eric-mt

The warranty replacement I received today has 0.1 and red lips. The card itself looks pristine, like a new card. The packaging was generic.
 

You're almost certainly looking at another RMA when NW is released. Most of the cards that failed were Rev 0.1 cards, even after the 'fix' was deployed in game.

It would not be a bad idea to run some benchmarks with Time Spy and/or Port Royal to see how the card is handling voltages. I'd be surprised if the old red lips cards were not susceptible to the same failures.

mine exploded after 56.5Hours on New world ( i had losing signal of monitor way before playing New WORLD / gpu drivers getting corrupted/ i reinstall windows like 5 times...... all this time i suspected CPU /MoBo) ...... something was wrong with the card from beggining ( i blame PSU= i bought 1600 T2 .....i blame heat= i put on water..... i blame memory = i bought 32gb 4000mhz Cl17 Memory )
-----on short i wasent enjoing my new gpu ever since i bought it
my GPu is sandwich on Water ( bykski waterblock + active watercooled backplate) ...i   NEVER saw more than 65 C Hot Spot ... thought dead after 10 months of using it ( losing Monitor signal was one a week...and more often and more often.... up to 4...5 times a day...... only a restart my monitor would come to life)

Not sure why eVGA said it's "noise on the i2c bus" when several users already confirmed directly that when the 2 million RPM appears, the fans ARE actually revving to maximum speed (and sometimes a black screen follows soon after).

eric-mt

Kylearan
Not sure why eVGA said it's "noise on the i2c bus" when several users already confirmed directly that when the 2 million RPM appears, the fans ARE actually revving to maximum speed (and sometimes a black screen follows soon after).

Noise on the bus can confuse the temperature control algorithm in the BIOS. The BIOS as a result thinks the card temperature is very very high. Like 2^16 (64K) degrees C. This could cause the BIOS to call for an extremely high fan RPM, say 2 million. In reality the BIOS should never call for a fan RPM higher than what the fan hardware is capable of, and the fan controller should ignore such baloney.

 
I don't think they ever said the fans weren't revving to maximum when the "2 million RPM" appears in GPUz. They just said the reading was wrong. Neither is the fan running at 2 million RPM, nor is the BIOS "requesting" it to. It's just a bad reading. The BIOS doesn't request RPMs...it requests %'s. These are PWM fans and all the BIOS can request is a duty cycle between 0 and 1 (100%). Even if it were to do something weird in the software and "ask for a 7 out of 1" (insert Kevin James's "Tim gave me a G!" bit here), the hardware can only drive to a "1" which is full-voltage all the time, which is 100% fan-speed. That readout in GPUz is not what anything is "requesting", it is the reading from the tach line on the fan. So when they say "noise on the line is causing the readings to be bad", it has nothing to do with what the fans are or are not trying to do.
 
What I find interesting and a bit concerning personally is that they're planning on releasing a firmware update that will "fix" this "noise on the i2c line". I don't know what the source of the noise is, but I don't think a firmware update is going to make it go away. I think at best, it will just ignore the noise with some logic filtering. I mean...maybe if they know what is causing the noise and they can adjust how that's operating with the firmware update, it will actually fix it, but I'm just a wee bit skeptical on that.
 
Also, the fan issue and erroneous RPM readings are a red-herring. Rather, a symptom of the actual problem, not the problem itself in relation to the card failures. What I find interesting here is that the cards failing have analog VRM controllers (no i2c lines), so there's no relation to noise on the i2c line causing erroneous fan RPM readings (and temperature readings?) and that corrupting the VRMs: they don't even talk to the fan controllers...or anything digitally for that matter.
 
