News Amazon's New World Is Out of Beta, Still Killing RTX GPUs

[renz496]

You do realise that these are often all made at the exact same factories, based on the exact same Nvidia reference designs, just with a custom designed cooler laid atop. The difference usually comes in the little tweaks to the numbers each vendor makes and the cooling solution.
I’m baffled why I’m explaining this to “Tom’s Hardware” of all sites…

AIB sometimes cut corners to the point their custom designed PCB quality end up being subpar against reference standard that being push by AMD or nvidia.

 

[GM_Turtle]

New reports have come in that Nvidia's GeForce RTX 30-series graphics cards are still dying after playing New World.

Amazon's New World Is Out of Beta, Still Killing RTX GPUs : Read more

Can confirm. Had no problems with my EVGA RTX 3080 XC3 Ultra Gaming until I played New World's second beta (where I had been told the GPU problem had been addressed) and 5 minutes into the game PC shut down lost all power and wouldn't come back on. After assuming it might be the PSU (it wasn't) then "Oh than it must be the mobo/cpu" (upgraded both and then test benched the GPU with different PSU/MOBO/CPU etc everything checked out fine. It was very costly and time consuming to find out that my practically new 3080 had fried. After paying a "collateral deposit" of a whopping $1800 EVGA sent me an advanced RMA and everything works now.
Current Build: i7 11700k, Gigabyte Z590 mobo, 16GB Corsair Vengeance 3200, EVGA RTX 3080 XC3 Ultra Gaming GPU, (3) Corsair LL120's and Corsair H100i AIO for cooling and this thing is never playing New World again. ¯\(ツ)/¯

 

[Zerk2012]

Their a lot more information I would like to see before saying a single game is killing cards.

The way social media spreads these days it don't take a lot to trigger something like this.

New game so their a lot of people that are probably exclusively playing it a lot of hours.

How many of the same cards have just failed doing other things, we all know these cards can have huge power spikes what was their 12V rail doing when these spikes happen. It could be as simple as the game causing the power spike to last longer than usual and the 12V rail going wacky out of specks or the power spike itself killing the cards.

 

[FITCamaro]

I have a 3070 and no issues. Runs the game mostly at 50+ FPS at 4K and Very High settings. I've seen it drop to 30-35 FPS in certain situations. Switched to 1440p (still looks fine on my 4K monitor) and now solid 55+ FPS at Very High settings all the time.

 

[Phaaze88]

It's not limited to New World.
Halo: Master Chief Collection was doing it too.
I believe there's more than just these 2 out there, but I suppose no one really touches on them.

Anyone know what these 2 titles have in common?

 

[alceryes]

It's not limited to New World.
Halo: Master Chief Collection was doing it too.
I believe there's more than just these 2 out there, but I suppose no one really touches on them.

Anyone know what these 2 titles have in common?

Haven't heard about the Halo MCC doing this.
Source?

They're two different engines. New World being a custom Amazon engine built on the CryEngine and Halo MCC using both the Saber 3D and Unreal 4 engines. Only commonality I know of is both of them being 'heavy' from a resource use standpoint,

 

[Daggre]

New reports have come in that Nvidia's GeForce RTX 30-series graphics cards are still dying after playing New World.

Amazon's New World Is Out of Beta, Still Killing RTX GPUs : Read more

If these cards are running overclocked, it's a risk that just comes with OC'ing your hardware, but if they're running at stock speeds, it's completely unacceptable that the cards let themselves get overheated like this without throttling themselves or at least shutting down. This is not something software should be able to do if it TRIED to.

 

[Phaaze88]

Haven't heard about the Halo MCC doing this.
Source?

So MCC is a rtx 3090 killer. (Gigabyte Vision OC 3090) :: Halo: The Master Chief Collection General Discussions

Well, just lost my second 3090 after the first refund (3 Months) playing MCC on PC. Watch out! https://forums.evga.com/tm.aspx?m=3150118&mpage=2 The Link is evga, but i got a Gigabyte GPU, same issue! Thanks for nothing.

steamcommunity.com

MCC a 30 Series killer? :: Halo: The Master Chief Collection General Discussions

So I am one of the lucky few who managed to get a 3090. An EVGA 3090 FTW3 Ultra specifically. While playing the Halo 3 campaign, my fans spun up to max, the image went dark, and I heard the game audio for about a minute before I had to reset. I assumed it was a weird glitch, so I loaded the game...

steamcommunity.com

https://forums.evga.com/Major-Issue...wed-by-no-video-output-from-GPU-m3150118.aspx

https://forums.evga.com/EVGA-3090-o...ow-as-it-might-kill-the-card-m3432165-p6.aspx

[You probably should use ctrl+f Halo, because trying to read through all the EVGA threads would take a while...]

Skimming through real quick, it looks like LoL and FFXIV pop up too...

 

[alceryes]

Damn.

I really think this is a manufacturer-cutting-corners issue.

 

[BadMrSnowman]

The issue is even deeper than that. Even with a power limit set well beyond what the vrm is capable of, the card shouldn't fail. The VRM should simply trip ocp and shut down safely.

Either ocp has been disabled via a modified vbios, which I doubt, or ocp is set improperly. Ooooor, the issue is a hardware fault.

I was thinking the same thing. I would expect something like "Everyone who played New World on a 3090 GPU had their card shut down, and a few unlucky ones had their card damaged." Poor soldering is a problem, but there should other safeguards in place that would also need to fail.

 

[loki1945]

There is no way that game will touch my PCs; with the current ridiculously greedy GPU prices I refuse to risk them for 1 game, and I'm sure I'm not alone. Though most likely this is 100% on nvidia and friends skimping on quality/testing; you have to have a crap design/cut corners for software to nuke your GPU.

