I've said this before - there's multiple power limits in VBIOS for Ampere, not just the overall board power limit.  Each power rail/connector has its own power limit. Memory has PL.  Etc.  There was a utility that someone wrote to read the PLs but it doesn't work on newer VBIOS, at least it didn't for my 3080 Ti FTW3 Ultra.  Below is from my former 3080 XC3 Ultra card...
 
https://imgur.com/ufztgyD
 
Buying any 2 power connector 3080 Ti card is a mistake if you're wanting to extract extra performance from it.  Same goes for the 3090.  Even a basic 3080 is held back.  EVGA is gun shy with all the power issues causing 3090 failures so they're going conservative now on limits...  hitting any one of these PLs will trigger Power limit state.
 
You can usually figure out what's causing it if you look in Hwinfo.  On the 3080 Ti FTW3 Ultra 8-pin #2 is usually the culprit before the max board power limit can be hit.

kram36
No, the 3080 Ti FE card is basically a two 8-pin card and it shreds the XC3. Both cards have the same 350w power limit.

Its not the same card and you can't see what EVGA has programmed in VBIOS vs. Nvidia.  Different cards, different PCB designs.  The 350W power limit is just one of the many limits that can trigger the Power limit state.  And the entire power delivery components will affect all these readings that the VBIOS has to respond to.
 
If you want to say EVGA has a worse designed card that's accurate based on your observation but there's more to these cards than just that single 350W figure you're looking at.  It is what it is and nothing will happen, you can complain all you want but short of a total redesign there's no fixing this.

kram36
We are not getting the performance these cards are supposed to get, we yet paid the price for that performance. If EVGA wants to refund part of my purchase price, then I would be fine, but I'm not fine paying full price and not getting the full price performance.

Get over yourself - we all know you paid the price for whatever card you could get that was available.
 
If you're not happy why don't you just return the card for a full refund instead?

kram36
EVGA will charge a 15% fee on the return.

Sucks to be you then...  you bought a card without doing the research to know how it would perform first based on real world results from others.  Enjoy your four cards and whatever could anyone be doing with 4 cards anyways - oh yeah.  Never mind.  Only things people who can't find a better way of making a living does.

neteng101

kram36
EVGA will charge a 15% fee on the return.

Sucks to be you then...  you bought a card without doing the research to know how it would perform first based on real world results from others.  Enjoy your four cards and whatever could anyone be doing with 4 cards anyways - oh yeah.  Never mind.  Only things people who can't find a better way of making a living does.

to be fair, there was no way to know about this issue before purchase. all reviewers were sent FTW3 cards or FE cards. no 3080Ti XC3 reviews. and personally I'm space-constrained, a watercooled 3-plug card will not fit in my O11D while keeping horizontal mounting. 
 
the FE being a non-standard design lets EVGA get away with the fact that there's no baseline/reference performance. so they don't really have a base floor level of performance to live up to. makes it harder to prove if a GPU isn't meeting performance expectations.
 
I don't agree with kram's torches and pitchforks approach, but I do think there's something going on that at least deserves EVGA's attention and comment. they should at least comment that they are looking into, and comment if something could be done, or conversely if something can't be done. we're all speculating about what the issue might be, and I have my own theories based on my own tests and observations, but so far EVGA hasn't even commented or attempted to look at their own BIOS to see if it could be adjusted. If they say "we looked at the BIOS, but unfortunately nothing can be done because of X reason or X design constraint", I would more or less be satisfied. I just want to know the exact reason for the inconsistent power limiting behavior with different loads, and if my assumptions about memory controller load being tied to the PCIe slot are correct. unfortunately the EVGA people we mainly interface with on the forums, are more PR and Marketing folks. they always have to send things up the chain to the more technical folks for technical issues. getting in contact with them is difficult, as most of them are probably in Taiwan.
 
I'm not planning to swap my card out because the card I received from RMA at least performs a little better than the first one i had and it's a 3080ti HC card that actually fits in my case.

neteng101

kram36
 There where no real world results on this card. We are the first people to get them and EVGA should fix the issue these cards are having.

Except its not an issue - what you bought is a 3080 Ti - there's no guaranteed performance guarantees that comes with any GPU.  And since you didn't wait for real world results you can either choose to live with the results you got or return the card with the restocking fee.  That's the price you pay to be on the vanguard and bleeding edge.
 
I'm done here talking to you - please continue your childish tirade and pouting in hopes that EVGA will fix your card for you - they simply won't and nothing will happen.  Sorry but the truth hurts.

No, you don't settle for lower performance than the card should be giving. The 3080 Ti FE card is outperforming EVGA's 3080 Ti XC3 Hydro Copper that cost $1,551.40 with tax and shipping cost. Yup, your the sucker born everyday. Yeah, you're done talking because as I said before, your talking out your backside.

So basically the evga 3080 ti cards are 3080 performance at 3080 ti price? Got it. 

lan2cp
So basically the evga 3080 ti cards are 3080 performance at 3080 ti price? Got it. 

For me it's only on the XC3 model that's under performing, the FTW3 model runs like a champ.

Just because I’m curious. What happens if you lower the memory clock speed. Run it at 0 memory offset. Keep the +165 core offset.

I want to know if it’ll clock higher or pull more power if you reduce the memory load.

kram36

gsrcrxsi
Just because I’m curious. What happens if you lower the memory clock speed. Run it at 0 memory offset. Keep the +165 core offset.

