Amazon's New World is bricking RTX 3090 graphics cards

[The red spirit]

Furmark GREATLY downclocks the card massively.
At 400W Power limit, a 3090 or 3080 card that would normally run at 1920 mhz would run at 1300 mhz in furmark.

You can destroy the card if it doesn't.
Proof? Try renaming furmark.exe to Quake3.exe or UnrealTournament.exe
Don't blame me for what happens.

At least Radeon cards (GCN era), have TDP, small power limit and maximum power limit. They try not to deviate from set wattage, even if there are spikes they are very fast and averages out at no more than power limit. There's also AMP limiter too. I can't really imagine any workload that could bypass so many protections. That's just vBIOS. There probably are some hardware protections too. I modify my RX 580's BIOS a lot (haven't modified for more wattage yet) and there are lots of things to tweak (and fuck up). So no, it isn't (or at least wasn't) Furmark specifically that killed cards, cards are made to not allow any workloads that demand more power than manufacturer sets. Setting wattage limit high, let's card run at full boost speed in Furmark and let's card consume far more power than it does in any game. But then again, same effect can be observed in Unigine tests and likely other synthetic benches.

 

[silentbogo]

Furmark GREATLY downclocks the card massively.
At 400W Power limit, a 3090 or 3080 card that would normally run at 1920 mhz would run at 1300 mhz in furmark.

You can destroy the card if it doesn't.
Proof? Try renaming furmark.exe to Quake3.exe or UnrealTournament.exe
Don't blame me for what happens.

Few issues with your argument:
1) It's not Catalyst from 13 years ago. Renaming Furmark won't make a difference. Just in case I'm too old and not keeping up with modern things, I've decided to run the latest v1.26 on my 1070Ti, and.... [drumroll] .... no difference whatsoever. It's not Furmark that downclocks your card, it's your card that downclocks your card for one reason or the other.
2) I think you misunderstand how Nvidia boost works. Just because you have 1900+MHz in one application doesn't mean you'll get the same boost in another app. More so... unless you are doing strictly manual OC with tightly controlled environment and decent pre-warmup, there's no guarantee that you'll get same clocks in the same app during consecutive runs. Boost 4.0 is so temperature-dependent that you may get a few MHz drop just by farting in a room with your PC.
3) I'm neither rich or lucky enough to torture 3090 personally, but I can find at least few screens|videos of RTX3090 running furmark at more than 1300MHz even at 400W PL.

Spoiler

4) Core boost clock depends on many factors, which include temps, power, voltage, and consequentially load type(intensity of load). Same reason why a stock 9700K won't boost as high in Linpack as in CPU-Z stress test.

Once again, it's not the software problem, it's a hardware problem. Could be bad VRM, could be defects in a PCB just like in EVGA's 10-series, but in either case it's all on EVGA. @buildzoid did a video not too long ago about this crap, there are some more details to this occurrence.

 

[Mussels]

Think of furmark and game menus as running the engine in first gear.
The RPM is up, heat output is up, but you're really not doing anything.

Throw it into 5th while carrying a heavy load, and the RPM is lower, but doing more.
It's just an unexpected way of loading the cards that shows up design flaws, looks like EVGA tweaked the values to win the performance charts, and its causing woopsies

 

[The red spirit]

It's just an unexpected way of loading the cards that shows up design flaws, looks like EVGA tweaked the values to win the performance charts, and its causing woopsies

In 2021, does anyone even care about card vendors anymore? Besides cooler design, VRMs and vBIOS, there's no real big differences between vendors. It's not anymore 3D accelerator era, when sometimes some vendors actually went bonkers and did something crazy. You can't even make claims about warranty and support anymore as basically everyone tries to reduce those expenses as much as they can, so thy all suck.

 

[maxfly]

So basically we still have no idea what was/is causing these cards to die?
What do we know other than EVGA has agreed to replace their 3090s asap and apparently heat isnt the primary factor in killing them?

 

[Shrek]

There is talk of a pop; which might leave some visable damage.

(3) Beware! This game could brick your GPU! WARNING! - YouTube

1:57

 

[Diverge]

The cards are to blame. Properly designed hardware adds failsafe's to prevent damaging hardware. Things are kept in check by both circuitry and firmware of devices. This applies to pretty much all electronics that have some sort of brains. EVGA knows this, and it's why they took responsibility. Depending on game developers to limit excessive frame rates in menus is not a long term solution and just a temporary Band-Aid.

 

[maxfly]

The cards are to blame. Properly designed hardware adds failsafe's to prevent damaging hardware. Things are kept in check by both circuitry and firmware of devices. This applies to pretty much all electronics that have some sort of brains. EVGA knows this, and it's why they took responsibility. Depending on game developers to limit excessive frame rates in menus is not a long term solution and just a temporary Band-Aid.

