Fair no I just scanned it obviously.
Well that seems to feed the speculation that its Nvidias screwup.
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Amazon’s New World game is bricking GeForce RTX 3090 graphics cards
Fair no I just scanned it obviously.
Well that seems to feed the speculation that its Nvidias screwup.
pendragon1
Extremely [H]
- Joined
- Oct 7, 2000
- Messages
- 52,226
that ^^^
No rage... we just disagree on the expected behvaiour of voltage protection systems.
I believe its silly to blame software any software for melting a GPU. These cards all have hardware level protections against exactly this. This probably is a soldering defect as per the company that makes the cards in question. Speculation goes that Nvidia changed the spec requiring higher voltage handling last min and some of the MFGs solutions are sub par, but neither evga or Nvidia have admitted anything like that nor would they. That is on them and perhaps Nvidia not Amazon and new world.
Indeed there was a time where if your CPU fan died it might catch fire as well. Then smart people figured out how to build better power delivery systems and chips that would throttle themselves. CPUs and GPUs don't in general catch on fire anymore if the cooling dies.
pendragon1
Extremely [H]
- Joined
- Oct 7, 2000
- Messages
- 52,226
this isnt about cooling.
its either crap solder, like you pointed out w/ evga, or a component being pushed beyond what is expected. just like back then...
It's definitely drawing more power than needed while running and keeping clocks maxed. I have mine capped at 60 fps, and accidentally left my mining clocks on and the game played exactly the same as it does when I left it free to clock itself (well, actually maybe slightly better) at what was effectively Half gpu clocks (I don't OC the memory for mining due to heat on the 3090 founders) and half power. So, yea, I can see there being some issue since it just seems to max out the card always.
The unlimited FPS issue is definitely fixed though, so I don't think that's the thing.
The unlimited FPS issue is definitely fixed though, so I don't think that's the thing.
Last edited:
GoodBoy
2[H]4U
- Joined
- Nov 29, 2004
- Messages
- 2,768
All you need to do is set the FPS cap in the nVidia settings to match your LCD's refresh rate. Works best if you set it to LCD Max - 2
Lateralus
More [H]uman than Human
- Joined
- Aug 7, 2004
- Messages
- 18,503
I haven’t tried the driver cap in NVCP yet. I use Rivatuner Statistics Server to set a cap, and have never seen it not work. Maybe give it a try.
cybrnook
[H]ard|Gawd
- Joined
- Jan 14, 2013
- Messages
- 1,339
What's nice about capping it in NVCP is it works all the time, just in case you forget to fire up afterburner or something.
Lateralus
More [H]uman than Human
- Joined
- Aug 7, 2004
- Messages
- 18,503
Sure, until it doesn’t work like in noko’s case lol. Maybe he just needs a clean driver reinstall?
In any case, I’m 99.9% sure that you can set Afterburner/RTSS to launch with Windows. Nothing wrong with the driver cap if it does work as designed though.
![Zarathustra[H]](https://cdn.hardforum.com/data/avatars/m/9/9298.jpg?1707758471){kind=avatar}
Zarathustra[H]
Extremely [H]
- Joined
- Oct 29, 2000
- Messages
- 38,866
I mean, EVGA already owned up to the fact that their solder workmanship issues are to blame for this. Why are we still pointing fingers at this Amazon game?
Sure the game was not designed using best practice, allowing runaway frame rates in menus, but there is no way that should ever kill a GPU that there is nothing wrong with.
Sure the game was not designed using best practice, allowing runaway frame rates in menus, but there is no way that should ever kill a GPU that there is nothing wrong with.
It seldom, and I mean seldom doesn't work in nvidia control panel. I have about 20 games installed, it only didn't work in one. It's nothing compared to amd drivers, almost none of the windowed mode configs worked (I think 2). When it comes to AMD, I use afterburner.
It's happening to multiple models from multiple manufacturers though so unless they're all having the same issues with solder there seems to be something else going on as well. I suspect that EVGA tested the model that was having the most failures and found an issue that was causing them to die at a higher rate than others while possibly leaving a more widespread issue undiscovered.
I think the cards should protect themself from that type of failure but there's obviously something screwy going on with the game too. Even if I was interested in the game I certainly wouldn't be playing it unless they figure out what's going on and fix it, on the other hand if I had one of the cards that's dying more frequently I'd still be a little concerned that other games might still kill it(just slower).
