Got a bit of an odd issue.
 
When running 3dmark TimeSpy Ultra stress test if I adjust the power target above 100% I can never complete all 20 loops (no matter which bios I use on the card).
 
In all instances the driver fails (but there is no blue screen, 3dmark just stops the test and reports a 0% score):
 
LKD_0x141_Tdr:6_IMAGE_nvlddmkm.sys_Ampere
 
Note that NO CLOCKS ARE CHANGED (I haven't "overclocked" in a very loooong time, I just adjust the power target as high as it will go...I water cool and the cards boost themselves) 
 
If I leave the power target at 100%, I pass all 20 loops, all the time with usually a 99.1% score.
 
I have the latest bios for my motherboard (and no, I do NOT have the infamous cap issue, I checked), latest vga bios, latest driver (yes, I used DDU in safe mode)(this was occurring on previous driver too), Windows 11 installed (this occurred on windows 10 as well).
 
I actually do not need the power limit upped from what I can tell, but I am concerned that the card might be faulty. 
 
Basically with the power limit at 100% the card does not draw over 400 watts, it appears that anything over this crashes the driver.
I do have new power supply (Seasonic Prime  1300 platinum), so I cannot rule that out, but this does not seem like a psu issue.
 
Any advice or insight is appreciated. 

its gets more weird.
 
I tried a +135 (was boosting as high as 1935, so that is why I chose this value) with staying at 100% power limit...stress test failed again, not really a surprise.
 
Then I tried +135 and maxed out the slider (113%) and to my amazement, stress test completed and boosted as high as 2070 mhz...****? This is the first time the card has completed the stress test at anything over 100% power limit. Power draw stayed between 397 - 410 watts (all values eyeballed via nvidia overlay)
 
I'll run it again, but this is such odd behavior. If I can can get a good overclock, perhaps my earlier concerns were misplaced and I just need to find the sweet spot for this chip...of course stability is the priority. 

AHowes
Ever try reinstalling 3dmark or just reinstalling timespy?

This happen on any orher 3dmark bench?? Firestrike or port royal?

I did reinstall, neglected to mention that :) 
 
Will try those others and report back, was trying to get stable on a single bench before moving on (but I do see the value in attempting those others) 
 

Also do you have https://www.hwinfo.com/ installed to see what the power and temps look like.. can see 3v/5v/12v as well as all the sensors on the card.

I should probably list my system specs :) :
 
12700k, no mce, flat stock
asus z690 hero latest bios
G.SKILL Trident Z5 RGB Series 32GB (2 x 16GB) 288-Pin PC RAM DDR5 6400 Intel XMP 3.0 Desktop Memory Model F5-6400J3239G16GX2-TZ5RK (yes this is on qvl) running in XMP 2.0
Intel 900p 280 gb - OS
Intel 660p 512 gb - storage
 
3X 360mm rads, (2 push/pull, 1 pull) 2x ddc variants in loop, ek mono block, ek 360 flt in phanteks 719 case
 
I tried Port Royal stress test with overlock and it failed, but this time with the "LKD_0x141_Tdr:6_IMAGE_nvlddmkm.sys_useroc" message.
 
Then I tried stock and it passed.
 
Actually, since this appears to be working at stock...I (should) have no complaints. Though I'd prefer a better outcome, the promised performance is there.
 
Furthermore, I am glad the driver crashes rather than pull power it does not need, nor can use...this is a good thing. Just because this behavior doesn't align with my habit/preferences does not mean it is "bad". I am thankful EVGA had this card at msrp...I logged on, and convinced myself to buy it...that easy. 
Worth noting that this is replacing a 3080 FTW3 Ultra (aftermarket water-cooled), so maybe I am looking for purchase validation (the 3080 ftw 3 replaced my 2080ti that is being gifted along with a system to nephew for good grades...he still pays around $600 for all of it, but he is getting steal) 
 
I am still seeing boosts of up to 1935mhz at stock at right around 400 watts. 
 
Currently running Firestrike extreme stress test at stock and plan on running memtest overnight (to rule out ram issues....this is built into asus bios, might as well)

Guess I spoke too soon. About 2.5 hours into playing Elex 2, I got a crash at stock settings: LKD_0x141_Tdr:6_IMAGE_nvlddmkm.sys_Ampere.
 
Currently running occt power test (with power at 113% for gpu). So far pulling ~ 650 watts, half of psu capacity.
 
Before I go to bed, will set memtest to run.
 
If both test pass, I guess it is rma time :( 

wanescotting
Guess I spoke too soon. About 2.5 hours into playing Elex 2, I got a crash at stock settings: LKD_0x141_Tdr:6_IMAGE_nvlddmkm.sys_Ampere.
 
Currently running occt power test (with power at 113% for gpu). So far pulling ~ 650 watts, half of psu capacity.
 
Before I go to bed, will set memtest to run.
 
If both test pass, I guess it is rma time :( 

:/ yeah if the card can't do defaults rma time. (

Hope ya get a new card in return.. has it been 14 days since purchase?

Whatever it is.. get the trouble ticket going now.. right now. Cause after 14 days they will give you a refurbished card.

If you ordered from evga, you only have a 14 day return policy.

So start it now while ya can.. keep working on it but. In the mean time hopefully they can start the exchange now.. with a new card.

Update, memtest has throw an error (already), test 7, cpu 14. 
 
Despite my careful planning, the pc gods have seen fit to bless me with faulty ram.
 
The description states the ram is at 4800 mhz; if this is the case it failed BELOW SPECS (well, XMP specs, that is).
 

wanescotting
Yeah I ordered from EVGA.
Okay support ticket entered, but it seems I cannot complete the RMA as it requires a support ticket number (which I do not yet have)

Weekend so assuming wait till Monday for a response.

If it is a mem issue.. test one stick at a time.to narrow it down if it is one of them.

Okay port royal passed all 20 loops.
 
I put ram back in xmp 2, fired up mem test and boom! error on test 8.
 
I'll test the sticks separately tomorrow. 
 
 

God I'm an idiot.
 
I selected the WRONG XMP PROFILE. I should have been using 1 not 2.
XMP profile 2 is not "Asus optimized".  once I switched to that profile: no memtest errors, no 3d crashes in any stress test. I realize now that memtest runs at the configured speed/setting buts only reports jedec speeds via spd.
 
I am on  roll today though, I figured out a nagging WMI process fatal error (Acronis need an update) and partially resolved a Corsair icue software thing.
 
Now I'm off to run dism and sfc...I did an inplace Windows upgrade with improperly configured ram...
 
Thanks for your help!