Info [PSA] Amazon's new game reportedly causing GPUs to brick

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Leeea

Diamond Member
Apr 3, 2020
3,599
5,340
106
So if I mine in it 24/7 without changing anything, is it still defective if it dies?

I would say yes.

It should hold up at least three years with that. Even when the fans do die, it should be smart enough to throttle itself back and not die.

This is nothing new, over heat throttling has been around for many years now.
 
  • Like
Reactions: lightmanek

Iron Woode

Elite Member
Super Moderator
Oct 10, 1999
30,854
12,353
136
Update: In a statement sent to PC Gamer today, Amazon says that it has observed "no indication of widespread issues" with RTX 3090s, despite "a few" reports of players who say they've experienced hardware failures. Although the game is already safe to play, says Amazon, it will release a patch that caps fps on the menu screen to "reassure" players. Here's the full statement:
Hundreds of thousands of people played in the New World Closed Beta yesterday, with millions of total hours played. We’ve received a few reports of players using high-performance graphics cards experiencing hardware failure when playing New World.
New World makes standard DirectX calls as provided by the Windows API. We have seen no indication of widespread issues with 3090s, either in the beta or during our many months of alpha testing.

The New World Closed Beta is safe to play. In order to further reassure players, we will implement a patch today that caps frames per second on our menu screen. We’re grateful for the support New World is receiving from players around the world, and will keep listening to their feedback throughout Beta and beyond.

Earlier in the day, an Amazon customer service rep acknowledged an issue with "Nvidia RTX 3090 Series & 100% GPU Usage" in a forum post. The developer recommends turning off "the overrides in the driver settings," but the only specific instruction given is to set 'Max Frame Rate' to 'Off' in the Nvidia control panel (its default state) and rely on the in-game frame rate limiting option.
 
  • Like
Reactions: Leeea

Hitman928

Diamond Member
Apr 15, 2012
5,177
7,628
136
Interesting take.

I don't agree with it, of course.

If a person could do anything they wanted to hardware and still get it warranted, things like heavily overclocking CPUs with dry ice and plotting chia on SSDs would also be covered. You'd hate to own the company that had to cover these kind of losses (kind of like the old "lifetime" electronic rent-a-center abuse Costco dealt with before changing it to 90 days)

Overclocking violates the terms of warranties though (at least all the ones I've ever had to use). There's no wording restricting compute loads from being run 24/7 (unless they've added this recently) and people have been doing very mining-esque loads for 24/7 (or very extended periods of time) for over a decade on consumer GPUs. The cards will (or at least should) be validated for a mean time to failure which should extend beyond the warranty period with some decent margin even when running 24/7. If not, then the manufacturer is not doing a good job of validation and the extra warranty claims are on them.

Chia mining may be a different beast, I'm not as familiar with SSD warranties as I've never had to use one but I'd be surprised if there isn't some kind of language in there about excessive writes voiding the warranty since SSDs have a fairly well known limitation in how many writes they can execute in their lifetimes.
 

Leeea

Diamond Member
Apr 3, 2020
3,599
5,340
106
I am thinking idiots built the system wrong with improper cooling or used a POS power supply, or its a OEM of HP or something.
We have every indication these systems have good power supplies and good enough cooling.

I do know the rear ram on the 3090 gets insanely hot.
Too hot sometimes ive seen it spike 80C before i loaded the rear backplate with heatsinks and have a dedicated fan to it now.

That there is what we call a hardware defect. If the manufacturer failed to cool the components of the card with the manufacturer supplied heat sink, the design is defective.

If the user is supposed to supply their own heatsink, that fact needs to be documented by the manufacturer. There is no documentation from anyone indicating this is the case.
 

