ckelas

firstnomad
Just chiming in here to add my experience: I purchased secondhand, but BNIB, an early-run 3090 FTW3 Ultra with red lips and a serial number beginning with 2014 (manufactured in Taiwan). It exhibited many of the same characteristics discussed in this thread and was perhaps one of the worst samples in comparison to those I've seen here - it would struggle to pull over 390W even on the stock OC BIOS with the power limit at 107%, and capped out at maybe 80W on the third 8-pin. The XOC BIOS only made it worse.
 
I called and spoke to a very accommodating service representative who wasn't familiar with the problem but was happy to hook me up with an RMA after a lengthy and detailed explanation of why power delivery isn't supposed to work like this.
 
Fast forward to today; I received the RMA - black lips this time, but still serialed 2014 and manufactured in Taiwan. It works perfectly - on the stock OC BIOS with power limit at 107% it sits right around 445W in Heaven with spikes up to ~460W. I honestly don't even have any inclination to flash the XOC BIOS at this point because heat is now my primary concern. 
 
I highly recommend anyone experiencing this problem just RMA their card and roll the dice on getting one that actually performs to spec. This is clearly a hardware problem with the early runs.

Thank you for sharing your experience bro. 
 
Can I please ask you to elaborate on your liaising with the representative?
What was it that you had to explain to them, and what did you tell them?

I simply explained its behavior: The PCIe slot was pulling power over spec (in my case, 83W; 8W over the standard of 75) while the third 8-pin was pulling a good ~40W lower than the other pins. Also in my case, the power limit was being hit while monitoring tools were still showing values under the appropriate setting. E.g. When the limit was placed at 107%, the power limit would engage with Precision still showing 97-98% consumption. I basically made the case that the 35-40W being lost off the third 8-pin pretty much directly corresponds to the proportion of power that is being "missed" by the monitoring tool, and furthermore suggested that there may be a blown sensor preventing the correct reading and affecting vBIOS behavior. I also mentioned the fact that this exact issue has been plaguing people on their forums, and that hardware revisions seem to be a factor.
 
I had called during after-hours support and the rep initially suggested that I could call back during regular hours the next day and speak to a technical service lead, but also conceded that it seemed like I knew more about the issue than he did and accepted my request to go ahead with the RMA regardless. He even upgraded it to a cross-ship RMA even though I have a secondhand card, and the new, correctly-functioning one got here in time for Cyberpunk.
 
 

firstnomad

ckelas

firstnomad
Just chiming in here to add my experience: I purchased secondhand, but BNIB, an early-run 3090 FTW3 Ultra with red lips and a serial number beginning with 2014 (manufactured in Taiwan). It exhibited many of the same characteristics discussed in this thread and was perhaps one of the worst samples in comparison to those I've seen here - it would struggle to pull over 390W even on the stock OC BIOS with the power limit at 107%, and capped out at maybe 80W on the third 8-pin. The XOC BIOS only made it worse.
 
I called and spoke to a very accommodating service representative who wasn't familiar with the problem but was happy to hook me up with an RMA after a lengthy and detailed explanation of why power delivery isn't supposed to work like this.
 
Fast forward to today; I received the RMA - black lips this time, but still serialed 2014 and manufactured in Taiwan. It works perfectly - on the stock OC BIOS with power limit at 107% it sits right around 445W in Heaven with spikes up to ~460W. I honestly don't even have any inclination to flash the XOC BIOS at this point because heat is now my primary concern. 
 
I highly recommend anyone experiencing this problem just RMA their card and roll the dice on getting one that actually performs to spec. This is clearly a hardware problem with the early runs.

Thank you for sharing your experience bro. 
 
Can I please ask you to elaborate on your liaising with the representative?
What was it that you had to explain to them, and what did you tell them?

I simply explained its behavior: The PCIe slot was pulling power over spec (in my case, 83W; 8W over the standard of 75) while the third 8-pin was pulling a good ~40W lower than the other pins. Also in my case, the power limit was being hit while monitoring tools were still showing values under the appropriate setting. E.g. When the limit was placed at 107%, the power limit would engage with Precision still showing 97-98% consumption. I basically made the case that the 35-40W being lost off the third 8-pin pretty much directly corresponds to the proportion of power that is being "missed" by the monitoring tool, and furthermore suggested that there may be a blown sensor preventing the correct reading and affecting vBIOS behavior. I also mentioned the fact that this exact issue has been plaguing people on their forums, and that hardware revisions seem to be a factor.
 
I had called during after-hours support and the rep initially suggested that I could call back during regular hours the next day and speak to a technical service lead, but also conceded that it seemed like I knew more about the issue than he did and accepted my request to go ahead with the RMA regardless. He even upgraded it to a cross-ship RMA even though I have a secondhand card, and the new, correctly-functioning one got here in time for Cyberpunk.
 
