sirien
109 pages and no response from evga on the issue

I see 37 pages, but either way I'm pretty sure there was a single response stating they were looking into it (the bios issue). 

Lol what does that mean hahahaha

2080Ti can't pull that much power simply because of the limited power connections. Do the math!! 2 eight pin connections and a PCI-e slot....

500 watts who are you?!?! I've got a 2080ti and it can't pull 400 because of its design...

LVNeptune

evgatorator

LVNeptune
Doesn't show up like that in the regular build.
https://i.imgur.com/Qxq0e0Q.png

Only shows % no wattage values.

Change that drop-down menu on top from Original to Nvidia BIOA

I've used GPU-Z forever and I didn't realize that was a dropdown. :D
 
I see, 420W default, 500W maximum. I guess I just really need a waterblock to take full advantage. I am really confused why GPU-Z shows Pwr being hit when it's well under 500W though.

Guys ever think that the "software" might need to be updated in order to properly recognize the hardware??

When was GPU-Z updated last? Did they release something new for the 3000 series cards?

Software like Aida64 usually updates when new hardware drops it just takes a few days maybe...

IMO the only disappointing thing about this card is the temp.  Why does it run so much more hot vs. other AIB cards?  Even the Gigabyte cards run cooler and they are usually the worst.
 

Jmalto7
IMO the only disappointing thing about this card is the temp.  Why does it run so much more hot vs. other AIB cards?  Even the Gigabyte cards run cooler and they are usually the worst.
 

The copper plate is also a factor - Gamers Nexus explains this. It is not a bad thing though. It is able to keep area around GPU cooler then other cards.

Akriant +80 on chip clock and +1000 on memclock. 

A recommendation - you are probably losing 2-3 FPS and some power draw by putting the memory into a full-time error correction state...  Back that mem OC to +250 and you will probably see some gains and at minimum absolutely will see some stability improvements in frame render time consistency.   This memory loses performance above approx +250mhz.  And that's the same for all cards, all temps, GDDRx6 is not good at overclocking without using ECC. 

kring

Akriant +80 on chip clock and +1000 on memclock. 

A recommendation - you are probably losing 2-3 FPS and some power draw by putting the memory into a full-time error correction state...  Back that mem OC to +250 and you will probably see some gains and at minimum absolutely will see some stability improvements in frame render time consistency.   This memory loses performance above approx +250mhz.  And that's the same for all cards, all temps, GDDRx6 is not good at overclocking without using ECC. 

orkan

kring

Akriant +80 on chip clock and +1000 on memclock. 

A recommendation - you are probably losing 2-3 FPS and some power draw by putting the memory into an full-time error correction state..  Back that mem OC to +250 and you probably see some gains and stability improvements in frame render times.  This memory loses performance above approx +250mhz.  And that's the same for all cards, all temps, GDDRx6 is not good at overclocking without using ECC. 

I had to back my memory down to +50 before it would stop with error correction overhead. Precision X1 scan recommended +50 on mem and +110 on core... and sure enough, that is the highest it will go without scoring lower... despite being "stable" at higher clocks on both core and mem. 

Use a benchmark that can be paused or a game scene that maintains a stable frame rate, then adjust memory OC while watching frame rate and memory power draw.

kring

Akriant +80 on chip clock and +1000 on memclock. 

A recommendation - you are probably losing 2-3 FPS and some power draw by putting the memory into a full-time error correction state...  Back that mem OC to +250 and you will probably see some gains and at minimum absolutely will see some stability improvements in frame render time consistency.   This memory loses performance above approx +250mhz.  And that's the same for all cards, all temps, GDDRx6 is not good at overclocking without using ECC. 

 How can you tell if its using ECC or not. Is there something in px1 that tells you?

Basgolfer

cerealkeller

kring

Akriant +80 on chip clock and +1000 on memclock. 

A recommendation - you are probably losing 2-3 FPS and some power draw by putting the memory into a full-time error correction state...  Back that mem OC to +250 and you will probably see some gains and at minimum absolutely will see some stability improvements in frame render time consistency.   This memory loses performance above approx +250mhz.  And that's the same for all cards, all temps, GDDRx6 is not good at overclocking without using ECC. 

Can you explain this in a little more detail? I’m not quite understanding you fully.

I'm not an expert, but this is how I understand it. 
 
GDDR6X memory (Used on 3080 and 3090) is a type of ECC (error correcting code) memory. If you push the clocks to high on the memory then you will start introducing memory errors. Since GDDR6X is ECC memory it will realize it's an error and try to fix it. This obviously takes time, and if you push the memory to far, you should actually see a decrease in performance (fps, frametime consistency, etc.) because the further you push it, the more errors you will introduce. You need to find the sweet spot for your memory. 
 
I would run a benchmark and leave everything the same except the clocks on the memory. At some point you'll start seeing a decrease in performance because your memory has to keep fixing the errors that are getting introduced by running it to fast.
 
