Nozler
And from something I spotted on Twitter here ya go>  https://www.youtube.com/watch?v=rpvsaedeeqg

Not legit. Jayz always licking manufacturers asses.
 
Also why he tested already patched NW??? That's no sense.

Not mine I don't have any issues more info for those that do I'm not debating anything..

Kylearan

jaketeck
Interesting. These boards have shunt resistors and power monitoring all over the place. I suspect that there may be a path that doesn't have power monitoring and that this game somehow maxes that path which leads to the failure. I'm reading through the entire thread now, has anyone inspected the failed cards yet?
 

StanJarensky3
For what it's worth, RTX 3080 owners are also reporting this. I had something very similar happen to me while playing FFXIV on my 3080 FTW3 Ultra so I'm assuming it's a similar problem. Theoretically, setting a maximum FPS globally should stop whatever application/game from drawing obscene amounts of power during unlocked loading screens, right? I'm about to RMA my 3080 to EVGA today but in the meantime, I'll be setting a global FPS cap to my monitor's maximum refresh rate. 

 
I think my 3080 was having a similar issue when playing RE:Village 

Ok this is going to be messy.
 
I keep trying to tell people this but not many people listen to me.  But the few who do actually go "hmm...interesting."
There is MORE than one power limit on the card, boys.
Not only that, the shunt resistors only shunt the INPUT RAILS that are fed from the main power rails to the main chips.
However there are more shunt resistors than there are main input rails.
On an 2x8 pin card, there's
 
1) 8 pin #1
2) 8 pin #2
3) GPU Chip Power (CPU Core NVVDD Input Power (sum))
4) PCIE Slot Power
5) MVDDC (FBVDD) / memory power
6) PWR SRC (Power Plane Input Source Power)
 
Shunt #6 is bizarre because the SRC chip not only has its own power limit (Most cards set this to 150W at 100% TDP slider and 175W at whatever TDP is the max slider position, which then becomes normalized past 100% to scale from 150W to 175W), but the SRC actually has several of its own power limits which equal the # of 8 pins (!).  The SRC power limit controls the maximum power that can be pulled from an 8 pin.  This is DIFFERENET from the individual 8 pin power limits!  The individual 8 pin power limits combined with PCIE Slot Power control the TDP VALUE when you add them.  The SRC value however does NOT.  The SRC value controls the max allowed from an 8 pin before a "TDP Normalized%" power throttle happens.
 
Even more bizarre is that the SRC power rail itself changes depending on the power coming from all the other shunts.  If you do a shunt mod and you shunt a rail with poor soldering and bad contact, SRC may show LESS watts at full load than at idle.  It can actually end up showing 0 watts at full load.  This may also cause MVDDC (memory power) to also show 0 watts at full load.  If you fail to shunt SRC, it may end up reaching its own power limit and throttling the entire card (VIA TDP Normalized%) or it can cause another rail to skyrocket (not a true reading btw) e.g. causing 8 pin #1 to report 175W and 8 pin #2 to report 80W, which triggers another power limit.
 
And these are just the shunts.  There are more sub-rails, SOME of which when "summed" together, add up to one of the primary rails, and some which report only to "TDP Normalized" and are not exposed via NVAPI (but might be accessible via I2C with Elmor's EVC2X).  All of the "Input" rails shown in HWinfo64 (both primary and secondary) respond to shunt mods.  None of the "Output" rails respond to shunt mods.  However MSVDD, NVVDD and PLL power rails also do not respond to shunt mods but ARE reported to TDP Normalized based on a "default" and "maximum" value.
I'm not talking about voltage rails.  I'm talking about the main power rails themselves which feed them.
 
MSVDD drawing too much power relative to what MSVDD voltage is supposed to allow will report a high "TDP Normalized" power draw (with respect to TDP%) and cause GPU effective clocks to drop, relative to requested clocks.  NVVDD drawing too much power relative to what NVVDD voltage (this is internal voltage, not VID--MSI Afterburner only supports setting VID) caused outright Power limit throttle by TDP normalized (without effective clocks dropping first).
 
Here is what MSVDD's unreported power limit does to TDP Normalized.
 
Path of Exile: 400W TDP (Shunt mod: Slider set to 58% TDP).
 
1) Shadows + Global Illumination Quality: High.  TDP=400W, 58% on TDP slider)
(Notice TDP Normalized %, TDP%, board power draw and the clocks).
 
https://i.imgur.com/FLAITFL.png
 
2) Shadows + GI = Ultra (TDP = 400W, 58% slider)
https://i.imgur.com/joz3YCc.png
 
Notice the difference in clocks and TDP Normalized, even though total board power remains the same.
 