 

[husker]

Remember the issue with capacitors? An EVGA response was:

"During our mass production QC testing we discovered a full 6 POSCAPs solution cannot pass the real world applications testing. It took almost a week of R&D effort to find the cause and reduce the POSCAPs to 4 and add 20 MLCC caps prior to shipping production boards..."

I'm not saying this issue is related to the POSCAPs, but supports the argument that the thorough testing of a new hardware generation just didn't happen.

 

[helper800]

I'm not saying this issue is related to the POSCAPs, but supports the argument that the thorough testing of a new hardware generation just didn't happen.

I am not aware of a new generation of hardware that went smoothly.

 

[Phaaze88]

JayzTwoCents has a video up, but what I found more interesting was the comments section.
-Killed cards:
iRacing
WoW
Greedfall
Ark
Cyberpunk
GTA V
[I just scrolled down the comments]
Can't blame Amazon devs for those...

-Software can damage hardware.
But there's several different samples out there, from different engines even... so they're all that badly written? 🤔
It doesn't help that some partner models did cut corners from the reference designs...

-Radeon 6000 isn't immune either.
We barely hear about them due to obvious reasons...


~So... a combination of both software and hardware?

 

[ConfusedCounsel]

No game or program should be able to do something outside of driver specifications. Driver specifications should ALWAYS prevent the GPU from doing something that is detrimental to its ability to function (overclocking and power limits included).

Until I hear of an FE card suffering this fate, I put the responsibility for these failed cards soley on the AIB manufacturers. If it turns out to affect FE cards as well, I put the responsibility on the drivers AND AIB manufacturers.

At no point will I say that the game is at fault. Drivers AND hardware should protect against poor programming - period.

Remember Furmark, Prime95, some builds of OCCT, software can definitely circumvent hardware protections.

 

[ConfusedCounsel]

I think it is probably a combination of software, hardware, and Power Supplies. The most vulnerable cards seem to be taking 300 Wats of Power or more. When asked for power, Power supplies generally give too much then back down. This can be seen by how many times the Kombuster stress tests in AfterBurner spike the power draw over the set limit. If the components have sufficient thermal headroom, this shouldn't be a problem, and the device should hold long enough for the software or PSU to back off, or the protections on the board to kick in. However, 300 Watts, that is a lot of power, and if you keep hitting a protection circuit again and again eventually it is going to fail. So, I suspect the software is just hitting the boards to hard, to many times, until something finally gives.

 

[alceryes]

Remember Furmark, Prime95, some builds of OCCT, software can definitely circumvent hardware protections.

Agreed.
I'm just saying that hardware SHOULD be protected by firmware/OCP, not that it's necessarily the case now.

 

[RodroX]

Huge hype from nvidia presentantion and media -> slow production due to pandemic -> high demand + mining -> high prices -> low stock for end user = solution lets skip some quality checks to have more cards on the street.

Im not saying this is the case for every maker out there but, you start with a poor silicon (cough samsung cough), and then some gpu makers maybe cutting corners on QA and QC to have more cards on street. If that wasn't enough a rough competition from AMD cards too.

 

[Diceman_2037]

video card manufacturers are getting a rude awakening on skimping the build quality.

definitely the case

Remember Furmark, Prime95, some builds of OCCT, software can definitely circumvent hardware protections.

hardware wouldn't need protecting if they were build to standard levels of quality.

Now, I'm just one person, so take what I say with some skepticism.

No, haven't heard of any FEs getting bricked.
Nvidia used some high quality components in their FEs. The cooler isn't the most effective out there, but I wouldn't be surprised if the rest of the gpu isn't plain designed and built better than most of the aftermarket models.

The partners didn't comply with everything laid before them - they tried to cut corners to save profit margins - because Nvidia is a dick to its partners.
They stood to lose money - perhaps make money slowly is more accurate - if they applied the FE design and bill of materials to all their models.


Once again, somebody, or someone, gets screwed over, because money.

The cooler is actually the best in the market, what its plagued by is flex causing the pads used to lack pressure, once addressed easily compete with the overenginered chunks of metal that aibs use.

 

[Phaaze88]

I wasn't saying the FE cooler was terrible. It's sort of in the middle.
The noise levels could be a little better.
For 3080 and higher cooler, there's a lack of mounting pressure as you said - I do love how they don't dump all their waste heat inside the chassis like the AIB models do though.

overenginered chunks of metal that aibs use.

I wouldn't even call them that. It's the same old dual and triple slot coolers they've been using the past several years, the cheap ~baa!
They didn't adopt Nvidia's cause of the cost, and they STILL cut corners on other components.

 

[gargoylenest]

all the while, Im running that game at maximum graphics (1080p)at around 50-60 fps with my very old overclocked sapphire rx480 nitro+ and a ryzen 2600x....

 

[hotaru.hino]

Agreed.
I'm just saying that hardware SHOULD be protected by firmware/OCP, not that it's necessarily the case now.

If you're relying on software to protect hardware, then you're doing it wrong.

 

[alceryes]

If you're relying on software to protect hardware, then you're doing it wrong.

Yup.
Unfortunately, that doesn't seem to be the case here.

 

[helper800]

If you're relying on software to protect hardware, then you're doing it wrong.

It is impossible to build a piece of computer hardware that is protected 100% via physical hardware installed on the PCB. You can take the best materials in the world and perfectly solder a board and all of its components together and put it in a system with a purpose built killer poke, and it will die. There needs to be firmware levels of protection IN ADITION to hardware protections in every consumer card.

So many people talking in generalities about what should and should not be or how things are vs how they should be without an understanding that malicious software can kill even the most secure hardware, and how a power source can take out any consumer electronic hardware built today.