The problem lies in the fact that the vRAM is erroneously hitting it's power limit & thusly the whole card is showing power limited when it's not actually hitting 500W.   If you don't believe me, try underclocking your vRAM by -250MHz & witness how much more stable & higher you can go on your GPU Core clock before hitting PWR PerfCap.  It's actually kinda ridiculous, I've seen up to 489W power draw while underclocking the vRAM & OC testing the core, in TimeSpy at least.  With the vRAM at stock, I can't get higher than around 473W, and only in that test, almost every game/benchmark besides that doesn't go much past 450W no matter what I do, unless I underclock the vRAM.
 
For now I'm still running my OC/undervolt of 2040MHz @ 1000mV on the core, it's completely stable in Cyberpunk 2077 for hours of play, and results in cooler than stock temps, with better than stock FPS, and I rarely see power limit as a )PerfCap reason due to the lower power draw by undervolting.
For reference, at stock, the 3090 FTW3 Hybrid typically boosts to 1890-1935MHz depending on temperature etc, usually sitting right around the 1920MHz mark, so my settings are basically a 120MHz OC (or stock clocks on a Kingpin card), with only 1000mV to the core, instead of the 1058-1075mV stock runs at.
Pretty sure almost every 3090 FTW3 Hybrid should be able to run 2040 @ 1000mV, so I highly recommend giving it a try... if it's unstable, bump it down to 2025MHz.
 
Speaking of Undervolting, has anyone noticed that even if you save a profile in Afterburner, the clock setting can randomly change itself after rebooting sometimes, and even occasionally while Windows is running, resulting in you having to re-set the MHz/voltage curve to get it at what you want it to be at again?  I think this is an unfortunate bi-product of Afterburner's Ampere support still being in "beta".

Dabadger84
The problem lies in the fact that the vRAM is erroneously hitting it's power limit & thusly the whole card is showing power limited when it's not actually hitting 500W.   If you don't believe me, try underclocking your vRAM by -250MHz & witness how much more stable & higher you can go on your GPU Core clock before hitting PWR PerfCap.  It's actually kinda ridiculous, I've seen up to 489W power draw while underclocking the vRAM & OC testing the core, in TimeSpy at least.  With the vRAM at stock, I can't get higher than around 473W, and only in that test, almost every game/benchmark besides that doesn't go much past 450W no matter what I do, unless I underclock the vRAM.
 
For now I'm still running my OC/undervolt of 2040MHz @ 1000mV on the core, it's completely stable in Cyberpunk 2077 for hours of play, and results in cooler than stock temps, with better than stock FPS, and I rarely see power limit as a )PerfCap reason due to the lower power draw by undervolting.
For reference, at stock, the 3090 FTW3 Hybrid typically boosts to 1890-1935MHz depending on temperature etc, usually sitting right around the 1920MHz mark, so my settings are basically a 120MHz OC (or stock clocks on a Kingpin card), with only 1000mV to the core, instead of the 1058-1075mV stock runs at.
Pretty sure almost every 3090 FTW3 Hybrid should be able to run 2040 @ 1000mV, so I highly recommend giving it a try... if it's unstable, bump it down to 2025MHz.
 
Speaking of Undervolting, has anyone noticed that even if you save a profile in Afterburner, the clock setting can randomly change itself after rebooting sometimes, and even occasionally while Windows is running, resulting in you having to re-set the MHz/voltage curve to get it at what you want it to be at again?  I think this is an unfortunate bi-product of Afterburner's Ampere support still being in "beta".

Does the 1000w Kingpin bios solve this issue by giving the vram more power? And if this is the case, then why does my card show 470w to 500w continuous draw depending on the game and yours does not? Not trying to be antagonizing. Just curious.

Dabadger84
The problem lies in the fact that the vRAM is erroneously hitting it's power limit & thusly the whole card is showing power limited when it's not actually hitting 500W.   If you don't believe me, try underclocking your vRAM by -250MHz & witness how much more stable & higher you can go on your GPU Core clock before hitting PWR PerfCap.  It's actually kinda ridiculous, I've seen up to 489W power draw while underclocking the vRAM & OC testing the core, in TimeSpy at least.  With the vRAM at stock, I can't get higher than around 473W, and only in that test, almost every game/benchmark besides that doesn't go much past 450W no matter what I do, unless I underclock the vRAM.
 