3) Shadows + GI : High, 800W TDP, 150 mhz core, windowed, TDP=114%
 
https://i.imgur.com/pO3sPMg.jpg
 
4) Shadows + GI: High, 800W, 150 mhz core, fullscreen, TDP=114%
 
https://i.imgur.com/C853Rg5.png
 
5) Shadows + GI High, 800W TDP, 90 mhz core, fullscreen, TDP=114%
 
https://i.imgur.com/nUnQJC7.png
 
6) Shadows + GI: Ultra, 800W TDP, +150 core clock (danger!), TDP=114%
(Fullscreen).
 
https://i.imgur.com/EmvAsiD.png
 
I couldn't run this very long.  I've black screened + 100 % screaming fans before from doing this at that TDP unless I lowered the clocks.
 
7) Shadows + GI: Ultra, 800W TDP (windowed), TDP=114%
https://i.imgur.com/0yWpWTM.png
 
Please compare the TDP Normalized difference between the windowed Ultra and High tests.
 
Notice that on the Ultra tests, the clocks and voltages are being throttled due to a high TDP Normalized even though total TDP is nowhere near 114% (hovering around 87% or so).
But look in hwinfo.  Only the VID (NVVDD) is being throttled.  SRAM voltage (MSVDD) is not being throttled.  1.013v NVVDD Vs 1.094v MSVDD.
 
TDP Normalized starts throttling the clocks when any individual rail (in this case it's the unreported MSVDD / NVVDD rails!) exceeds 108%.  It starts mildly throttling the clocks.
 
Superposition 4k custom / Timespy Extreme has had TDP Normalized go up to 119% with some massive clock throttling.  However neither of these programs managed to black screen 100% fan the shunt modded FE Card (at most it just crashes the driver).
 
But POE + Ultra + 800W TDP has managed to hard crash the power phases (past 90 mhz core clock overclock) because the MSVDD rail just gets overloaded with the power draw going way past 600W with temps past 75C.  Card didn't die though but it was scary.
 
What I'm curious of is if New World is doing something similar.  If it is, TDP Normalized should be able to report something being way higher than it should.
 

I don't think that people don't listen, but they are probably struggling to understand. I had to read this about 3 times to really understand it all. Not because you wrote it poorly or something, but just not a trivial subject.

Are you following that 3080ti XC3 thread with everyone up in arms over the cards only showing 300-350w in GPUZ? I'd be curious to know what your thoughts are on that too. I'm assuming it's hitting some internal limit that we can't see (or directly see) in the monitoring tools.

Outofstock4ever
it's easier to patch a game then to recall all the defective cards. I sure wouldn't stand for it if i paid that much for a defective product. Damn EVGA is worst then the NZXT fire hazard case. The case at least was insanely cheaper.

 
Go deal with another gpu company and enjoy THAT service.  Stop kidding yourself,you WOULD stand for it, and you would wait for an RMA just like everyone else. You're just like the rest of us otherwise you wouldn't be here!

SinisterGamer

Outofstock4ever
it's easier to patch a game then to recall all the defective cards. I sure wouldn't stand for it if i paid that much for a defective product. Damn EVGA is worst then the NZXT fire hazard case. The case at least was insanely cheaper.

 
Go deal with another gpu company and enjoy THAT service.  Stop kidding yourself,you WOULD stand for it, and you would wait for an RMA just like everyone else. You're just like the rest of us otherwise you wouldn't be here!

Exactly.  EVGA's customer service and willingness to make things right is second to none.  They have not been perfect in their card designs historically, but are always proactive, investigating the problem and offering solutions.  If a solution isn't readily available, they are quick to offer no-beg RMAs when they are at fault.

babayega
what abt the kpe cards?

Has anyone reported a WC card that died? All reports I’ve seen is air cooled. I would hope the basic stuff radiator and fans with the KPE should be enough to combat this issue. Otherwise hold off till you can get in a custom loop. At that point I’ll never worry about heat again, then again I’m a bit more overkill then most

ShawnB420

babayega
what abt the kpe cards?

Has anyone reported a WC card that died? All reports I’ve seen is air cooled. I would hope the basic stuff radiator and fans with the KPE should be enough to combat this issue. Otherwise hold off till you can get in a custom loop. At that point I’ll never worry about heat again, then again I’m a bit more overkill then most

Governator

squall-leonhart

Governator
I've read through this entire thread and found it to be very interesting to say the least, curious if this has affected anyone who's water cooling.

considering one of the videos was a Hybrid, figure it out.

Sorry, I should've been more specific I was referring to full coverage custom loops, active backplates etc.

 
I've seen a few people with hybrid cards report their card dying. I'm on a custom water cooling loop with my 3090 FTW3 HC and haven't run into any issues. (Two 360mm radiators and a large EK reflections front distro for my Lian Li O11 Dynamic XL.)
 
