New World (Closed Beta), an MMO, Found Bricking GeForce RTX 3090 Graphics Cards

[InVasMani]

Do closed betas usually wreck your PC physically?

Did the game wreck the PC or was the underlying hardware faulty in the first place and the game exposed it!!? Frankly if the game did manage to brick the hardware that is a problem itself that the hardware needs to do better to prevent. A game itself shouldn't even be possible to brick the hardware and if it can that's something GPU makers need to rethink and prevent in hardware. I mean I've seen people test CPU's w/o heatsinks on briefly and not fry them if a GPU is frying from a game menu with fans and heatsink attached that's a huge design error of the hardware. Even if the the GPU fan fails it shouldn't outright fry the GPU should throttle to prevent damage.

I can fully understand overclocking software physically altering settings bricking the hardware by pushing it too hard and something breaking, but software operating like normal just trying to render graphics shouldn't be a causing a GPU failure as a rule of thumb. This scenario that allegedly is breaking the GPU sounds more probably to cause a catastrophic CPU failure as well if anything at 9000FPS that poor CPU is getting a workout.

Just to reiterate unless the game overclocked the GPU in some way I have strong doubts that the game caused the hardware to fail rather than the hardware failing because it was over stressed from being pushed too hard while being near the breaking point in the first place.

 

[Auer]

I get the feeling Amazon has already moved on and I very much doubt they'll ever offer any kind of meaningful explanation at all.
EVGA might not offer any insight either. (or any other boardpartner for that matter)

 

[Mussels]

Game menu: didnt cap FPS, cause user chose to disable Vsync (common)
Some hardware: poof cause people had hidden issues in the systems

people: DeLiBerAtE MUUUUUrrrrDeeeeerrrr!

 

[InVasMani]

Probably just a EMP attack by aliens at Jeff Bezos empire of evil giving SETI that long awaited response back.

 

[Jism]

Proberly a fuse blowing out. Simple to check with a multimeter.

Look for these type of things. They blow out when the current of card exceeds what they are designed for.

 

[medi01]

That's a lovely title for "nVidia's 3090 can easily overheat into death" news.

Anyone care to make a TL;DR?

NV cheaped out on components of a premium card, amazon is to blame.

 

[efikkan]

Did the game wreck the PC or was the underlying hardware faulty in the first place and the game exposed it!!? Frankly if the game did manage to brick the hardware that is a problem itself that the hardware needs to do better to prevent. A game itself shouldn't even be possible to brick the hardware and if it can that's something GPU makers need to rethink and prevent in hardware. I mean I've seen people test CPU's w/o heatsinks on briefly and not fry them if a GPU is frying from a game menu with fans and heatsink attached that's a huge design error of the hardware. Even if the the GPU fan fails it shouldn't outright fry the GPU should throttle to prevent damage.

I can fully understand overclocking software physically altering settings bricking the hardware by pushing it too hard and something breaking, but software operating like normal just trying to render graphics shouldn't be a causing a GPU failure as a rule of thumb. This scenario that allegedly is breaking the GPU sounds more probably to cause a catastrophic CPU failure as well if anything at 9000FPS that poor CPU is getting a workout.

To be clear, high FPS can't damage a card. High FPS in the thousands will be CPU limited before anything else. Also several have reported deaths ingame, so that theory is bust anyway.

While it is theoretically possible that a game can trigger a bug in firmware or hardware which leads to damaged hardware, but this is extremely rare.
It is far more likely that either many of these cards have been overclocked or that some cards may have weak VRM designs, and for that reason be close to a breaking point. I would like to see the statistics of how many of these cards have been overclocked, even in the Reddit thread quite few of them have modified BIOSes. People tend to be quite superstitious when it comes to finding causes without sufficient evidence, easily blaming the last thing they changed on the machine.
Also remember that with hundreds of thousands of players, we should expect a few cards to die daily, so we need to separate the actual "problem" from the background noise.

I do find it interesting that so many of the repored cases are the same model; EVGA RTX 3090 FTW3. Who knows, there could be a couple bad batches of these.

 

[basco]

so 3dmark2001 with its 999 fps should have killed a lot of cards-lol
on computerbase they say it could be a special fan controller on the evga cards.