Or instead of being the games fault. Perhaps the early reports of current generation Nvidia cards dying early on are also related. EVGA specifically had a bunch of early pre production cards they sent to reviewers that where crashing in plenty of games. I don't know if the 3090 issues are in anyway related at all... although as this seems to be going beyond a handful of bad solder jobs in a few production runs. Multiple MFGs all having the same issue. Perhaps other RMAs just haven't grabbed news... anecdotally we know people have had to RMA 3000 cards for dying doing things other then new world.
Those early issues seem to revolve around Nvidia changing the power delivery specs for the chip not long before product was due to ship. If I was guessing I would assume many of Nvidias engineering partners choose the cheapest possible fix if they potentially already had a bunch of finished product. Either replacing a few capacitors or perhaps even just applying more solder. (half kidding on that)
https://www.tomsguide.com/news/nvidia-rtx-3080-crashes-caused-by-capacitors-says-evga
I'm not suggesting there aren't plenty of good 3000 cards around... just perhaps that the earliest designs might have had some design changes that are at fault. I mean it takes two seconds on google to see people complaining about cards dying.
https://www.reddit.com/r/EVGA/comments/llxsxk/do_we_have_a_high_failure_rate_for_rtx_30_series/
I know this is super anecdotal... and every MFG for every card ever has a non zero defect rate. New world isn't doing anything crazy... who cares if the menu runs at 20 billion FPS, if it was a bajillion poly scene it was running would it not also be stressed. I think New world simply brought the 3000 defect rate to light. Big well covered Beta... users complaining in one place. Instead of complaining about failures playing 100 different demanding titles. (I mean the google machine tells me Halo kills 3000 cards as well pretty regular)
It looks like the MFGs are making things right for the most part for people effected. I would have to imagine though if this is Nvidias fault with some last minute change in capacitor spec (which come on we all know that happened... Nvidia is widely reported to have pushed the chips just a little bit more right before launch to ensure they still had a lead on AMD, and the cards that seem to be failing most also happen to be the MFGs known to design their CAP layouts to hit min spec for the cheapest cost. The cards with higher end cap setups don't seem to be effected. While companies like EVGA admit adding a bunch of MLCC caps after the initial runs. Who knows if its the caps or something with the hybrid setups that doesn't switch right I'm no engineer) I imagine Nvidia will quietly make it right for their partners behind the scenes. I just hope in the future Nvidia, AMD and Intel (as I assume the other guys are paying attention) won't futz with their designs so close to launch. (and I know AMD has done the same BS changing specs at the last min in the past as well)
Last edited:
pendragon1
Extremely [H]
- Joined
- Oct 7, 2000
- Messages
- 52,226
that was for their one model, its happening to other brands too. furmark used to warn about killing cards...
Those warnings where at a time before most chips even adjusted their frequencies never mind real time fine tuning of voltage and power delivery. Completely different era where ya a sub par cooling setup could be exposed by far worse then black screens and reboots.
Also when many peoples video cards hand not been used for anything but basic 2d output... I remember the first time I got Quake to run in hardware mode on a old AMD card. Different time. First time that card had to really work and the room got warmer. No software stress test should be capable of smoking a modern GPU or CPU if its made properly.
pendragon1
Extremely [H]
- Joined
- Oct 7, 2000
- Messages
- 52,226
weve already cover this crap on the last couple pages, someone is behind.
Your still suggesting its possible somehow for software to kill GPUs by displaying graphics. Something that hasn't been true on any GPU produced in over a decade.
I get it Jay mentioned it also... but he is also wrong.
pendragon1
Extremely [H]
- Joined
- Oct 7, 2000
- Messages
- 52,226
yes, 'cause it is and its not just "bad solder". we fucking covered this already.
Furmark can destroy GPU's, as can Prime95 can CPU's. Software is very much capable of reducing the lifespan of hardware by revving (mandatory car analogy) them to the max on first gear and keeping it there. Yes modern rigs should be able to clock down to prevent damage but shit happens, especially if cheaper components are used.
I get it we agree to disagree.
Its not bad solder. I agree completely. IMO its clearly a bad power design. From all reports the companies most effected used an earlier Nvidia power requirement an designed their cards to only use a couple of the more expensive Conductive Polymer Tantalum Solid Capacitors... using the excuse that Multilayer Ceramic Chip Capacitor are better at higher frequencies. (truth being they are a hell of a lot cheaper then caps made using Tantalum... interesting side bar I worked for a supply company for years that used to supply an actual Tantalum mine, they mined the stuff in a underground mine that went under a lake and was constantly under water craziest place I have ever been). Of course the truth is there isn't anything wrong with MLCC chips... just the way it goes is there are no cheap Poscaps... but MLCCs can be had for fractions of a penny, and I would assume the quality on those can't be great.
