Amazon's mmo New World is bricking 3090 gpus

[Edward78]

 

[TetraSky]

No they are not "exploding", it affects more than just Nvidia GPUs according to various reports and we got a tech news thread about this already

 

[Abdullah Bhutta]

Now they are not, because the Game has probably been patched.

 

[Chris Pratt]

You're about a week late, with a random YouTube video no one is going to watch and none of your own thoughts. What was the point of this? 

[Moonzy]

gAmEs dOn'T hUrT gPu LiEk MiNiNg

[WkdPaul]

@Edward78

 

Your thread was merged with the news thread.

[Mister Woof]

On 7/21/2021 at 3:13 PM, Yoinkerman said:

Ah the classic menu screen 9999 fps problem.  You'd think devs would know that, it's like the oldest trick in the book

I've got 9999 fps problems

 

but a New World ain't one of em.

 

Pass on that game anyway

[RejZoR]

And yet again showing that RTX 3000 series have several serious flaws in their designs, power delivery of AiB's, VRAM temperatures, and also drivers that allow such things. If they know card can implode because of too high framerate, that should be soft limited to I don't know 750fps in drivers. No one needs such framerate for gameplay, but it would be limited to prevent implosion.

[Kisai]

1 hour ago, Abdullah Bhutta said:

Now they are not, because the Game has probably been patched.

 

So the EVGA 3090 FTW card is literately experiencing a different problem from the rest. Given he said it's an early model, which might mean an early VBIOS as well. Wouldn't that be funny if the card he had was "tweaked" to give better scores on benchmarks as a review card. hitting 110% power while the GPU clock is at 52%.

 

Also brings up MLCC vs POSCAP's. The model he has very clearly has the same configuration as a FE model (4 POSCAP's), not the 6- POSCAP models the reviewers had before.

 

Eventually he got it up to 530w and didn't blow up the card.

 

At any rate, my suspicion, again is that those who "killed their card" probably had VSYNC off and/or a high-refresh monitor that is actually able to max out the GPU and somehow tripped defects in those cards. The FTW card itself appears to be exceeding the power by a rather large margin if it's allowed to, so that might be specific to the FTW card that allowed those cards to fail. 

 

I also suspect that EVGA will probably just repair cards if all it did was blow a cap on the card and then use those as warranty exchange inventory.

 

 

[Kisai]

2 minutes ago, RejZoR said:

And yet again showing that RTX 3000 series have several serious flaws in their designs, power delivery of AiB's, VRAM temperatures, and also drivers that allow such things. If they know card can implode because of too high framerate, that should be soft limited to I don't know 750fps in drivers. No one needs such framerate for gameplay, but it would be limited to prevent implosion.

They probably don't framelimit in the driver because of VR HMD's, (which could drive 240fps per eye)

[RejZoR]

20 minutes ago, Kisai said:

They probably don't framelimit in the driver because of VR HMD's, (which could drive 240fps per eye)

Still, something like 750fps as hard limit in driver is very much reasonable. Nothing requires such high framerate yet it would prevent this assuming it's actually the source of problem. Unless you have 4 eyes and you need more fps in VR

[poochyena]

2 minutes ago, RejZoR said:

something like 750fps as hard limit in driver is very much reasonable.

I can hear the CS:GO players running after you now

[Touch My Hamm]

21 hours ago, Mark Kaine said:

Everything looks like a voltage issue that insta fries cards, not enough  time to heat up the vram that much.

 

so the issue is most likely some voltage regulator, maybe in combination with not ideal drivers…

 

more nerfs through drivers incoming if so… 

Strange that the Nvidia drivers and a few layers of software would allow such a thing to happen. If this happens with somewhat finished product being a game you would think this would be common in development of these assets and testing. Strange its happening now lol. 

[Zando_]

20 hours ago, Bombastinator said:

This one disturbs me a lot.  If GPUs are over heatable despite throttle systems then CPUs could be too.  The argument for old school style stress tests and low temperatures is bolstered.  I’m pretty sure my equipment is immune to this because by modern standards it is overcooled.  The stuff can’t run so fast it melts itself.  

ASUS' first gen X99 boards beg to differ. Would randomly feed 2v or more into the CPU, were never fixed either (thus why you should avoid them even second hand). Even some pretty beefy water setups that was a death sentence for many 5960Xs in X99 Rampage Vs. I actually had a 5960X for a while that came out of a Rampage V, perfect visual condition but an absolute brick. Nice paperweight I guess. 

Stands to reason if modern boards derped similarly, they'd kill a CPU no matter the cooling solution. 

[wanderingfool2]

On 7/22/2021 at 3:18 PM, Kisai said:

No, it's pretty clearly the lack of frame-rate limiting. You know how people test synthetic benchmarks and never seem to catastrophically cause the video cards to fail, even if OC'd? The question is why the cards own thermal/power throttling isn't kicking in before this failure happens.

 

Just like why people shouldn't pick up pre-owned GPU's. You don't know how it's been abused, and this is one of those reasons which puts a cloud over the ability to resell them, since someone might sell a "defective" card that self-destructs and not even have the warranty.

If it was a frame-rate limiting issue, guess what...other games would have fried cards as well.  My hypothesis was that they were running an API call in the menu that isn't common/ever used for things such as menus...thus getting high fps and an API call that might add that extra tax on top of it all.

 

To blame it on a lack of frame-rate limiting though is not correct, as clearly other games can do it without blowing up the GPU.  It's like blaming a car crashes on high-speeds, when in reality what started the issue was a blown tire (and the high-speeds were just what made things worse)

[Bombastinator]

23 minutes ago, wanderingfool2 said:

If it was a frame-rate limiting issue, guess what...other games would have fried cards as well.  My hypothesis was that they were running an API call in the menu that isn't common/ever used for things such as menus...thus getting high fps and an API call that might add that extra tax on top of it all.

