gsrcrxsi

GlaucomaPredator
 
 
I haven't looked up the data sheet for the fan controller, but I suspect it can in several ways. 
One if it's reporting and trying to generate a high frequency and the crystal responsible for it can not handle it it will fail. Then the card goes into some protection mode. 
If the fans are being controlled digitally by the IC via a duty cycle and the fans can not read or use the value, fans stop spinning and card goes into some sort of protection mode, hence the black screen. 
Then a hard reboot and power cycle resets the memory and some users report being able to bring the card back to life. 
But in the end I think whatever is generating the reading or the duty cycle/frequency for the fans, fails, and puts the card into some sort of protection mode as not to fry the GPU core or memory. 
 

what does "generate a high frequency" mean in this context? high frequency of WHAT?
 
crystal oscillators are used for frequency generation based on multipliers against the natural frequency of the crystal. they are not driven to higher or lower frequencies, and they run at insanely high frequencies natively. there is no such thing as "the crystal can't handle it" they aren't driving the crystal, it just runs and they extract frequencies via multipliers against the crystal.
 
the "protection modes" that most of these cards are experiencing is very likely the OCP modes built into the power stages themselves. nothing to do with frequencies or oscillators. And if you watch Son of a Tech's video (not that I recommend his channel, as I hate his deadpan delivery and dry commentary, but his video on this topic has some good clues) he specifically noted that he had a short on the 8-pin connectors which tripped his PSU's OCP. A very common cause of a short here, is a failed power stage.
 
"fans stop spinning" is the exact opposite behavior than people are reporting. everyone says "black screen + fans at max speed". not sure how you think the fans are shutting off
 
it sounds like you're just using random words and phrases but don't actually understand what they mean or how they work on a GPU (or electronics in general).
 
but still, Igor's article is pretty bogus on this subject. there's no real analysis that shows the false readings are in any way connected to the cards failing from the New World game. not only did his card, which has these bad fan readings, NOT fail during New World gameplay, but also exhibited the same bad readings in other games. there's nothing connecting this issue with false fan speeds to New World game 3090 failures.

Thank you. This was also my interpretation, but I admittedly stopped reading the poorly translated Igor's lab article about 3/4 of the way through when it just kept repeating itself about the RPM readings without actually showing any correlation or evidence that fans were doing anything other than running to 100% (as they're PWM fans...and cannot ramp up past 100%). There's clearly an error in the RPM reading, but I don't see how that would brick a GPU.

There are 2 articles on this topic from him
#1: https://www.igorslab.de/en/evga-geforce-rtx-3090-vs-amazons-new-world-design-caused-evga-problem/
#2: https://www.igorslab.de/en/evga-geforce-rtx-3080-rtx-3090-and-not-only-new-world-when-the-graphics-card-goes-amok-because-of-design-failures/
 
 
from #1, he admits to not have any hands on time with an EVGA card, just hearsay from some "other colleague directly in the Asian R&D departments of the biger AICs". his whole argument seems to be "EVGA seems most affected, and EVGA is the only one running iCX. therefore I conclude the problem is a failed fan controller... somehow. case closed!" it's nonsense. there's no evidence presented that supports this. he did not even have an EVGA card on hand to test, and only verified high FPS numbers with other AIB cards, without failure.
 
he also continually mentions "burned up" fan control chips, with no proof of such. no pics posted, no references. nothing.
 
#2 first off, testing with a 3080 FTW3. not even a 3090. and only 1 GPU is referenced, and it didnt fail. again claims this issue has been "known for months" but he's the first person I've heard of claiming it's an issue lol. everyone else is just referring back to his article(s) about it. circular logic. the only evidence he shows here are some GPU shots showing faulty readings (in more than just New World).

