compguru910
I don't know if anyones suggested it or not, but flashing the bios from a 2x8pin 3080 might work for you. I say this because thats what worked for my 3090. Specifically the XC3 bios. The power reporting was off after that, but the 3rd 8pin pulled the full 150w after that and consumption was somewhere around 500w. 

My first RMA failed. They sent me a faulty 3090 FTW3 Ultra Gaming again. 
The second RMA has been a complete success.
They gave me a brand new Rev. 3, EVGA 227, which reaches easily 450 W. Even reaching at last 501 Watts isn't a big deal for the card.
EVGA did a great job exchanging the power buggy card. Just look at it by yourself

Audioboxer
Sorry, one more post, most conclusive evidence that the card is NOT drawing power correctly
 
400w standard BIOS
 

 
450w OC BIOS
 

 
Same curve profile in afterburner, barely any difference between the two. In fact, for the bigger laugh, the power draw spike of 403w on the default BIOS is higher than the 401w on the "450w BIOS". Pegged around same frequency. Same iCUE 12v power reading. There has been no increase in the ability for the card to draw more power, even GPU-Z correctly states performance cap due to Power.
 
Will forward this onto EVGA. Either the BIOS is faulty (unlikely) or EVGA has lied about this being a proper 3 pin power delivery card. In which case it's wrongly advertised, though chances are they'd weasel their way out of this saying it's sold rated for up to 400w. Or, it's a total lottery if your card is manufactured better than other people's 3080s and you MIGHT have the ability to draw power correctly on the third pin.
 
Now I understand the total silence from EVGA/Jacob. Poor show given how good EVGAs customer service normally is and how they handled the New World blows up GPUs drama.
 
Given other manufacturers of 3 pin cards do not suffer from this, my conclusion at the moment is AVOID all EVGA 3080/3090 products (other than Ti, apparently they didn't cheapen out there, unsurprisingly). Can this be fixed with a BIOS/Firmware update? Doubt it, seeing as EVGA have done absolutely nothing for around a year now and they seem to be refusing any RMA requests based on power draw. Seems they seriously cheapened out on whatever the controller is that manages card power delivery on the 3080/3090 FTW3. Incredibly stupid move given the FTW3 cards are always some of the most expensive RRPs.
 
I will stand corrected if wrong and/or this can be fixed, and as I said above I am recognising EVGA's usual great customer service. Radio silence on explaining this though is very poor.

Audioboxer
It's nothing to do with memory, that has been tested at 0. I've even been able to run at +2000 without crashing, but the one thing memory obviously does do at higher clocks is help hit that 400w wall quicker.

In general gaming I've found +1600 strikes a good balance in power hungry games like Metro Exodus to not pull the core clock down due to the card being near 400w.

Customer service are now away to check again with technical support if the BIOS is compatible with LHR cards. Bit of a farce really, I know it isn't, but I like to be nice and civil with support staff.

Should really have been Jacob and other well known front facing media figures of EVGA who confirm why some cards can't seem to go above 400w, or struggle to reach the 450w of the EVGA BIOS. Where others will happily sail right up to it if a load wants more power. EVGA released this 450w bios, it is therefore up to them to explain why it doesn't work on some cards.

In related news, Veii from Overclock is reverse engineering the BIOS of cards to pump up the power draw https://www.overclock.net....1753930/post-28969627 Gives me hope EVGA could do something at a BIOS level. Not nvflash compatible yet though. Nvidia and/or some of the manufacturers seem to be artificially throttling the controllers from what I can gather. But it's then extra confusing why some cards will draw more power than others with same BIOS.

Audioboxer

talon951

Audioboxer
It's nothing to do with memory, that has been tested at 0. I've even been able to run at +2000 without crashing, but the one thing memory obviously does do at higher clocks is help hit that 400w wall quicker.

In general gaming I've found +1600 strikes a good balance in power hungry games like Metro Exodus to not pull the core clock down due to the card being near 400w.

Customer service are now away to check again with technical support if the BIOS is compatible with LHR cards. Bit of a farce really, I know it isn't, but I like to be nice and civil with support staff.

Should really have been Jacob and other well known front facing media figures of EVGA who confirm why some cards can't seem to go above 400w, or struggle to reach the 450w of the EVGA BIOS. Where others will happily sail right up to it if a load wants more power. EVGA released this 450w bios, it is therefore up to them to explain why it doesn't work on some cards.

In related news, Veii from Overclock is reverse engineering the BIOS of cards to pump up the power draw https://www.overclock.net....1753930/post-28969627 Gives me hope EVGA could do something at a BIOS level. Not nvflash compatible yet though. Nvidia and/or some of the manufacturers seem to be artificially throttling the controllers from what I can gather. But it's then extra confusing why some cards will draw more power than others with same BIOS.

Well the bios can absolutely be changed to fix it as it controls all the limits, but didn't you show that the 3rd 8pin on your card either is low or at least reads low relative to other 3080 FTW3s? I would think that would be the best way to get EVGA to RMA it.

The 3rd pin wattage? Like other people with this issue it struggles to go anywhere above 60-75w.