I assumed we all knew the cards were the problem and the game simply exposed the as of yet, still unknown fault. Or does there still remain some debate?
From what ive read some have said it may simply be a fuse, others have said an overloaded cap or vrm or...this is my question. What is it exactly? Has anyone taken one apart yet and actually found the part or parts that are failing or are we still playing the waiting game? Or worse yet relying on AIBs to tattle on themselves

 

[Mussels]

In 2021, does anyone even care about card vendors anymore? Besides cooler design, VRMs and vBIOS, there's no real big differences between vendors. It's not anymore 3D accelerator era, when sometimes some vendors actually went bonkers and did something crazy. You can't even make claims about warranty and support anymore as basically everyone tries to reduce those expenses as much as they can, so thy all suck.

Yes, especially with the 30x0 cards each variant has major changes

The FE and RE cards are all the same with a different cooler, but oh lawdy every other model has a dozen small changes. Go look at the amount of blocks EK had to put out for them all!
Power inputs (1x12, 2x8, 3x8) connectors (fan out, RGB out, ARGB out, RGB and ARGB in), VRAM cooling (front and back), noise levels, and so on

It's not the old days where we'd see a 3090 48GB GDDR3 or anything nuts, but the differences between the cards sure can stack up despite not affecting benchmark scores much

I mean the Aorus 3080 waterforce i had, had my PSU *screaming* with coil whine.... yet my replacement 3090 was dead silent on the same PSU. 100% noise from the PSU and not the GPU, i slid it out of the PC and confirmed that

 

[The red spirit]

The FE and RE cards are all the same with a different cooler, but oh lawdy every other model has a dozen small changes. Go look at the amount of blocks EK had to put out for them all!

That's something way out of budget and I won't ever care about that. And water in PC, ew.

RGB out, ARGB out, RGB and ARGB in

There's no point in it.

VRAM cooling (front and back), noise levels

O kay, this actually matters.

It's not the old days where we'd see a 3090 48GB GDDR3 or anything nuts, but the differences between the cards sure can stack up despite not affecting benchmark scores much

So, it's just different coolers then. Basically if OEMs didn't lock you out of tuning this stuff yourself, vendor wouldn't matter. It's mostly just artificial differentiation to cash in from gamurz. The only real thing that changes things a bit is heatsink.

I mean the Aorus 3080 waterforce i had, had my PSU *screaming* with coil whine.... yet my replacement 3090 was dead silent on the same PSU. 100% noise from the PSU and not the GPU, i slid it out of the PC and confirmed that

lol even more poor engineering from 3090s, no wonder they are burning up.

 

[Mussels]

It's not different coolers when they all have different PCB's
It's like the GPU and VRAM are the only consistent things

 

[The red spirit]

It's not different coolers when they all have different PCB's
It's like the GPU and VRAM are the only consistent things

That they are different, doesn't meant that they are important. Pretty much all of them perform very similarly and there aren't any big differences between them.

 

[R-T-B]

Proof? Try renaming furmark.exe to Quake3.exe or UnrealTournament.exe
Don't blame me for what happens.

lol that does nothing. You don't seem to understand how power limits work. They aren't exe name based driver profiles. Heck, they aren't driver profiles at all.

Besides cooler design, VRMs and vBIOS

Those are all reasons to care.

You can't even make claims about warranty and support anymore as basically everyone tries to reduce those expenses as much as they can, so thy all suck.

EVGAs warranty sucks less in my experience. They do tend to put effort into that image at least.

Of course, at times, it's needed, because their hardware can be hit and miss. Case in point? This thread.

 

[The red spirit]

Those are all reasons to care.

With cooler it's easy. Bigger is better. VRMs are usually fine with reference cards and only rarely can cause issues, unless you overclock (nowadays, there's really no reason to mess with card overclocking imo), they don't matter much. vBIOS is somewhat of crapshoot. You really don't know what you will get, unless you analyze them in hex editor. Essentially all cards past Polaris and Vega have it locked and not editable, so you are more or less stuck with what you got. The only thing that matters in vBIOS is fan curve, everything else is mostly irrelevant (either it's reference spec or it's a bit upgraded, but essentially it's mostly the same).

 

[R-T-B]

vBIOS is somewhat of crapshoot.

Not really. You look at the power limits on a database, like we have.

EVGA XOC bioses tend to be the highest, but there are limited gains past a point.

My point is, it's the vendors ability to blend all three of those things WELL that make a good card. And no, not all succeed. There are certainly more budget options out there with lesser performance.