I want to know if it’ll clock higher or pull more power if you reduce the memory load.

Do you care which bios is used?

not particularly, since it doesn't seem to matter.

kram36
I did some playing around with changing the bios. I loaded the Zotac bios gsrcrxsi linked, the Hydro Copper bios xReason dumped and put the Hybrid bios back on. I ran Port Royal and Fire Strike Extreme on each bios. I had to use Afterburner as PX1 would not allow me to adjust the fan speeds with the Zotac or Hydro Copper bios. Yes I now the rest of the system isn't giving the best score possible, but there are differences in the score depending on the bios I used. In Afterburner I set the GPU to +165 the Memory to +1100 and these are my results.
 
Fire Strike Extreme
 
Zotac bios 12660 Hydro Copper bios 12392 Hybrid bios 12300
 

 
Port Royal (Note: Not positive I had the Zotac bios run GPU clock set the same, it's showing a lower Clock Frequency, but had a better average clock speed)
 
Zotac bios 12761 Hydro Copper bios 12527 Hybrid bios 12483
 

 
However, looking at the power logged by GPU-Z, I don't see a difference in total board power, pcie slot power or the pcie plugs power.
 
 

Dang! So it looks like some improvements could be made with bios tweaks from EVGA!
I wonder how they could be motivated to revisit their bios design, or if they would consider whatever the ZOTAC bios is doing differently to be unsafe for sustained functioning. 
 
Regarding all the memory discussions, I read these last night and played with underclocking the memory. It didn't lead to more power draw, and generally led to slightly lower scores in benchmarking. 
It did lower memory temps which seem to cap at about 80c. 
 

Kokin
Well if anyone who has the 3080Ti XC3 and has a lot of time on their hands, I'm still interested in seeing how power limits behave for 1080p/1440/4K in several games (of your choosing). 

Firestrike isn't a perfect gaming representation, but it is 1440p, Firestrike extreme is 4k. Running either on my card shows similar power draw. 
It seems to depend on workload more than resolution. The max power that's been represented by gpuz on my card is 330W.
 
 

Yeah, I think the memory controller is the main villain in this case. During Witcher 3 gameplay, during the cutscenes where the GPU is maxed out but the load on the memory is pretty low, the power usage easily goes into 330ish or a bit higher.

gsrcrxsi
the reason I ask, is that when I run Einstein@home (which has a good bit of reliance on the memory subsystem), the card runs pretty low average clock speed ~1665-1725 core and ~285W when the memory is clocked at 19.5Gbps, and this gives quite good performance because Einstein really benefits from fast GPU memory/bandwidth
 
if I remove the memory OC, and run the card at stock memory speeds (compute = P2 clocks = 18.5Gbps), suddenly average core clocks increased to ~1710-1800 range, power draw increased to ~290W, but actually worse performance because of the nature of this computation.
 
ideally I'd want to be able to run the fast memory, AND high clocks and have the option to pull all the way to 350W if I wanted. but I get the feeling that pushing the memory hard is the weakest link and causes hard throttling when under load.

What memory temps are you seeing in this workload? I didn't find a strong correlation between the core clock and the memory clock/temp with synthetics. I was pretty surprised to see 80C on the memory with the core at least 10c lower. 
 
I think it's pretty clear why they added a third 8 pin power connector on the FTW3 cards.

Cats4Life
Seems like the memory controller and the memory is a major issue for the card. Has anyone tried to see how the card runs with very high loads on the VRAM?

The card is not getting the 350w. Why? I don't know, but it's surely not because of temperature on the GPU or Memory. The Zotac bios might have better timings on the memory and scored better. This card is a head-scratcher for sure.
 
Sure wish EVGA would give us an answer on what's going on with these cards.

sentient12

gsrcrxsi
the reason I ask, is that when I run Einstein@home (which has a good bit of reliance on the memory subsystem), the card runs pretty low average clock speed ~1665-1725 core and ~285W when the memory is clocked at 19.5Gbps, and this gives quite good performance because Einstein really benefits from fast GPU memory/bandwidth
 
if I remove the memory OC, and run the card at stock memory speeds (compute = P2 clocks = 18.5Gbps), suddenly average core clocks increased to ~1710-1800 range, power draw increased to ~290W, but actually worse performance because of the nature of this computation.
 
ideally I'd want to be able to run the fast memory, AND high clocks and have the option to pull all the way to 350W if I wanted. but I get the feeling that pushing the memory hard is the weakest link and causes hard throttling when under load.

What memory temps are you seeing in this workload? I didn't find a strong correlation between the core clock and the memory clock/temp with synthetics. I was pretty surprised to see 80C on the memory with the core at least 10c lower. 
 
I think it's pretty clear why they added a third 8 pin power connector on the FTW3 cards.

No idea on memory temps. I run Linux and memory temp monitoring is not available here. but when running Furmark in Windows, which had a fairly high memory controller load (~60%), memory temps were about 66C.
 
I never claimed that memory TEMPERATURE was any culprit. just memory LOAD. if i'm right that the memory or memory controller is getting power from the PCIe slot only, then more load = more power draw = hitting the PCIe slot power limit earlier under heavy memory load situations, with power limiting in place, it restricts core clocks (doesn't pull back on memory at all) and reduces overall power use, even if there's headroom. that is all just an estimated guess based on observed behavior so far.