So it seems like perhaps a few of the GPU companies (and it seems to follow its the guys known for doing things on the cheap) may have put 20 or 30 ceramic caps on $1000+ cards that where the $0.007 cent verity instead of ones that might have cost 10 cents each but been made properly. These companies cheeping out on parts on high end GPUs seems to be the issue here... but none of us know for sure and no one is going to ever admit it.
I agree... software stress tests can uncover FLAWs in hardware. Built properly no GPU or CPU made in at least 10 years should cook itself under any stress test, assuming you aren't trying to kill it by purposely removing cooling ect. Blaming this issue on a game menu... is crazy. Its 100% flawed hardware. Early 3000 cards have been dying in all sorts of game loads... there where obvious issues with the design. Those issues also appear to have been sorted... and never effected some of the companies that purposely over engineered their cards. By all accounts the companies effected seem to have all updated their designs and the cards people are getting from RMAs are a newer design.
Zarathustra[H]
Extremely [H]
- Joined
- Oct 29, 2000
- Messages
- 38,866
Absolutely not.
It may very well be that there are more than just EVGA's video cards that have workmanship issues or other defects in them making them vulnerable here, I don't know.
What I do know is that short of messing with a GPU's firmware and altering its throttling behavior there is no effing way any software load will ever harm a CPU or GPU. Not in the modern era. Doesn't matter if it's 9000 FPS or 9 billion FPS, if it gets too hot it will throttle. It's a completely closed loop.
Things that could have an effect are as follows:
1.) Video card hardware design issues (wrong materials selected, etc) causing the video cards to not be able to withstand the parameters set in the firmware for throttling, voltage, and other considerations. (This is what we saw with Nvidias solder issue on the 7xxx through 9xxx series GPU's 10-15 years ago)
2.) Video card software (firmware) design issues, causing the throttling, voltage limit and other settings to not work properly
3.) Inadequate testing to ensure the above work right
4.) Manufacturing / Workmanship issues resulting in the hardware not being able to withstand the set parameters.
All of these are the fault of either the GPU manufacturer or the Video Card AIB, NOT the software.
Short of going in and sabotaging the firmware there is no way, no how any software load, no matter how extreme can harm a non-defective modern GPU or CPU. Simply cannot happen under any circumstance at all. No way, no how.
If software harms your GPU (and it is not software that is messing with the firmware) in every case it is because that GPU was defective to begin with.
Let's not give the GPU and AIB manufacturers a pass on this one. If their GPU's are failing due to a software load, any software load, they need to recall the whole lot of them and fix them, and make their customers whole. For this I'll give EVGA some praise. They did the right thing in acknowledging that they had found a problem, being transparent about it, and taking action to immediately replace any affected video cards. This is how a good company responds. If other AIB's have the same, or similar issues, they need to take the same action.
I agree 100% that the design of this New World game - allowing menu scene FPS to run rampant - was not best practice, but under no circumstance what so ever with correctly working hardware should it have been able to damage anything.
Last edited:
pendragon1
Extremely [H]
- Joined
- Oct 7, 2000
- Messages
- 52,226
is it really a flaw if something is pushed past is known/expected limits?!
serpretetsky
2[H]4U
- Joined
- Dec 24, 2008
- Messages
- 2,180
which limit are you referring to? Is this part of some spec that software developers can reference?
pendragon1
Extremely [H]
- Joined
- Oct 7, 2000
- Messages
- 52,226
when they are building a card, the components have rated specs. this is pushing one of them to past the spec and frying it.
Last edited:
Zarathustra[H]
Extremely [H]
- Joined
- Oct 29, 2000
- Messages
- 38,866
Which component are you referring to?
If this is the case, they either selected the wrong component, or failed to properly design the firmware to enforce limits that protect that component. Either way it is a hardware / firmware design flaw.
The software (game) goes through so many layers of abstraction (Graphics API -> Driver -> Firmware -> GPU Hardware) that it is agnostic of hardware limits and cannot be held responsible. If a component has specification limits that can result in damage if exceeded, at some point in that chain there should be a protective mitigation. Preferably this would either be in the hardware itself or in the firmware, but if not, even the driver would do.
serpretetsky
2[H]4U
- Joined
- Dec 24, 2008
- Messages
- 2,180
Softwave developers (theoretically) know nothing about components. They have an API to a black box. How are they supposed to know this?
pendragon1
Extremely [H]
- Joined
- Oct 7, 2000
- Messages
- 52,226
i misread, no software wont know. hardware builders will know a spec, but the software is pushing it beyond whatever it is.
serpretetsky
2[H]4U
- Joined
- Dec 24, 2008
- Messages
- 2,180
Its unreasonable to blame software for a limit that is not part of some specification or part of the API. I agree there is an informal recommendation to "keep FPS low", but this to me is not good enough. the main fault whenever something like this happens is the hardware.