 

[The red spirit]

EVGAs warranty sucks less in my experience. They do tend to put effort into that image at least.

It seems to be USA exclusive. They don't give a damn outside of USA.

 

[R-T-B]

The only thing that matters in vBIOS is fan curve,

lolwut, that probably matters the least. You can set a custom one easy afterall.

Power limits.

 

[The red spirit]

Not really. You look at the power limits on a database, like we have.

EVGA XOC bioses tend to be the highest, but there are limited gains past a point.

My point is, it's the vendors ability to blend all three of those things WELL that make a good card. And no, not all succeed. There are certainly more budget options out there with lesser performance.

I would only want lower power limits at this point. RTX 3000 series murder power supplies and RX 6000 series, aren't saints either. More than performance, noise and heat output matters. Nobody wants a space heater or a loud card. High power limit is a bad idea, especially on high end cards.

lolwut, that probably matters the least. You can set a custom one easy afterall.

Yes you can, but not permanently. If you use Wattman, you may as well don't bother with that as it never works well after reboot. You can likely do that with Afterburner, but it's janky to use Afterburner and AMD software (or nV software), which is supposed to do exactly the same thing.

 

[R-T-B]

I guess what it comes down to is a lot of enthusiasts would disagree with you, frankly. I guess you've made an effective argument as to why YOU don't care, but others do.

 

[ixi]

Looks like whatever the "lobby" is in this case is ignoring vsync and that spike in FPS is burning something out. That sounds like it falls on amazon more than EVGA/NVIDIA imo, something in the engine they're using is ignoring FPS limiters. Definitely something that should have been easily noticeable in QA but, what game company does real QA anymore. Forget microtransactions, being able to easily push out hotfixes via launchers did more to kill quality.

Also sounds like a bug that would affect any card, green or red. 3090's might just be running super hot to begin with and not leave as much wiggle room.

Even if amazon fked it up with avoiding fps limiter. That is no excuse for AIB's to not set safety measurements for not overheating or what is the case.

 

[GreiverBlade]

so... New World is the new Furmark but limited to 3090?
edit: scrolling up ... too late someone already mentioned the "universal GPU killer" (beside mining) well at last that one was indiscriminate ...

heck i am glad i do not have 3000chf to buy a 3090, nor the envy to play that mmo (or to idle on the menu screen )

one thing tho .... it only affect the 3090? ... if so, that's more, like, Nvidia at fault... surely if it was the game engine lack of fps limiter it would affect other cards ... (or EVGA ... )

/joke could not resist : also .... 9000+ fps on menu? wait ... "It's Over 9000!"? seriously?

 

[erocker]

If anyone who plays this game who cares, best way to reduce power/temps is limiting frames in the Nvidia control panel as opposed to in game. I cap mine at 100 fps and it runs cooler than capping at 60 in game.

 

[The red spirit]

I guess what it comes down to is a lot of enthusiasts would disagree with you, frankly. I guess you've made an effective argument as to why YOU don't care, but others do.

It's just that overclocking graphics cards is usually a lot harder due to various limitations and manufacturer lock downs, also there's usually not enough benefit from overclocking, power consumption of cards is already way too damn high (with those wimpy cooler that they put) and no overclock will beat just one architecture improvement. Cards today hide too many things from user and to really tweak card, you have to modify vBIOS (which is recently made near impossible thanks to nV and AMD), likely modify VRMs, make custom mounts for different cooler and have a cooler, that not only cools core, but also VRMs and VRAM. So imo graphics card overclocking is a huge mess, for small benefits.

 

[Chomiq]

You can cap the max FPS using nVidia's control panel in 3D settings. Why not using this ? Especially when you know that a lot of games don't have a FPS cap feature and you can hear your card working for nothing in the menus of many games.

Default setting is OFF, users shouldn't have to adjust settings globally on the driver level to work around lazy game devs.

Best example here - Squad, a game by a small indie studio. They have two fps limiters in place - one for actual gameplay, one for menus. That way you can set your cap to 141 for the actual gameplay and have game hover at 60 when you're in the menu changing settings, connecting to servers etc. de facto lowering GPU usage when you're not playing.

 

[erocker]

Default setting is OFF, users shouldn't have to adjust settings globally on the driver level to work around lazy game devs.

Best example here - Squad, a game by a small indie studio. They have two fps limiters in place - one for actual gameplay, one for menus. That way you can set your cap to 141 for the actual gameplay and have game hover at 60 when you're in the menu changing settings, connecting to servers etc. de facto lowering GPU usage when you're not playing.

Just wondering how it is that it's the devs fault?