 

[kane nas]

For 3090 ftw3 the maximum power limit is 107%,the most likely scenario is driver, game and hardware failure

 

[Mussels]

For 3090 ftw3 the maximum power limit is 107%,the most likely scenario is driver, game and hardware failure

i thought those numbers seemed absurd, then i saw jayzs video - and yeah, thats with a modded unlimited wattage Vbios (and probably a high wattage intel system)

that video shows how nuts an unleashed 3090 can get tho

 

[Ravenas]

Do closed betas usually wreck your PC physically?

It appears you didn't read my reply.

 

[Mussels]

Jayz video is onto the cause i believe:

MSI: 99-100% power usage ~370W
EVGA: 119% 410W (Gah, 530W with the unlocked BIOS)

I can see this being whats happening, chewing too much power and burning cables/GPU parts
Also, Jay spotted the issue got worse at 4K... so did EVGA assign more power to the VRAM and its going over the internal limits as more VRAM is used? What did they change to over-ride Nvidias VBIOS limits?

Is this also people who flashed unlocked BIOSes to their cards, and are covering it up to claim warranty?

 

[Auer]

Jayz video is onto the cause i believe:

MSI: 99-100% power usage ~370W
EVGA: 119% 410W (Gah, 530W with the unlocked BIOS)

I can see this being whats happening, chewing too much power and burning cables/GPU parts
Also, Jay spotted the issue got worse at 4K... so did EVGA assign more power to the VRAM and its going over the internal limits as more VRAM is used? What did they change to over-ride Nvidias VBIOS limits?

Is this also people who flashed unlocked BIOSes to their cards, and are covering it up to claim warranty?

I don't think EVGA denies you an RMA based on the bios you use, They have released 500W+ bioses themselves.
I could be wrong of course, but ppl put their high end EVGA GPU's thru all kinds of stuff with mostly EVGA's blessing.

 

[Mussels]

I don't think EVGA denies you an RMA based on the bios you use, They have released 500W+ bioses themselves.
I could be wrong of course, but ppl put their high end EVGA GPU's thru all kinds of stuff with mostly EVGA's blessing.

If my stock 350W BIOS melted PSU cables, whats a 530W gunna do?

Just because the card can theoretically handle it, doesnt mean the PSU and cables arent gunna have a seizure and send out of spec voltages in

 

[Aquinus]

If my stock 350W BIOS melted PSU cables, whats a 530W gunna do?

Just because the card can theoretically handle it, doesnt mean the PSU and cables arent gunna have a seizure and send out of spec voltages in

Imagine the stress and instantaneous current on the VRMs when the GPU is running at 500w and OVP gets triggered via a shunt to protect the GPU. Anything already hot is going to get a whole lot hotter.

 

[Auer]

If my stock 350W BIOS melted PSU cables, whats a 530W gunna do?

Just because the card can theoretically handle it, doesnt mean the PSU and cables arent gunna have a seizure and send out of spec voltages in

You gotta admit tho, it's kinda laughable that the current benchmark undisputed champ GPU is getting fried by a MMO menu screen.
But here we are, and EVGA is already doing RMA's, like they always have done.
And Amazon patched a problem they didnt have. So maybe it's all over and done with.

 

[efikkan]

Jayz video is onto the cause i believe:

MSI: 99-100% power usage ~370W
EVGA: 119% 410W (Gah, 530W with the unlocked BIOS)

I can see this being whats happening, chewing too much power and burning cables/GPU parts

At 14:57 he tested another benchmark, and got the exact same behavior.
What he is showcasing is a card with an abnormal behavior. Regardless of this being caused by faulty sensors or an actual higher power draw, it's still a hardware defect. If it's the latter (which is most likely), it may cause damage over time, and lead to hardware failure at some point.

Games don't manage power, this is managed by the GPU itself. No game is to blame here.

 

[Ravenas]

At 14:57 he tested another benchmark, and got the exact same behavior.
What he is showcasing is a card with an abnormal behavior. Regardless of this being caused by faulty sensors or an actual higher power draw, it's still a hardware defect. If it's the latter (which is most likely), it may cause damage over time, and lead to hardware failure at some point.