Software should patch it (because they can actually quickly create a patch in days/weeks) and hardware needs to fix this on next revision using the original software that caused the fault as a case study.
Zarathustra[H]
Extremely [H]
- Joined
- Oct 29, 2000
- Messages
- 38,866
Agreed.
Ideally this should be a limit in hardware/firmware, but if not even putting it in the drivers or the API wouldn't be a bad idea.
the highest refresh monitor on th emarket is what? 360hz? Even that is stupid IMHO, but put the limit at 360fps in the driver or API and call it a day. This shouldn't negatively affect anyone.
pendragon1
Extremely [H]
- Joined
- Oct 7, 2000
- Messages
- 52,226
lol you think that putting a limiter on the gpu isnt going to cause reeeeeeing?! look at the lhr thread.
"this card needs a limiter!!"
"how dare they limit MY hardware"
Zarathustra[H]
Extremely [H]
- Joined
- Oct 29, 2000
- Messages
- 38,866
I think those who complain are assuming that the limit will be lower than 360fps and that's why they are complaining. That's what I first thought when I heard the term "limit".
Either that, or they really don't understand how things work.
Both are equally plausible.
There used to be a time when devs knew the intricate workings of the hardware, so they could extract 99% of the performance from it. At best, today's GPUs sit at 97% on "full" load, but that number might as well be made up because we know they can't possibly use any significant portion of the entire pipeline fully.
serpretetsky
2[H]4U
- Joined
- Dec 24, 2008
- Messages
- 2,180
Yes, but it's good to have abstraction, modularization, and interface specifications. A modern developer cannot be expected to understand every detail of this system he is developing for. In any system design you always have these layers, and it's very important as systems get more and more complex.
cybereality
[H]F Junkie
- Joined
- Mar 22, 2008
- Messages
- 8,789
As someone who has worked in game development for a while, and even written basic 3D engines, a high FPS should not kill a card. I regularly have tested things that were several thousand fps with no problems. It is not the software developer's job to know what a piece of hardware can handle (at least not these days) but it is the hardware manufacturer's job to ensure that the software API they are supporting (like DirectX 12) works fully with all features. I agree that it is good practice as a developer to set an internal frame limit (hopefully that can be changed with a config file) maybe something high but reasonable like 300 fps. And as a user, it is a good idea to set a frame cap at or 2 below your refresh rate, and also use V-Sync/G-Sync/FreeSync. That should eliminate that possibility. But turning V-Sync off has been a feature of 3D games since the beginning, and a GPU should not die just cause it's on a menu screen with V-Sync disabled. And, in any case, there are many games that have this behavior (high fps on menus) and I've never heard of widespread failures. Also, any AMD cards die with this game?
pendragon1
Extremely [H]
- Joined
- Oct 7, 2000
- Messages
- 52,226
Lateralus
More [H]uman than Human
- Joined
- Aug 7, 2004
- Messages
- 18,503
Yep. Post #120. Also some other posts early on reporting the 6900XT as affected.
cybereality
[H]F Junkie
- Joined
- Mar 22, 2008
- Messages
- 8,789
So that video was interesting, but he shows he really doesn't know about game development. Many new games are using borderless fullscreen now (in fact, the way modern OSes and APIs work, this is the preferred method), and the resolution you set in the menu is an internal render target.
Setting the game to 4K will just change the internal render resolution, the window size doesn't necessarily need to change, and it would make sense that the card was being pushed harder (as 4K is almost twice the pixels content of 1440p) and thus drawing more power. So his confusion there is kind of embarrassing.
But the rest of the video was pretty good. It definitely seems like some sort of power delivery problem. I still don't think a game can kill a card, it's probably just that the game is demanding and exposing hardware flaws that already existed.
Setting the game to 4K will just change the internal render resolution, the window size doesn't necessarily need to change, and it would make sense that the card was being pushed harder (as 4K is almost twice the pixels content of 1440p) and thus drawing more power. So his confusion there is kind of embarrassing.
But the rest of the video was pretty good. It definitely seems like some sort of power delivery problem. I still don't think a game can kill a card, it's probably just that the game is demanding and exposing hardware flaws that already existed.
Last edited:
pendragon1
Extremely [H]
- Joined
- Oct 7, 2000
- Messages
- 52,226
ie: the game is killing the card. this isnt rocket surgery....
could be the extended extra voltage draw killing a component.