Kylearan

kristoferen
 
The only way I can understand that being relevant to me, is if the pcie slot power is hitting limit and thus the card hitting power limit - despite the 8-pins not being anywhere near power limit. IF thats the case, then what decides to use pcie power instead of 8pin and how do I fix it? If thats not it, then I'm still not sure how that post applies to me... Sorry for being thick. My PSU has each 8pin cable is on its own 12v rail and each rail is rated at 40amp/480w.
 
 
FWIW the 3080 FTW3 Ultra w/ XOC bios hit 440w no problem

What decides is probably the vBios itself.  I highly doubt power balancing is hardwired.  But modding the PCIE Slot shunt fixes the early throttle for everyone.
However please read Drunknfoo's post last page.
 
He implied that the PCIE shunt has continuity with MVDDC (Actually I found this out myself, by seeing people who weren't modding PCIE on their Founder's Edition cards, but modding all other shunts, having their MVDDC Report sky high like 238W, then when the only change they did was modding PCIE Slot after, MVDDC dropped to about 120W), and he tested himself that "SRC" shunt has continuity with 8 pin #1 and 8 pin #2, and 8 pin #3 has continuity with both "GPU Chip Power" and "SRC".
 
The rest is up for people with multimeters and soldering irons to test.
 
There are also two "mini" 005 shunts on FE cards, linked to GPU Chip Power and PCIE Slot.  No idea what they do--they do NOT affect main TDP--they might affect the "strange" random throttling at high loads...
 
https://www.overclock.net...0/page-9#post-28668863

To add and stress...

The beta was able to draw 500w and 490+ continuously without any hardware mods in my taiwan based card. (But extremely rare to replicate)

The strix bios likely the same... When i observed the killawat. (Granted when i first tried the strix, i was only paying attention to gpuz without realizing no 3 pin readout)

You guys need to give up on beta until another build is released

Some of us have been saying this since almost day 2

drunknfoo

To add and stress...

The beta was able to draw 500w and 490+ continuously without any hardware mods in my taiwan based card. (But extremely rare to replicate)

The strix bios likely the same... When i observed the killawat. (Granted when i first tried the strix, i was only paying attention to gpuz without realizing no 3 pin readout)

You guys need to give up on beta until another build is released

Some of us have been saying this since almost day 2

You mean you have a RTX 3090 Strix or you flashed the ASUS BIOS to the FTW3 UG ?
 
Highly doubt that the solution will be a new firmware, for me EVGA will come up with another PCB another batch, wish they could solve it with just an update but more than a month now that's not good at all.

lobstar
 
 
I've rarely seen my card over 60c under load.  I've rarely had my ambient temps more than 22c.  The rest of my test bench is water cooled so very little latent heat from other components.  I have directed two 140mm noctua industrials at the front and back to ensure no pockets of hot air exist.  Nothing changes the 450w max power usage on the xoc bios.  I can get my card to hit 72c if I add 20mv, underclock the gpu to -64 with zero mem OC.  The resulting scores are abhorrent.  If I OC to 1448 mem and 240 gpu temps stay around 52-60c and I get much more normal (but still low for a 3090) scores. I still think some nerd hand tuned the tables in these things and didn't know what they were doing.

Yes we said these many times before IF you manage to keep it up to 60C on all sensors it behaves more stable, over that it acts up, but this is another problem that adds up to the XOC Beta BIOS. As I told above I use the XC3 BIOS for gaming and when I game some hours and the card can't get cooled even showing 63Cish...as I'm on Air I have to turn off all and let it cool to see stable Clocks again, otherwise will start with the jumps, not as unstable as XOC.

Got a new card via RMA today, Taiwan card. 3090 ftw3 non ultra with the 500w bios. If temps were lower, maybe it would have drawn more power.
 
https://i.imgur.com/YQHniDv.png

Was playing around with different bios today (mainly for a separate issue) noticed that the scores on port royal seems much lower than before? Anyone else have this issue or is the card just deteriorating lol

I’m back on the standard bios. Running the same sweeps I have documented before and I’m about 300 points lower consistently. Clocks are the same but scores are lower. It seems something changed

Can confirm, even with other BIOS the card power throttles due to pcie slot power draw. 

drunknfoo

andressergio

drunknfoo

To add and stress...

The beta was able to draw 500w and 490+ continuously without any hardware mods in my taiwan based card. (But extremely rare to replicate)

The strix bios likely the same... When i observed the killawat. (Granted when i first tried the strix, i was only paying attention to gpuz without realizing no 3 pin readout)

You guys need to give up on beta until another build is released

Some of us have been saying this since almost day 2

You mean you have a RTX 3090 Strix or you flashed the ASUS BIOS to the FTW3 UG ?
 
Highly doubt that the solution will be a new firmware, for me EVGA will come up with another PCB another batch, wish they could solve it with just an update but more than a month now that's not good at all.

I have a ftw3 ultra

Ok so what ASUS BIOS are you using? I saw the EVGA XC3, the EVGA XOC, the EVGA KPE and Gigabyte but not ASUS, how it works or compare with the others we tested? Sorry if I missed something.
 
Thank You!

Anyone get anymore info on this? For an $1800 card, this sucks.

green4191
Anyone get anymore info on this? For an $1800 card, this sucks.

Unfortunately we are just waiting. Some cards seem to work fine, some limited to 450w and below. 

I purchased a 3090 FTW3 Ultra direct from EVGA yesterday. I should get is Tuesday so I‘ll report the production location, and power draw. I used to have the 3080 and had no issues with achieving 450w with its XOC Bios


I have an Asus Strix X570-E MB, 5900x cpu, and 850w PSU. 

sweet

QueueCumber
TBH, I actually scored 100 higher on PR after the updates. That is too small a difference to be meaningful, but it illustrates my point...

I scored 14250 early last night.  14647 early this morning...

PadinnPlays
I got my RMA card (3090, 2014 series, black accents) since my old one blew something. This one does peak at close to 500w in Superposition 8k. Temperatures around 82c and I'm drawing near 700w total per my kilowatt, so its working better. Wasn't a sustained 500, but definitely better than my original card.
 
 
 
 

Your card is way too warm.  Did you leave your fans on default? My max temp in SP 8K is 61C (fans at 90%) and the score is 7852. The power draw is below 450W.

vlcapata
best I've done is a 145xx at like +200 and +1400 and still on air with like 450w pull
 
what are you at?

3090 FE with shunt mod.  +150 core +500 Vram.
 Scored 22, 247 Timespy Graphics score right after that. 
Correction: PR score was 14,658.
 
The above results are "good" scores.  The "bad" scores are the ones 300-400 points lower.
 
Yet under almost same conditions, I've had PR as low as 14,250 and Timespy as low as 21,840.  Same clocks.
And no, it's not temps.  The most temps have ever cost me are 100 points.  When I run the tests at the same time, they are extremely consistent.
If I do something like play a video game, Overwatch, Modern Warfare, or somethinig, then come back to run the test, I'll score wildly different (could be good scores or bad scores).
The interesting thing is, if I instantly shut down and turn off the computer, then turn it back on and load windows, I'll STILL score the bad scores if I was scoring bad before the shutdown.
And it's not power limit either.  If I had to take a guess, I'd say it's getting "Drivered".  Maybe something to do with the shader cache or which segment of VRAM is being used...