B0baganoosh
Just curious on the memory clock question. Have you done testing in port royal, or time spy with a range of memory clocks? This Gen is a bit different than previous cards with memory clocks. You'll see no artifacts or crashing if you start hitting error conditions, these have error correction, so you won't see visual issues as much, but you will lose performance. It looks a bit like a bell curve. I've tested repeatedly to see that my performance starts going back down if I exceed a +1100 OC on memory. Your number will likely be different, but there will be a number far below where crashing or artifacts happen where your performance will start going back down. I've always looked at the FPS performance, but I haven't looked at the power draw once I head into the error zone on memory. It would make sense.

I'm not saying that's definitely your issue, but I do think it's worth checking with a repeatable benchmark to find your own peak point before performance starts going back down. Maybe yours is +2000, but I haven't seen anybody ha e thag high of a number unless they're throwing some serious cooling at it.

ty_ger07

CraptacularOne
I find it humorous that there has been an EVGA employee in here saying everything looks fine and yet still here we are with conspiracy theories and nonsense.

It's not nonsense. There is a 239 page thread on this subject for the 3090 where EVGA agreed that it was a problem and created a special RMA process to correct the issue. Why, as soon as it isn't a 3090, everybody believes that the exact same symptoms aren't a problem?
What is nonsense is an EVGA employee saying that an obvious defect -- which can patently be shown to limit the performance of the card -- is something that no one should be concerned about.

CraptacularOne
So if your car has a 8000RPM redline....do you drive it around town constantly at it's redline just because it says it can?

No, but when you try, it should be able to. If you try -- even though you don't normally every day -- and you find that it won't go over 6000 RPM because the cam timing was set way off, you don't say that it isn't a problem. It is a defect you discovered which should be addressed. It still affects you, even when you aren't running at 8000 RPM.

It's like Crap woke up one day and said 'benchmarks are stupid, and I have always thought they were stupid'. Like, what? Who are you any more? A power-limited product line has a specific card that is more power-limited than it should be, and Crap says that is fine? Not if it was yours. Tone deaf.

rjbarker
400w or 450w or 460w......really doesnt matter imho...as long as the Card lasts longer 3-4 months with out another visit to the Red Light District I can live with any power balancing issues or power delivery issues.
 
For what its worth my highest PR and TS Scores peaked out my Card around 445w...whether that means anything or not....at this point dont really care.
 
Real world gaming performance is what matters to me ....after the initial becnhmarking with a new Card.
This 3080Ti has been quite solid thru 120 Hrs of CP77....and now FC6 w High Texture Pak is solid as well....Valhalla etc.
 
Really...after all the initial benchmarking is said and done (dont know anyone that "plays" PR or TS), best tests for the Cards is to find a decent OC and game on....I have settled on +120/+750....no crashes in any games.
 

ty_ger07

CraptacularOne

ty_ger07

CraptacularOne
I find it humorous that there has been an EVGA employee in here saying everything looks fine and yet still here we are with conspiracy theories and nonsense.

It's not nonsense. There is a 239 page thread on this subject for the 3090 where EVGA agreed that it was a problem and created a special RMA process to correct the issue. Why, as soon as it isn't a 3090, everybody believes that the exact same symptoms aren't a problem?
What is nonsense is an EVGA employee saying that an obvious defect -- which can patently be shown to limit the performance of the card -- is something that no one should be concerned about.

CraptacularOne
So if your car has a 8000RPM redline....do you drive it around town constantly at it's redline just because it says it can?

No, but when you try, it should be able to. If you try -- even though you don't normally every day -- and you find that it won't go over 6000 RPM because the cam timing was set way off, you don't say that it isn't a problem. It is a defect you discovered which should be addressed. It still affects you, even when you aren't running at 8000 RPM.

It's like Crap woke up one day and said 'benchmarks are stupid, and I have always thought they were stupid'. Like, what? Who are you any more? A power-limited product line has a specific card that is more power-limited than it should be, and Crap says that is fine? Not if it was yours. Tone deaf.

You are really take a leap here aren't you? I never said benchmarks are stupid, but furmark and kombuster are flagged as a power virus in the drivers and they will not let the card exceed a certain point. Have you forgotten how to read? Who are you anymore? 
 
That 239 page thread is maybe a handful of people that actually know what they are talking about and observing the rest is just a bunch of "mee too" people that think they see something and then open up GPUZ and see "pwr" as the perf cap reason and immediately start shouting from the hilltops. They don't know what they are looking at nor do they know how to interpret it. This whole "your card should be drawing it's max power every single time I use it" is just nonsense. The card may momentarily blip to it's peak but often the resolution of the monitoring software won't catch it. They don't know how to lower sensor refresh rate to get a better idea of what's happening. Even in benchmarks like 3Dmark the card doesn't "need" to pull it entire rated peak power through the whole damn benchmark. That's not a realistic ask for your average air cooled or even water cooled card. 
 
You can stop your holy campaign here man, you've really gone off the hinges as of late. 

Okay, then what do you think this person should do different?
Everything they have shown so far is "not fine". You say there is a fine in there somewhere to find. So, lead them.

It's not about "not always running at peak", it's about never even coming close to peak.

EVGA never said that their cards were purposely designed to have broken power balancing. All they say is "don't worry about it" or simply no comment. Except for the 3090 where they did admit that it was a defect.