Igor
But if there are already errors at 60 FPS, what about at 6000 FPS++? That’s exactly what you don’t even want to think about anymore.

more baseless claims. he's seeing some behavior at some FPS and making the assumption that the phenomenon will get progressively worse with more FPS. no evidence to support this.
 
he also claims to have issues setting the fan speeds (implying this as more evidence for a faulty fan controller), but never mentions what software he's trying to use. I think it's common knowledge that 3rd party fan control software doesnt work well with these cards and you need to use PX1. maybe he wasnt. but we can't know since it's not mentioned. I'm going to guess he's trying to use MSI AB or something.
 
and at the end of the day does not even make any claims as to how this could kill a GPU. Zero evidence for it.
 
so where is the evidence? cause it seems like people are reading too much into this. it seems his 3080 FTW certainly has something wrong with the fan controller. but probably isolated to his one sample. I haven't seen this shown elsewhere. but again, nothing to do with the actual GPU failures being seen

schoolofmonkey
I'm curious, say your motherboard has it's own protection on the PCI-E slot not allowing a device to draw over 75w, would that still allow the fuse to blow.

I don't see over 72w on the PCI-E power with my 3090 XC3 Utlra, but I haven't played New World, not my thing.
Kingdoms of Amalur has a unlocked framerate in the menu as well (saw 3000fps at one point), it never caused a issue.

I understand where the fault it, but was just curious.

As far i know your XC3 is safe from this issues that's why only FTW3 are getting fried.

gsrcrxsi

GlaucomaPredator
 
 
I haven't looked up the data sheet for the fan controller, but I suspect it can in several ways. 
One if it's reporting and trying to generate a high frequency and the crystal responsible for it can not handle it it will fail. Then the card goes into some protection mode. 
If the fans are being controlled digitally by the IC via a duty cycle and the fans can not read or use the value, fans stop spinning and card goes into some sort of protection mode, hence the black screen. 
Then a hard reboot and power cycle resets the memory and some users report being able to bring the card back to life. 
But in the end I think whatever is generating the reading or the duty cycle/frequency for the fans, fails, and puts the card into some sort of protection mode as not to fry the GPU core or memory. 
 

what does "generate a high frequency" mean in this context? high frequency of WHAT?
 
crystal oscillators are used for frequency generation based on multipliers against the natural frequency of the crystal. they are not driven to higher or lower frequencies, and they run at insanely high frequencies natively. there is no such thing as "the crystal can't handle it" they aren't driving the crystal, it just runs and they extract frequencies via multipliers against the crystal.
 
the "protection modes" that most of these cards are experiencing is very likely the OCP modes built into the power stages themselves. nothing to do with frequencies or oscillators. And if you watch Son of a Tech's video (not that I recommend his channel, as I hate his deadpan delivery and dry commentary, but his video on this topic has some good clues) he specifically noted that he had a short on the 8-pin connectors which tripped his PSU's OCP. A very common cause of a short here, is a failed power stage.
 
"fans stop spinning" is the exact opposite behavior than people are reporting. everyone says "black screen + fans at max speed". not sure how you think the fans are shutting off
 
it sounds like you're just using random words and phrases but don't actually understand what they mean or how they work on a GPU (or electronics in general).
 
but still, Igor's article is pretty bogus on this subject. there's no real analysis that shows the false readings are in any way connected to the cards failing from the New World game. not only did his card, which has these bad fan readings, NOT fail during New World gameplay, but also exhibited the same bad readings in other games. there's nothing connecting this issue with false fan speeds to New World game 3090 failures.

As I said several times before, several users on this very thread posted VIDEOS and were able to confirm by their own testing that their black screen crashes *PRECEDED* the fans revving up to maximum speed.
They *also* demonstrated that when the fans showed 2 million RPM, the fans were actually revving to maximum speed.  One of them also showed the second fan revving up (but it only went from 1200 RPM to 1600 RPM).  If he didn't stop what he was doing (pulling up tooltips in Anno 1800), the card black screened in short order.
 
Many users also mentioned that their fans were sometimes revving up (by sound) in the games which killed the cards long before New World--Halo and League of Legends and Halo:MCC randomly --BEFORE the black screens happened.
 