For now I'm still running my OC/undervolt of 2040MHz @ 1000mV on the core, it's completely stable in Cyberpunk 2077 for hours of play, and results in cooler than stock temps, with better than stock FPS, and I rarely see power limit as a )PerfCap reason due to the lower power draw by undervolting.
For reference, at stock, the 3090 FTW3 Hybrid typically boosts to 1890-1935MHz depending on temperature etc, usually sitting right around the 1920MHz mark, so my settings are basically a 120MHz OC (or stock clocks on a Kingpin card), with only 1000mV to the core, instead of the 1058-1075mV stock runs at.
Pretty sure almost every 3090 FTW3 Hybrid should be able to run 2040 @ 1000mV, so I highly recommend giving it a try... if it's unstable, bump it down to 2025MHz.
 
Speaking of Undervolting, has anyone noticed that even if you save a profile in Afterburner, the clock setting can randomly change itself after rebooting sometimes, and even occasionally while Windows is running, resulting in you having to re-set the MHz/voltage curve to get it at what you want it to be at again?  I think this is an unfortunate bi-product of Afterburner's Ampere support still being in "beta".

The problem is it's not the memory power limit at all that's throttling you!
The memory power limit is 121W (default), 130W (max TDP), and I don't think you're reaching that.
However I have seen on the Founder's Edition cards is that Memory and PCIE slot shunts are linked to each other, meaning PCIE throttling can cause memory power draw to skyrocket (due to power balancing fail).
Take a screenshot of your GPUZ with MVDDC set to "max" in the column and see if you're reaching 121W.  If you're not, it's not the memory rail that's throttling you.
 
It could simply be underclocking the memory also lowers the PCIE power draw (remember what I just told you about how MVDDC and PCIE slot shunts are linked?).
 
https://i.imgur.com/FAKrSJL.jpg

So I tried the 500w bios, & ran the superposition benchmark the GPU uses around 490w with highs slightly above 500w and lows of about 485w, drops to 470's when the scene changes.  OC settings for this test were +669 memory, +120 core, +100 voltage, 119% power limit, seems, temp 75c @ 100 fan speed.
GPU clock during benchmark was around 1980mhz
question:
I didn't save my original BIOS, and flashed the OC bios onto the Normal jumper switch position Bios, I downloaded the original bios from the 1st page of this thread and re-flashed it. But in GPU-Z the bios version shows 94.02.26.48.15 with a 420-450W PL, is this correct? I thought the Default Normal Bios PL was 400w PL, and the Default OC Bios was 450w PL, I redownloaded the original normal and OC bioses and re-flashed them to the right position jumper switch, but the only thing that seems different between the normal and OC is the bios# normal is 15 OC is 16. is this correct? 
anyway see attached GPU-Z screenshot from the Superposition Benchmark while using the 500W bios (right click and open link in new tab too see the bigger picture)
 
edit: just tried some gaming, and I'm pulling very close to 500W all the time when not CPU limited that is and gpu load is maxed. So yea at least in my case it seems the 500W bios works. The air cooler is not enough to keep the temps in line even at 100% fan speed at 500w  and the clocks drop below 2000mhz and temp settles around 75c. in the end you just use way more power for pretty much the same performance.
My GPU is the china 2012, and I'm using the Corsair HX1000i PSU with a seperate 8 pin cable for each plug.

Turbo-12R
I love watching all the self-proclaimed pros jockeying for who has the best answer for as what the actual problem is.    Ever consider a job at EVGA as a engineer to rectify the problem? 

This is Mostly a User to User support Forum
 
Everyone is entitled to express their opinion as to what they believe is Helpful or can challenge the wisdom given - but its best to say why you think their information is inaccurate & your source of data to convince members that your point is worth considering
 
The Beta BIOS is Not officially supported ... use at your own risk ... results Not guaranteed

Your situation may be an outlier? Or maybe mine is, but I play cyberpunk at +1300 Ram and 2050 core all day long. The card will only draw the watts when it needs it.

evgablee
Your situation may be an outlier? Or maybe mine is, but I play cyberpunk at +1300 Ram and 2050 core all day long. The card will only draw the watts when it needs it.

I can guarantee you that clocking ur ram to 1300+ is losing you fps. Try around 850-1000 in games and see ur fps jump. 

Your wrong dude - I am getting 200fps in cod Cold War at 4K 21:9. I don’t think I have seen anyone getting that many frames in that game at 4K

It’s funny to me how kingpin lays out exactly how he goes about over clocking his cards in a YouTube video yet people don’t even pay attention. The same method can be used for all evga cards nit just a kingpin card. Anyway, good luck with your overclocking.