 

Thank you so much for elaborating on this. 
I am nowhere near to that level of knowledge and I literally wouldn't know how to word the problem.
Now I guess I have to wait a few more hours, get my card delivered and try it. 
Will be back with figures. 
Thank you again bro. 

firstnomad
Just chiming in here to add my experience: I purchased secondhand, but BNIB, an early-run 3090 FTW3 Ultra with red lips and a serial number beginning with 2014 (manufactured in Taiwan). It exhibited many of the same characteristics discussed in this thread and was perhaps one of the worst samples in comparison to those I've seen here - it would struggle to pull over 390W even on the stock OC BIOS with the power limit at 107%, and capped out at maybe 80W on the third 8-pin. The XOC BIOS only made it worse.
 
I called and spoke to a very accommodating service representative who wasn't familiar with the problem but was happy to hook me up with an RMA after a lengthy and detailed explanation of why power delivery isn't supposed to work like this.
 
Fast forward to today; I received the RMA - black lips this time, but still serialed 2014 and manufactured in Taiwan. It works perfectly - on the stock OC BIOS with power limit at 107% it sits right around 445W in Heaven with spikes up to ~460W. I honestly don't even have any inclination to flash the XOC BIOS at this point because heat is now my primary concern. 
 
I highly recommend anyone experiencing this problem just RMA their card and roll the dice on getting one that actually performs to spec. This is clearly a hardware problem with the early runs.

 
Thanks for the update! It sucks, both for me and for EVGA, that I have to go through an RMA just to get base power corrected.

Look it here everyone.... you think with all the phone calls, emails and this thread that EVGA isn't looking into this? Or that businesses know 1 upset customer tells 10+ people about the bad experience and that a happy tells maybe 3??? And with how EVGA has been around with supporting their customers with products that can overclock and still stands behind their warranty...... all they really have to provide is a card that works at STOCK setting!! There is no guarantee that you are gonna have a world record card or a card that can only handle stock settings! A lot of this is luck! I knew a guy back in the day that had a intel 920 chip that clocked over 5ghz on air alone and I went through 5 to find 1 that could just touch 4.7ghz without crashing.... so what ya it appears the 3090 ftw3 ultra has something going on with it and ya I agree because my personal experience with mine makes me thing it's the thermal pads and putty.... but until there is an official EVGA statement or announcement.... let's keep the frustration under control because atleast we have a 3090 and there are a whole bunch more people that wish they were in our shoes having a card! Have some faith.... sit back and leave your card STOCK and see what happens! Yes long and fan boy-ish but you didn't buy an EVGA product because they have a bad reputation..... I'm sure they are working on it!

I think the problem lies in the lack of transparency. Some people, myself included, bought this card under the assumption that the EVGA supplied bios would work.

I have a limited return window. And with the limitation seeming like an issue with the PCI Power limit, it may be an issue at the hardware level that can’t be fixed without a new card.

What’s more frustrating is that people with a STRIXX or gaming x trip can fully utilise a bios not even designed for their card.

jl88jl88
I think the problem lies in the lack of transparency. Some people, myself included, bought this card under the assumption that the EVGA supplied bios would work.

I have a limited return window. And with the limitation seeming like an issue with the PCI Power limit, it may be an issue at the hardware level that can’t be fixed without a new card.

What’s more frustrating is that people with a STRIXX or gaming x trip can fully utilise a bios not even designed for their card.

This is a valid point.
This isn't a $699 video card.  If this were a $699 video card, people would not be flipping out so much.  But this is a $1799 video card!  As much as an E-bike or a really cheap car.  That's a LOT of money for some people.  And this is a high end enthusiast card (the Kingpin is for LN2/world records/competitive overclocking).  So people simply don't expect for these things to happen.  For something this expensive, it's like getting blindsided, almost like complete betrayal.
 
When was the last time something like this even happened?  (No I'm not talking about VRM issues or cards catching fire--something that was supposed to just work, not working and then it being blamed on "silicon lottery"?).  Everyone here knows it has nothing to do with silicon lottery.  The cards either work or they will crash.  Boosting is based on power limits(TDP and rail limits), not on SL.  How far your card will overclock with offsets is SL.

I totally agree something is going on! My personal experience so far was 200mhz 1400mhz on my card out the box with 430w max draw..... once I changed the TIM and pads..... I can maybe do a 170mhz and 1200mhz run much lower score and 460w during the run on port. In time spy pre TIM and pads I would peak 480w's and now during the run I average 460-470w with peaks in the 490w's.... it's strange to say the least but hey mine seems to be something with thermals? Does that mean they have a problem with the TIM, pads and putty? Who knows! And almost 2k for a video card with tax and shipping is a good chunk of money! I am frustrated with not having answers as well.... but until EVGA comes out with something to say.... continuing to point out the obvious isn't gonna make them move any faster

And I'm sure if EVGA finds a problem that is hardware related as much as it is going to suck to send our cards back in..... I'm sure that is what they are going to do

Kingpin 520W Bios:
 
https://www.techpowerup.com/vgabios/227017/227017
 
Not sure if this was posted yet.  Supposedly people over on OCnet are trying it on the FTW3.

awsome