 

To blame it on a lack of frame-rate limiting though is not correct, as clearly other games can do it without blowing up the GPU.  It's like blaming a car crashes on high-speeds, when in reality what started the issue was a blown tire (and the high-speeds were just what made things worse)

I expect the actual cause will be looked into.  It’s the kind of thing that is discoverable and all this supposition can be put to rest. That it effects only one maker and one model implies a problem with the maker of that model. That’s only implication though.  I think it likely Someone is going to crawl through things with a fine tooth comb, pull out the wriggling thing, and present it for all to see, and then we will know.

[xAcid9]

4 hours ago, wanderingfool2 said:

If it was a frame-rate limiting issue, guess what...other games would have fried cards as well.

I had 2 cards that will coil-whine when FPS reached over 1000. Maybe VRM is more stress when your FPS is over thousands?

[Video Beagle]

Is there a relation of frame rate a card will produce to what the monitor it's attached to will display? (I don't even know if this is a question that makes sense).  My monitor is 75hz. Will my card (1070ti) produce higher fps than 75?  (I assume 75 fps is the highest I can see).

[Spotty]

6 minutes ago, Video Beagle said:

Is there a relation of frame rate a card will produce to what the monitor it's attached to will display? (I don't even know if this is a question that makes sense).  My monitor is 75hz. Will my card (1070ti) produce higher fps than 75?  (I assume 75 fps is the highest I can see).

Yes, the framerate when playing games can be higher than your monitors refresh rate. However, if you enable Vsync then your framerate will be limited to your monitors refresh rate, in your case 75FPS. Vsync will lock the systems max framerate to the refresh rate of your monitor.

[Video Beagle]

20 minutes ago, Spotty said:

Vsync will lock the systems max framerate to the refresh rate of your monitor.

Is that a good thing to do?

[Kisai]

6 hours ago, wanderingfool2 said:

If it was a frame-rate limiting issue, guess what...other games would have fried cards as well.  My hypothesis was that they were running an API call in the menu that isn't common/ever used for things such as menus...thus getting high fps and an API call that might add that extra tax on top of it all.

 

To blame it on a lack of frame-rate limiting though is not correct, as clearly other games can do it without blowing up the GPU.  It's like blaming a car crashes on high-speeds, when in reality what started the issue was a blown tire (and the high-speeds were just what made things worse)

 

I think you misunderstood something here.

 

There are three ways to control the GPU processing power:

- VSYNC - Which limits the frame rate to in-step with the monitor refresh. GSYNC permits this to work at the monitor refresh rate and below. If you turn Vsync off, then you are expressly telling the game to min-max the GPU redraw rate.

- Frame Limiting, which sets a hard ceiling, not a floor. It's not the same as Vsync. If the game's internal physics or communications probably operates at 15fps since it's a MMORPG, but I don't have the game, and and not interested in it. If the game permits you to set a ceiling, that's to keep the GPU from operating at the monitor refresh rate, as many games simply do not work, especially those built on Unity, and those built on the Creation engine used by Fallout4/Skyrim/etc, as their physics and collision systems don't work correctly.

- Frame Skipping, which sets a floor. "This game must be responsive and must achieve minimally 30fps, adjust these knobs if not achieved"

 

There is no secret "set GPU on fire" API.

 

I'm sure other games have fried cards, but they just chalked it up to a bad card. In Alpha and Beta tests of games you have the ears of the developer, where in most retail games, especially single player ones, you do not.

 

 

Like I said, this stuff happens, and it's just rarely reported on, and always hardware when it is.

 

Even on this forum

Guess which manufacturer's GPU's catch fire? EVGA and MSI.

 

Like it did not surprise me, at all, that the EVGA card's were actually dying. I'm pretty sure you will find that anyone who actually had a card fail, likely had the exact same EVGA card, and everyone else who just had "black screen" problems likely had a factory OC model of any manufacturer.

 

Because that's the trend. Every time a new game comes out that people start complaining about (the last one was Cyberpunk) how the game is unoptimized or some other reason to deflect the fact they they either OC'd their GPU or bought a factory OC'd model and it failed because the new game uses the GPU at a higher level than the games that were released the year they bought their GPU.

 

Many games you play today, are DX11 games. Not DX12, not Vulkan. "Amazon Lumberyard" is a fork of Cryengine. What's Cryengine known for doing? Breaking on AMD cards. 

 

At any rate, no I don't believe the game engine is doing anything special to kill GPU's, and if you want to test it yourself, you can download Lumberyard and build a demo game and try it yourself.

 

https://docs.aws.amazon.com/lumberyard/latest/releasenotes/lumberyard-v1.28-known-issues.html

 

The current version is from May of this year.

 

[joaopt]

On 7/22/2021 at 10:22 AM, That Franc said:

There's a fresh update from Jay:

 

This was a lie, apart from that tweet you can't find any broken cards besides the evga ftw3. I guess it was just intended to divert attentions.

[WkdPaul]

3 minutes ago, joaopt said:

This was a lie, apart from that tweet you can't find any broken cards besides the evga ftw3. I guess it was just intended to divert attentions.

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.

[Bombastinator]

4 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.

So the it only being evga is not true, they’re just fixing stuff anyway.  Is there a pattern to which cards then?

[WkdPaul]

4 minutes ago, Bombastinator said:

So the it only being evga is not true, they’re just fixing stuff anyway.  Is there a pattern to which cards then?

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).