In other words, the fans revving up to 100% speed (Without the card crashing) and the card black screening are related to each other in some way which eVGA Taiwan engineers must know about --othewise a revision 1.0 PCB and digital VRM controller wouldn't exist.

gsrcrxsi
How could they see that the rpm reading was “2 million RPM” if the screen went black?

But the fans can’t be more than 100%. They just can’t. It’s only an incorrect reading. It has nothing to do with the failures and certainly isn’t causing it.

Hook your GPU up to a power source without a live computer. 12V with no control = 100% fan speed. This is what’s happening when people are seeing fan speed ramp to 100% when the GPU craps out. It’s still getting power but fan control stopped working since the GPU went dead so fans got the full 12V and run full speed. Literally all of my computers do this when I power up my system and before it POSTs.

I pointed this out elsewhere, the ran is rated for 3500 RPM at 100% PWM or no signal, but the card only runs it at 3000 RPM when set 100% on the slider during normal operation.  This extra bonus 500 RPM does sound very fast and loud.

Update: my friend's Gigabyte RTX 3090 Xtreme just dead during playing GTA V, 100% fan, black screen and not posting any more, just same as my first dead EVGA RTX 3090 FTW3U. He testes his card's 3 8pin connector, the first 8 pin is broken.
 
FTW3's faulty PWM controller and fried RTX 3090 could be just two different issues.

Surely New world wouldn't brick the Kingpin 3090?

cryptomasters
Surely New world wouldn't brick the Kingpin 3090?

Kingpin cards generally have extremely low failure rate. Kingpin is made for extreme OC and to last. It is basically made up from the best of everything, unlike the FTW3 cards which are middle of the road and have some compromises. Currently FTW3 cards seem to have the highest failure rate. XC3 cards have low failure rate due to the 300w PL, and Kingpin the PCB components are too good for a game to destroy it. 

justintylerwright

hallmark888

cryptomasters
Surely New world wouldn't brick the Kingpin 3090?

Kingpin cards generally have extremely low failure rate. Kingpin is made for extreme OC and to last. It is basically made up from the best of everything, unlike the FTW3 cards which are middle of the road and have some compromises. Currently FTW3 cards seem to have the highest failure rate. XC3 cards have low failure rate due to the 300w PL, and Kingpin the PCB components are too good for a game to destroy it. 

The 1800 dollar flagship card is middling and not built to last with compromises. Noted.

Well I guess what he meant is slightly different, you pay mostly for the GPU and the memory on a 3090... The rest is a set of compromise.

A KP card is an out of the league in comparison, they make probably a lot less money on KP than any other cards. It's like a hyper car versus a super car. Every single bits of components are cherry picked (not talking about binning), to get the most durable product under the most extreme environment, -130C for LN2 overclock.

An FTW3 ultra is a supercar, you have compromise, to keep cost low at a reasonable price. (Yes the price is reasonable, as it is a better attempt to basic models). But sometimes supercar fail... Think of these Ferrari on fire, you have the equivalent.

So yeah it is meant to last, but some design flaw can affect the assumptions.

gsrcrxsi
How could they see that the rpm reading was “2 million RPM” if the screen went black?

But the fans can’t be more than 100%. They just can’t. It’s only an incorrect reading. It has nothing to do with the failures and certainly isn’t causing it.

Hook your GPU up to a power source without a live computer. 12V with no control = 100% fan speed. This is what’s happening when people are seeing fan speed ramp to 100% when the GPU craps out. It’s still getting power but fan control stopped working since the GPU went dead so fans got the full 12V and run full speed. Literally all of my computers do this when I power up my system and before it POSTs.