The 3090 HC came out in April though so it might have an updated(?) design. I'm 9.9 hours in without a single crash/studder.
 
Edit: Should also mention that a few people on Reddit are reporting that their 3090 FTW3 cards that had a waterblock put on them, also failed. Wish more people shared their board revisions and whether they were hybrid, air cooled or water cooled.

neteng101

talon951
Are you following that 3080ti XC3 thread with everyone up in arms over the cards only showing 300-350w in GPUZ?

I tried to tell those guys why and most are utterly clueless but refused to listen.  Can be summed up in this one sentence really from Kylearan below.  Total power draw is just one of many limits.  And power monitoring in software is not accurate enough to tell us the full picture, plus the polling interval means no software measurements are truly real time either.  If you truly want to geek out watch many of Buildzoid's older videos where he shows stuff crashing because of transient spikes using an oscilloscope.
 
There is MORE than one power limit on the card, boys.

Indeed. The reference 3090 has I think 9 shunt resistors meant to monitor different power paths. Then we have 3 rails of power with several power stages per rail, so I'm actually speculating here but there has to be at least 4 different power limits. One for total draw power, another for the core, another for GDDR6 modules and another for MSVDD otherwise why Nvidia would implement so much power monitoring if there isn't going to be used.

babayega
what abt the kpe cards?

KPE has a much better designed power management and PCB design than the FTW3 ever did. The FTW3 problems existed from day 1.

neteng101

talon951
Are you following that 3080ti XC3 thread with everyone up in arms over the cards only showing 300-350w in GPUZ?

I tried to tell those guys why and most are utterly clueless but refused to listen.  Can be summed up in this one sentence really from Kylearan below.  Total power draw is just one of many limits.  And power monitoring in software is not accurate enough to tell us the full picture, plus the polling interval means no software measurements are truly real time either.  If you truly want to geek out watch many of Buildzoid's older videos where he shows stuff crashing because of transient spikes using an oscilloscope.
 
There is MORE than one power limit on the card, boys.

Yep, on the 3080 Ti FTW3 you'll usually ride one of the power limits like a pony all the time even though the overall card power is about 425w and rarely hit a 450w max, maybe 440w.  But then if you go back after a half hour and add up all the individual 8-pin and pcie max power readings, they sum to around 450w or more.

talon951
Yea that's pretty much what I told them too. Although less frequently, I've seen that kind of behavior with my 3090 also depending on what bios I run. Must be something set a bit lower (or too low for the 3080ti) because most of them seem to exhibit that behavior in a larger number of games/benchmarks, not just EVGA cards. Still need an updated version of ABE that can read the 3080ti bios.

On my 3080 Ti FTW3 Ultra - PCIe slot power has intentionally been lowered from my observations, and PCIe #2 is the number 1 cause of power limiting, it draws the most power of all the 3, while 1 draws closer to 2, and 3 is always way below.  I believe the PCBs on FTW3's likely have hardwired inputs to power phases, and the way the power delivery is on the PCB it causes the power unbalance issues that have been seen on all FTW3 3-connector cards since launch (3080/3090 and now the 3080 Ti).  The only improvements seen are to deal with the early card failures especially 3090 FTW3s (reduction in PCIe slow power draw, change of VRM controller getting rid of the uP9511).
 
And someone on OCN measured with a clamp so even with the 450W BIOS the 3080 Ti is running like a 480-500W TDP card, #2 is reporting lower than it really is drawing in software.  Considered trying the Galax HOF 1000W BIOS but it doesn't seem worthwhile now that I know the card is actually overperforming already.

Although it's not proven all of the 3080ti FTW3s exhibit that higher power draw. Could be they do but I haven't seen anyone else verify it.

Hosting fake queues, selling faulty gpus. Wouldn't be surprised to see their downfall if they keep this up.

neteng101
And someone on OCN measured with a clamp so even with the 450W BIOS the 3080 Ti is running like a 480-500W TDP card, #2 is reporting lower than it really is drawing in software.  Considered trying the Galax HOF 1000W BIOS but it doesn't seem worthwhile now that I know the card is actually overperforming already.

That was the guy you're responding to.  Lawl.

mikeinmissoula
So after reading this entire thread. You're telling me that League of Legends is going to brick my brand new 3090 FTWUG?? Because if thats the case, my 3090 FTWUG and 3080Ti FTWUG will be going on ebay today. I mainly play WoW, LoL, Doom, and planned on trying New World. I've spent $3400 on EVGA GPU's. I don't want them to die on me and have to do all this RMA nonsense. First time EVGA owner by the way. Always used ASUS cards before. But the elite sign up and price got me to go EVGA. I haven't built my new computers yet, so if the cards are a no go for the games I love. They are gone..... 

Is LoL really bricking 3090's?