Leeea

Diamond Member
Apr 3, 2020
3,599
5,340
106
I have nothing to say honestly...
If you don't want to tune its on you.

and if the user does tune it frequently voids the warrantee when said user alters cooling systems/heatsink, which is also on the user
 

Furious_Styles

Senior member
Jan 17, 2019
492
228
116
Overclocking violates the terms of warranties though (at least all the ones I've ever had to use). There's no wording restricting compute loads from being run 24/7 (unless they've added this recently) and people have been doing very mining-esque loads for 24/7 (or very extended periods of time) for over a decade on consumer GPUs. The cards will (or at least should) be validated for a mean time to failure which should extend beyond the warranty period with some decent margin even when running 24/7. If not, then the manufacturer is not doing a good job of validation and the extra warranty claims are on them.

Chia mining may be a different beast, I'm not as familiar with SSD warranties as I've never had to use one but I'd be surprised if there isn't some kind of language in there about excessive writes voiding the warranty since SSDs have a fairly well known limitation in how many writes they can execute in their lifetimes.
A lot of manufacturers specifically say you void the warranty by mining.
 

Mopetar

Diamond Member
Jan 31, 2011
7,797
5,899
136
If a person could do anything they wanted to hardware and still get it warranted, things like heavily overclocking CPUs with dry ice and plotting chia on SSDs would also be covered.

If you just run the card with factory settings for 24/7 and it dies why shouldn't a warranty cover it? Most warranties specifically include language indicating that they don't cover use modifications, use outside of specified parameters, etc. but I don't think I've seen any that say you can't use the GPU more than X hours per day.
 
  • Like
Reactions: Leeea

Hitman928

Diamond Member
Apr 15, 2012
5,177
7,628
136
A lot of manufacturers specifically say you void the warranty by mining.

They must have added after people were overclocking their GPUs to the extreme during the BTC mining craze (when GPUs were still viable). Other mining algos don't benefit from overclocking so they shouldn't cause failures like people overclocking for BTC would. If your particular brand prohibits mining then it's on you (if you are honest) and it breaks. Probably some if not all of the factory overclocked models are more prone to failure as well due to being factory overclocked and they know this so they say the warranty is void if you do any non-gaming type of work because they don't want to be responsible for their own lifetime shortening overclocks. Caveat emptor and make sure you know the warranty limits you are buying into (which will also vary depending on region sold).
 

Leeea

Diamond Member
Apr 3, 2020
3,599
5,340
106
VRAM... not VRegs..
And to be exact the REAR side on card...

Simple fix... ADD surface area to backplate... put active fan on it..
That is a great fix.

In order for this not to be a manufacturer defect, the manufacturer needs to disclose to the user they need to install said heatsinks at time of purchase

And that is why the manufactor will accept the RMA.
But again, it something which could of been avoided had you tuned.

Here is a FE Fix for you guys as well without voiding your warrenty.
That is really clever. I was figuring anything added would void the warranty, but it looks like those can pop right off. Nice design.

can't blame a game for killing your GPU because it tax'd it.
I agree 100%.

YOU:
"You can't blame a game for killing your GPU because it tax'd it."

One of the stranger statements ever made here.
I agree with him though. A game should never be able to brick hardware.

Hopefully they will figure it out quickly, and fix whatever the issue is. I personally believe if it was a true hardware defect issue, this sort of thing would have popped up all over the internet by now since the cards have been out for a while, and they definitely have been pushed hard by all the miners. It could end up being something like just some bad batches of cards that used problematic components (much like the bad capacitor issues on motherboards in the past).
But miners do not really push there cards in the same way. Mining loads are steady, constant, and not at all transient. They also likely do not use the same circuitry in the same way as the menu system.

Maybe NVidia will be in for another class-action, for designing/building/selling video cards with GDDR6X memory running at nearly 100C under normal load. Sounds ridiculous to me.
I remember back when Nvidia was sued for the laptop GPUs that melted off the board and in a few cases completely out of the laptop.


I have heard this may happen on some 6900XTs as well?
Ekk! This is starting to get close to home! Any other news on that?