ECC memory is usually used in data center applications where reliability and uptime are the most important things. Not sure why they decided to put it on these GDDR6X modules. I think this is the first generation of consumer grade cards to feature ECC memory. 
 

Exactly what I was looking for as well! Thank you for the explanation!
 

Sure.  basically the default clock speed is what they certify the memory to run at under high temps and conditions normal to the card...  in some cases, it can be overclocked if it's good memory and clean power... that's your sliders in PX1 to try to force the memory higher, which unlike the GPU isn't self-over-clocking or variable clocking.
 
Now in ALL prior video cards used from GDDR5 to GDDR6... the memory is NON-ECC,  which means if there is an error in the memory you will get an impact anywhere from couldn't tell to HARD system crash.  early memory faults can look like sparkles, noticeable artifacts in game and odd colors, push it a little farther and you'll get game lockups, crashes and black screens.
 
For the first time, the 3000 series comes with GDDR6x, which brings with it ECC.  which is "Error Correcting Code"  and has the capability to recover from minor errors.  it does this by actually detecting the memory error, pausing, going back and re-reading/writing, checking again... and in cases of this memory it seems we can still run an overclock in the +1000 to +1200 where these errors occur but the memory can still recover and not "crash" the system or show artifacts... so this is REALLY GOOD!
 
Now for the specifics,  each time ECC catches an error that cycle I mentioned STOPS the entire processing cycle for say 1-3ms.. now if these are occurring regularly on every frame those pauses add up and reduce your FPS.
 
The sweet spot is the MAX overclock where ZERO ECC error correction cycles occur.. and so far that's between +50 & +300.... most seem to be safe around +150 to +200.
 
Does that help?
 
 

Regarding the memory...
 
Don't think I can post links yet so..
 
Youtube: JayZTwoCents
Video Title: RTX 3080 Overclocking... Holy S@&T
Published: Sep 23, 2020
 
Watch from: 14:40 to 16:30
 
He demonstrates pushing your memory to fast while benchmarking. 
 
 
 

 I dont know if you see the new driver ready for cold war also stop boosting like a rollercoster each time you change game and need adjust overclock or put lower clock for all game.
 I boost more equal just near 2000mhz and even i put some overclock i was crashing on shadow of tomb raider, now i dont crash anymore but my boost stay under 2000mhz, around 1990mhz.
 With new driver i didnt got a crash even i put +140 core and +1000 on memory but the card dont boost higher. Maybe they put something in driver for rtx-3090 over 50c wont boost over 2000mhz, i dont know but its a good thing if they have really resolve the crashing issue.

LordGurciullo

drunknfoo

cerealkeller
At least one other person mentioned the video on Frame Chasers YouTube page.  The way he described it, I'm inclined to believe he is correct when he says the way this card is designed, it's hard wired to limit power draw once the PCI-E slot hits 75 watts of power draw, which is why this BIOS is not working.  If that's not the case, it would be awesome if we could expect a fix.  But since the BIOS works on other cards, but not ours, it does seem to be the case.  I'd suggest anyone with an interest watch the video, I just searched "EVGA 3090 500w BIOS" and it popped up, I actually came across not even looking for it.  That being the case, it's unlikely this will ever be resolved for us who already have cards.  I hope I'm wrong though.

its not, i can avg 490w and sustain given some random condition sometimes and peak 520 without any shunt
 

 
DrunknFoo, did you use the original firmware with the xoc bios? It seems thats the only possible way to do this..
You are the only person able to do this as far as I know..
 
EVGA - please release a firmware and bios update so we can get on with our lives and have the experience Drunknfoo is having!!
 
Regarding memory - I used valley benchmark and noticed my mins drop crazy after 500 mem with the sweet spot for me being 275-300. 
Yes Port royal will take 1000 but I prefer higher minimums in actual fps gaming. 

that was on the updated firmware, as i mentioned previously, i was playing around with driver display settings that eventually pulled this.
as i stated earlier, it is difficult to replicate as some settings toggled i.e power / quality gave me random numbers
 
tried the newest drivers released and same behavior... other than layouts among pcbs, icx is the only unique characteristic of the ftw3, so maybe firmware could be culprit. /shrug
 

Hey Everyone,
I just received my 3090 ftw3 ultra yesterday. I was excited to see there was a 500w beta bios, but unfortunately did not read into the issues everyone is having with it. I have also not been able to get it to work. My card is unable to go over 400w in any benchmarks or games. Currently I have it OC'd at 150 core and 400 memory. Am I looking at a bigger issue with my card since it won't even tip over 400w. I am using a 1000w G3 with three separate PCI-e cables. Temperatures sit around 59-60c and GPUZ perfcap reason is pegged at pwr the entire time. 
 
It would be nice to have an update from EVGA about this.
 
Thanks everyone!