really how a popped fuse of that type should look like on a circuit board?

No. If a fuse like that blows it might not even have visible signs of damage, you would use a multimeter to measure continuity across the fuse to check that it's still working. No continuity = fuse blown. It might also look damaged/burnt but no damage to the nearby board. Then you could easily desolder the blown fuse and replace it with a new one (though usually there's a problem elsewhere which caused the fuse to blow in the first place, so you would need to address that first).

My guess is on the card in that photo the fuse blew open but it just arced across the open fuse, causing the damage you see to the board.

Edited July 25, 2021 by Spotty

[CarlBar]

Whilst we have no idea yet if this is the failure mode for the majority of affected GPU's, if it is it raises all kinds of issues.

 

1. Why did the board get a hole blown in it. The voltage shouldn't go much above or below 12v since this is PSU feed, and arcing is voltage based. The only thing i can think of is a really odd fuse failure mode of some kind that made it blow up or only partially fail. But that then raises the question what it was and what triggered it.

 

2. How did the cards pull that much current. These fuses are generally rated with OC'ing in mind, i doubt everyone involved was raising the voltage on their cards, and without that the card shouldn't have been able to pull enough to blow the fuse. So WTH is happening here exactly.

[Bombastinator]

My worry is that if a computer program can be written that will physically damage a computer it will be.  Even if it’s ridiculously difficult and the damage is only slight.  Many years ago it was determined that it was impossible to physically hurt someone through the internet and the best that could be done was emotional hurt so people settled on hiding the most horrible images they could find inside innocuous things.

 

Of course deliberate misinformation hadn’t come up yet.    There’s always something worse.  If there is a way to damage a machine through code it won’t just be something that happens.  It will become something that is hard to avoid.

Edited July 25, 2021 by Bombastinator