Something New World is doing (and other titles like LoL and Halo) seems to be causing these cards to operate in such a manner that they overheat or request huge amounts of current. We don't know if VRMs are actually blowing, or if fuses are popping, etc. So we don't know what part is failing, but the fan issue is just a side-effect. Something is getting over-stressed and when that happens, yes, the fans will ramp up to 100% (or whatever their programmed maximum is anyway) as the card tries to save itself. And just to clarify, New World and other titles should not be able to cause this. Clearly they do create a situation where the cards are operating in a state that doesn't allow them to protect themselves (they often try to, which is the earlier shut-downs before they actually fail), but the cards should protect themselves better and more reliably. It sounds like what is failing is something intended to protect the card only a limited number of times. For example, MOVs are a device that can protect from over-voltage spikes, but they usually have a limited number of uses before they fail short. Sort of a "well how many times is this toaster going to be hit by lightning?" type of protection. The built-in OCP in the VRMs is usually made in such a way that they can hit it again and again without actually damaging anything. It's almost like it triggers the OCP a couple times, but in doing that, causes the OCP to relax it's set-point each time...which eventually allows enough current to pop a fuse. This is not a something I'm suggesting is happening, but a theoretical idea intended to illustrate what could cause the failures people are describing. Until they give us any details (if they ever do), that's all we can do. Even if that were the case (which it probably isn't) and you could solve the OCP shifting issue, you'd still have to ask yourself "why are games like New World pushing the card into OCP protection?!" That's the many thousands of dollars of RMAs and bad publicity value question: What is it about the design that could allow a poorly coded game or an older title not designed for this hardware to create a situation where the card can't protect itself adequately? 
 
TLDR (because before I knew it I've written another book in here): fan RPM issue not related to card failures, just bad data on the lines. Card failure issue still unknown. Conjecture! Mystery! Rabble Rabble Rabble!

Does anyone know if the Kingpin 3090 is susceptible to this issue?

Xavier Zepherious
I think the issue is relate to PSU OCP or OPP

I disagree. Makes zero sense. All you are doing is creating another patch to protect a bad card from destroying itself.

The problem manifests itself as OCP on the card tripping over and over again (black screen / 100% fans) until finally it happens too many times that the VRM pops and the fuse blows (or the fuse blows before the VRM pops, but less likely).

PSUs with too aggressive of OCP were an issue initially, right? In your theory and solution methodology, the PSU tripping problem would simply happen more and more often and then people would again blame the PSU.

What if the GPU is the problem? ;)

What if all the symptoms we have seen since release are all adding up and all confirm the same problem? I feel like I am beating a dead horse, but all the evidence really does line up. All the cards with the worst power balancing have had the most problems. It's no coincidence that EVGA created a PCB revision. It's no coincidence that the old PCB had the majority of problems. It's no coincidence that EVGA is being silent about the root-cause and is instead RMAing as fast as possible to reduce complaints about the issue.

After a 10 days finally getting new card shipped out. Except I received 2 emails with 2 serial numbers in sequence. Wouldn't that be nice to receive 2 cards for 1..  LOL

That's pretty sad for a flagship card costing as much as car from the 60's LOL

my card poped after 56 Hours of New world ....  i got it replaced 3 weeks ago
sooooo...... i sould not play New World on release with the new card ???
anybody knows >???
 
btw i saw somebody talking about psu or temp problem ( i got evga 1600 T2 Psu  and watercooling my 3090ftw3 on both sides with bykski waterbock 65C max temp)]

this isnt isolated to this issue with the 30 series cards..no way in heck i would even consider buying one..they are way over priced..not readily available and arent using quality parts. 

So now there's this.
 
Thoughts?
 
https://www.pcworld.com/a...geforce-rtx-3090s.html

Kylearan
So now there's this.
 
Thoughts?
 
https://www.pcworld.com/a...geforce-rtx-3090s.html

Interesting. A rare soldering issue eh?

they forgot about all the cards that died before New World... lol...

Tom's has the article as well... Myself and several others had cards fail and only played NW after the menu frame limiting patch. Is there any official source of the findings about the soldering? They mention it was findings from EVGA but I don't see anything directly from them

I certainly hope they do tell us specific serial numbers so we can at least check and see if we need to be worried / send ours back.