Games don't manage power, this is managed by the GPU itself. No game is to blame here.

The power management relies on the game for exceution. If no game was being played, no overlimit would have been reached.

 

[ThrashZone]

I don't think EVGA denies you an RMA based on the bios you use, They have released 500W+ bioses themselves.
I could be wrong of course, but ppl put their high end EVGA GPU's thru all kinds of stuff with mostly EVGA's blessing.

Hi,
Depends on which vbios you altered
ftw3 usually has dual vbios and only the oc vbios are you allowed to play with.
Normal vbios can only be flashed with evga updates.

 

[Aquinus]

The power management relies on the game for exceution. If no game was being played, no overlimit would have been reached.

No power limit would also have been reached if you never turned the computer on.

The point is that a GPU should be able to handle 100% load conditions safely regardless of how that situation arises. The bottom line is that shoddy hardware run full tilt is going to experience issues regardless of if that load is from Furmark, a game menu, or regular gameplay.

 

[Ravenas]

No power limit would also have been reached if you never turned the computer on.

The point is that a GPU should be able to handle 100% load conditions safely regardless of how that situation arises. The bottom line is that shoddy hardware run full tilt is going to experience issues.

Stating the game has nothing to do with the power management is false. I never said the power management wasn't to blame, you just can't say that the game has nothing to do with it. Why didn't any other games cause this issue over the past since 2020 release?

 

[Aquinus]

Stating the game has nothing to do with the power management is false. I never said the power management wasn't to blame, you just can't say that the game has nothing to do with it. Why didn't any other games cause this issue over the past since 2020 release?

It doesn't. The GPU is responsible for figuring out how to scale clocks to keep power consumption in check. The game only applies load to the GPU, it does not tell it how to run. The GPU responds to load placed on it.

As for why some games cause this, certain parts of the GPU are stressed differently depending the the case. A game menu is rendering a lot of frames and relatively few triangles/polygons. It's pushing a ton of pixels.

 

[Ravenas]

It doesn't. The GPU is responsible for figuring out how to scale clocks to keep power consumption in check. The game only applies load to the GPU, it does not tell it how to run. The GPU responds to load placed on it.

As for why some games cause this, certain parts of the GPU are stressed differently depending the the case. A game menu is rendering a lot of frames and relatively few triangles/polygons. It's pushing a ton of pixels.

The game and the power management are independent and dependent variables. They are correlated. I’m not arguing your explanation of power management.

The statement “some games” is now extending this problem which has only been widespread reported for this single closed beta game.

 

[efikkan]

The power management relies on the game for exceution. If no game was being played, no overlimit would have been reached.

Stating the game has nothing to do with the power management is false. I never said the power management wasn't to blame, you just can't say that the game has nothing to do with it.

You are dead wrong on this one.
Power management is done in hardware, and isn't even aware of the game.

Why didn't any other games cause this issue over the past since 2020 release?

No one has proven any kind of correlation or causation yet, people are just jumping to conclusions.

It's not even possible for a game to cause damage by itself. The game only executes generic DX/OpenGL/Vulkan API calls to the driver.
A game can only cause indirect damage, but only if there is an underlying hardware defect, which is extremely rare.
It's far more likely that some cards have a defect where they break after some time with load, it happens all the time, but now a game is blamed just because a few cards broke with an alpha/beta version.

 

[Ravenas]

You are dead wrong on this one.
Power management is done in hardware, and isn't even aware of the game.


No one has proven any kind of correlation or causation yet, people are just jumping to conclusions.

It's not even possible for a game to cause damage by itself. The game only executes generic DX/OpenGL/Vulkan API calls to the driver.
A game can only cause indirect damage, but only if there is an underlying hardware defect, which is extremely rare.
It's far more likely that some cards have a defect where they break after some time with load, it happens all the time, but now a game is blamed just because a few cards broke with an alpha/beta version.

The game and gpu are independent and dependent variables, and that's all I have stated. Explaining how power management works has nothing to do with that statement. The game itself is absolutely in question due to it causing widespread reports of GPU damage. When did any of this happen prior to this during late 2020 and 2021 YTD?