If a person could do anything they wanted to hardware and still get it warranted, things like heavily overclocking CPUs with dry ice and plotting chia on SSDs would also be covered. You'd hate to own the company that had to cover these kind of losses (kind of like the old "lifetime" electronic rent-a-center abuse Costco dealt with before changing it to 90 days)
Dry ice is clearly an alteration of the manufacturer's cooling system, and clearly voids the warrantee. Overclocking is clearly an alteration of the designed or recommended specifications, and if said over clocking is done with non-first party tools, clearly voids the warrantee.

Plotting chia likely exceeds the writes on the SSD (which the manufacturer discloses to you on sale), and would not be covered by warrantee. However, if a user plotting chia had a SSD die before its write limit, legit warranty claim.
 

Stuka87

Diamond Member
Dec 10, 2010
6,240
2,559
136
I didn't read the first three pages of this thread. But this issue basically reminds me of when the 3090's first launched and they were bricking/crashing, so nVidia release a 'patch' to detune them a bit. Everything about the 3080, and especially the 3090 suggest nVidia clocked them too high, and the power draw is causing issues that typically were never issues. Like having transient current spikes that result in power consumption hitting 600-800W that trigger the current protection on PSUs.

While having an upcapped FPS is stupid on Amazon's part, but the video card should not destroy itself trying to run as fast as it can. In previous non-Turing cards, the worst we (as users) would experience is a whining sound. Which is annoying, but would never damage a card. But when said card draws 450-500W, any component that is right on the edge, risks outright failure.

So while Amazon may be the catalyst for these issues, nVidia should take most of the blame.
 
Last edited:

Furious_Styles

Senior member
Jan 17, 2019
492
228
116
I didn't read the first three pages of this thread. But this issue basically reminds me of when the 3090's first launched and they were bricking/crashing, so nVidia release a 'patch' to detune them a bit. Everything about the 3080, and especially the 3090 suggest nVidia clocked them too high, and the power draw is causing issues that typically were never issues. Like having transient current spikes that result in power consumption hitting 600-800W that trigger the current protection on PSUs.

While having an upcapped FPS is stupid on Amazon's part, it video card should not destroy itself trying to do run as fast as it could. In previous non-Turing cards, the worst we (as users) would experience is a whining sound. Which is annoying, but would never damage a card. But when said card draws 450-500W, any component that is right on the edge, risks outright failure.

So while Amazon may be the catalyst for these issues, nVidia should take most of the blame.

That just was a driver bug where it was trying to boost core clocks too high. You can still pretty much do it with any card by going into AB and trying to put your core clock boost way too high.
 
  • Like
Reactions: Leeea

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
Personally I find it fascinating from an electrical PoV, though I guess people affected don't see it the same way, especially considering the state of the market.

Definitely. Especially if you bought a card from a scalper on eBay at a ridiculous price.

Interesting take.

I don't agree with it, of course.

Mining is nothing but a GPGPU algorithm. Plain and simple. There's nothing magical about it. Ethereum mining in particular doesn't even utilize all of a GPU's resources, so much so that the Claymore author figured out how to mine another coin (DCR) at the same time that used different resources on the card with little effect on Ethereum hashrate. It's not at all clear that what Amazon's game inflicts upon a dGPU when rendering the menu screen at unlimited framerates, or how that is even related to mining. Nor is it clear WHY unlimited framerates at a menu would cause any problems whatsoever.
 
Last edited:
  • Like
Reactions: Leeea

GodisanAtheist

Diamond Member
Nov 16, 2006
6,716
7,004
136
I didn't read the first three pages of this thread. But this issue basically reminds me of when the 3090's first launched and they were bricking/crashing, so nVidia release a 'patch' to detune them a bit. Everything about the 3080, and especially the 3090 suggest nVidia clocked them too high, and the power draw is causing issues that typically were never issues. Like having transient current spikes that result in power consumption hitting 600-800W that trigger the current protection on PSUs.

