gsrcrxsi
well I called EVGA support and as expected there was nothing I could do from here to troubleshoot the RGB. their suggestions were the same as what I've already tried (reseat the GPU, power cycle the GPU, etc).
 
The started a cross-ship RMA for me. apparently the crosship now requires 2x retail price lol, so I'll have to put 3k on a CC or something. which is fine. I don't want to do the loop teardown more than once so this was OK for me. at least they're paying the return shipping and providing me a label with the new card.

Hopefully the replacement works correctly. 

Hi, could you please post here your results after you receive a new card? I am also aiming for the hydro copper, so I'm slightly intrigued by this thread. Thank you.

RMA approved, $3000 collateral paid, shipment pending.

got the new card. and it seems to have fixed the issue. the RMA reason was RGB being non-functional, and the replacement works there, so RMA satisfied.
 
as far as power consumption. it seems a lot better, and in line with what xReason reported. now pulling 335-340W while running Kombustor (vs 310ish before), and running ~300W running furmark (vs 270W before). there's still some behavior preventing it from going to 350W i think, but it at least seems to be in line with the others now.
 
in general, I can see the card pushing a bit more clock, and pulling a bit more power. definitely something wrong with the old card. I'm almost glad that the RGB issue popped up to give me an excuse to RMA.
 
oh and even though the replacement card came in a retail package, the terminal top area (where the RGB logo is) had a good amount of scuffs on it, underneath the protective plastic film. probably why this one was set aside for RMAs...

I think it could be fixed in the BIOS, but EVGA doesnt seem interested in giving it any attention. since the XC3 owners are a small minority, there likely wont be enough users asking about this to get them to take any action.
 
RMA will be the only way. but,
 
1. you'll have to convince them it's a problem (they'll say things like "we don't guarantee power draw" which is an asinine oversimplification)
2. you run the risk of the replacement card performing the same way.
 
but the sooner the better. they'll pay for the shipping costs if you start the RMA within the first 30 days or so. I paid the 2x MSRP collateral because I wanted to ensure I get a totally separate card, and I only wanted to drain and fill my loop once (hydrocopper card). but you can do a standard RMA too, if you can deal without a card for like a week, they should still send you a shipping label though, make sure you ask if they don't mention it.
 
good luck.

Just noticed that the Normalized Total Power is maxing out at 100% while Total Power power is staying at 88 maxed. So whatever Normalized Total Power is, that's what's capping the card it seems.

Aruzedragon
Just noticed that the Normalized Total Power is maxing out at 100% while Total Power power is staying at 88 maxed. So whatever Normalized Total Power is, that's what's capping the card it seems.

I think that just indicates it's hitting a limit other than the tpd.

Power use is determined by the PCIe slot it seems. It’s like when it maxes the PCIe slot, it inadvertently throttles the whole card, which prevents more from being pulled from the PCIe cables even though there’s headroom. Seems to be a power balancing problem. EVGA should fix the bios to allow more through the PCIe cables or not throttle the GPU so early

Someone on overclock was saying the cards are actually pulling more than we think and none of the software is reporting it correctly. 

gsrcrxsi
apparently the crosship now requires 2x retail price lol, so I'll have to put 3k on a CC or something. which is fine.

The policy seems to be in flux, so day to day I don't know what the current policy is.  But, basically, the issue EVGA was having is that people would buy 1, do an RMA cross-ship to get a second one, scalp both, and even though they lost money on the collateral, they made money overall with the scalping.  So, the double collateral would have been used to prevent the scalping and bleeding of inventory.
 

xReason
Someone on overclock was saying the cards are actually pulling more than we think and none of the software is reporting it correctly. 

 
Is there any proof provided?

still no word from EVGA on this. multiple people with the same issue now.
 
We need one of the big Tech-Tubers to cover this. a head to head XC3 vs FTW3 (without any crazy BIOS, just the stock ones) would be eye opening and might put some pressure on EVGA to publically address it rather than sweep it under the rug.

Just got my 3953 email today. How much does this power issue affect the performance of the card?

Aruzedragon
Basically if your card has the power limited at 300watts or so, expect slightly below a 3080ti FE  which scores about 19000ish. My card is scoring at 18000ish back to back. There's no overclocking room for the card other than undervolting to get better power usages and clocks. You can check using hwinfo or gpuz how much power is being used up by the slots and the two 8 pins to see if it's affecting your card too.

So you have the power issue and you're seeing around 18000 vs the 19000 of the FE. How much power are you drawing during this benchmark? How does the card perform vs a FE or something in non-synthetic situations (i.e. games)?

Aruzedragon
It doesn't matter what you run on it, my card will pull 304watts max. Maxing out the pci slot at 67 watts and the two 8 pins at 110ish watts each. From what I am seeing, it's best to leave it at stock clocks and/or undervolt so it stops the sudden clocks spiking and dropping.

Ok, I won't overclock it then. How much of a difference does that 300 watt vs 350 watt power draw make? Does it make it bad enough that something like a 3080 ftw3 ultra could surpass it?

Aruzedragon

Cats4Life

Aruzedragon
It doesn't matter what you run on it, my card will pull 304watts max. Maxing out the pci slot at 67 watts and the two 8 pins at 110ish watts each. From what I am seeing, it's best to leave it at stock clocks and/or undervolt so it stops the sudden clocks spiking and dropping.

Ok, I won't overclock it then. How much of a difference does that 300 watt vs 350 watt power draw make? Does it make it bad enough that something like a 3080 ftw3 ultra could surpass it?

yeah, a 3080 FTW3 overclocked can easily get 19000ish scores

What about a 3070 ti ftw3 overclocked? I managed to get one of those from micro center and I'm deciding whether or not to buy the 3953.
 
(also what benchmark are we talking about? I'm unfamiliar with most of them)