Same way as on the 3090 FE.
Max fan RPM at 100% speed is 2600 RPM
But the fan can actually run at 3000 RPM
You can hear it visibly when you black screens to overtemp protection or overload protection and the fans sound MUCH louder and whinier than 100%.
I can approximate the difference by running my IPPC industrial fans at 2600 RPM vs 3000 RPM, even though that's a very terrible comparison, comparing Noctua industrial fans to puny fans.
What I'm saying is that (again--don't ask me--read the users posts who commented on this AND who uploaded videos!!  I'm not that person!), when that user had the fan show "2 million RPM", it sounded VERY loud, louder than what 100% sounded like.  And it happened when he pulled up that Anno tooltip.  If he didn't close the tooltip quickly, a black screen followed.
 
It's not hard to put 2 and 2 together.  I'm starting to wonder if some people are intentionally trolling and trying to act dumb just because they think it's cool to do so.  Or maybe COVID really did destroy people's critical thinking skills...

Kylearan

gsrcrxsi
How could they see that the rpm reading was “2 million RPM” if the screen went black?

But the fans can’t be more than 100%. They just can’t. It’s only an incorrect reading. It has nothing to do with the failures and certainly isn’t causing it.

Hook your GPU up to a power source without a live computer. 12V with no control = 100% fan speed. This is what’s happening when people are seeing fan speed ramp to 100% when the GPU craps out. It’s still getting power but fan control stopped working since the GPU went dead so fans got the full 12V and run full speed. Literally all of my computers do this when I power up my system and before it POSTs.

Same way as on the 3090 FE.
Max fan RPM at 100% speed is 2600 RPM
But the fan can actually run at 3000 RPM
You can hear it visibly when you black screens to overtemp protection or overload protection and the fans sound MUCH louder and whinier than 100%.
I can approximate the difference by running my IPPC industrial fans at 2600 RPM vs 3000 RPM, even though that's a very terrible comparison, comparing Noctua industrial fans to puny fans.
What I'm saying is that (again--don't ask me--read the users posts who commented on this AND who uploaded videos!!  I'm not that person!), when that user had the fan show "2 million RPM", it sounded VERY loud, louder than what 100% sounded like.  And it happened when he pulled up that Anno tooltip.  If he didn't close the tooltip quickly, a black screen followed.
 
It's not hard to put 2 and 2 together.  I'm starting to wonder if some people are intentionally trolling and trying to act dumb just because they think it's cool to do so.  Or maybe COVID really did destroy people's critical thinking skills...

There are a lot of dumb gamers out there that don't like fans over 1500 rpm.  You should hear the quad of 120x38 3400rpm San Ace server high pressure fans I put on my Hybrid.  Like twice the static pressure of the Noctua 3K and rated at ~.8 amps each.  Sounds like a circular saw at full speed.

gsrcrxsi it's like thinking you can get your car to go faster when you're already at 100% throttle by pushing the throttle harder. nothing happens, you cant give it anymore.

Where can you go from there?
 
https://youtu.be/uMSV4OteqBE
 
 

Let me know when your GPU can shift gears.

Bruno747
So are we done, did evga succeed in sweeping this under the rug, shipping out a few more questionable cards under RMA and hoping for the best?
 
Was really hoping evga would finally provide some details on why the 3090 FTW3 seems so fragile.

I said this would be forgotten in a week or two.  Clowns over on reddit still recommending underclock for overheating cards.

deathlessdd

Bruno747
So are we done, did evga succeed in sweeping this under the rug, shipping out a few more questionable cards under RMA and hoping for the best?
 
Was really hoping evga would finally provide some details on why the 3090 FTW3 seems so fragile.

Clearly to evga it wasnt alarming enough since its not the only game that has killed gpus so id say the cards that failed had issues to begin with. So not enough eyes on the issue it must not really be that big of problem.

They should just force a Bios update that puts the 100% PL at 350W like the FE cards.  Unfortunately that would probably drop the stable boost clock for some cards so they wouldn't be making advertised spec anymore.  Keep the max PL.
 
I bet EVGA Taiwan thinks the card is fine when operated in controlled circumstances. Likely the design's projected failure rate in the field is about what they are getting, ie. high, what do you expect for a 420w air cooled card.