While having an upcapped FPS is stupid on Amazon's part, it video card should not destroy itself trying to do run as fast as it could. In previous non-Turing cards, the worst we (as users) would experience is a whining sound. Which is annoying, but would never damage a card. But when said card draws 450-500W, any component that is right on the edge, risks outright failure.

So while Amazon may be the catalyst for these issues, nVidia should take most of the blame.

- I wonder if this entire gen is going to end up having a "cursed" reputations given that so many cards were built during supply shortages that manufacturers relaxed their standards on the quality of components in order to get goods out there combined with the fact that these components (especially NV this round) are basically pushed to the hilt (far less margin of error on crappy parts).

Throw mining abuse in the mix as well (in terms of mechanical wear on the fan bearings) and hey baby, you got yourself a strew going.

Will be interesting to see the longevity of these cards play out.
 
  • Like
Reactions: Stuka87 and Leeea

CP5670

Diamond Member
Jun 24, 2004
5,508
586
126
I didn't read the first three pages of this thread. But this issue basically reminds me of when the 3090's first launched and they were bricking/crashing, so nVidia release a 'patch' to detune them a bit. Everything about the 3080, and especially the 3090 suggest nVidia clocked them too high, and the power draw is causing issues that typically were never issues. Like having transient current spikes that result in power consumption hitting 600-800W that trigger the current protection on PSUs.

The cards are actually capable of running at significantly higher clocks though, and benefit from it. The problem is the power usage and power limits. They just use way too much power beyond 2ghz. I think the 6800XT/6900XT clock higher because of their lower power usage, and are faster at lower resolutions because of it.
 

Mopetar

Diamond Member
Jan 31, 2011
7,797
5,899
136

Mopetar

Diamond Member
Jan 31, 2011
7,797
5,899
136
I can totally see some users stress-testing their builds in the future by running Prime95, Memtest86, OCCT, FurMark, and finally the toughest one yet: Sitting in the menu system in New World. :p

I think we need to make something even tougher on the GPUs. Using what we've learned from this as long as we don't cap frame rates a perfectly black screen should have the theoretically highest FPS possible. Of course it will be impossible to tell if it's actually bricked the card or not or even at which point this has occurred, but I suppose we can always hope that the test proves so intense that the GPU actually catches fire.
 

Shmee

Memory & Storage, Graphics Cards Mod Elite Member
Super Moderator
Sep 13, 2008
7,381
2,415
146
I know someone who had a Strix 3090 basically become defective after playing the New World game. Hopefully Asus does not give him trouble with his RMA.
 
  • Wow
Reactions: Leeea and Stuka87

DAPUNISHER

Super Moderator CPU Forum Mod and Elite Member
Super Moderator
Aug 22, 2001
28,272
19,912
146
Jay did a follow up vid -


TL;DW -They patched in the frame limiter already. He tried to make his FTW 3 blow up playing the game, and failed. Including using a custom bios with normal power limits removed. And using the verboten pig tailed power connector. The card did exhibit higher power draw v. the MSI 3090 under the same test conditions. It should also be pointed out this is an open air test bed.

I share his position on EVGA; I have had the opposite of bad experiences with them over the years. That goes for both their products and services. That said, there was a strong EVGA damage control vibe in this vid. He admits he is a bit of a fanboy at the end. Do with that info what you will.
 
Last edited:

CP5670

Diamond Member
Jun 24, 2004
5,508
586
126
Maybe this game pushes the 400-500W AIB cards to the limit, beyond the reference or FE cards.

The most intensive games I've played are Cyberpunk and Metro Exodus with RT. The latter runs at noticeably lower clocks than anything else, probably because it hits the power limit. RT games in general are more power hungry than anything non RT.
 

aigomorla

CPU, Cases&Cooling Mod PC Gaming Mod Elite Member
Super Moderator
Sep 28, 2005
20,839
3,174
126
  • Like
Reactions: Leeea