So i can add 5-10mhz more on the core clock without crashing which gives me small gains but definitely better stability

sirien
Looks like Frame Chasers got it to work
 
https://www.youtube.com/watch?v=i8BAb3KzX5w

It sounds like he's saying that there isn't a vBios EVGA can release to give us a higher power limit than 450W. Am I understanding this correctly? It sounds like hes saying the PCIe shunt mod is the only way to get around that power limit, as if its a physical design barrier and not a bios/firmware/software limitation.


 
I'm just really confused, because it looks like my 3090's PCIe power, according to HWinfo, is coming in around 82 watts at max. If you put that together with ~150w per 8-pin, that should yeild you about 530 watts. But the problem is my max 8-pin power is coming in at around 117, 123, 129 watts respectively.
 
Can anyone chime in on this? I understand why EVGA probably won't say anything specific here, as it would be sort of foot-in-mouth to admit their flagship card cannot compete with lower tier/competitor cards without a mod they absolutely could not or would not endorse.

Well, the way he did it seemed quite easy. One single shunt and you dont even need to take the card apart. And u dont even need to glue it on just stick it there and place a thermal pad and the compression will hold it in place. In theory it would take less than 5min from start to finish
That said, im not advocating shunting but im very intrigued by his results

So i played around with heaven a little bit and i did see board power draw around 475 watts. Unfortunately its not stable at that power draw and usually crashes.
Tbf im running half the window which makes things hard on the application

Also, i was able to set a new pr in Port Royal by actually LOWERING my voltage to the minimum...



Safe to say im a little confused. Not 100% sure its working but...maybe. Should do some more testing tmrw

arcky

EVGANewMember2719

arcky
I mean the fact that folks with MSI cards are able to flash theses EVGA bios and get 497w in TimeSpy (https://www.overclock.net/threads/official-nvidia-rtx-3090-owners-club.1753930/post-28655352) while we with actual EVGA 3090 FTW3 Ultras can't get over 450w with the same bios, says without a doubt that something is either wrong with the bios implementation on the FTW3 cards, or with the FTW3 cards themselves. There's no other possible explanation. @Jacob it'd be nice to have some acknowledgment that there is a problem and the team is looking into it, rather than some one-off screenshots of Quake 2. Happy to be patient considering it's a beta and all, but up until this point the reaction has sounded more like denial of a problem and pointing to other limitations which I don't think is going to lead to an adequate solution. Thank you, as always, for everything, and of all the problems to have, this is certainly a problem I'm happy to have (so long as it's eventually fixed ;) 

I can hit 500w with FTW ultra with this bios.

What was your update process? Which bios on which switch? Which nvidia driver? which version of precision? can you link your precision settings? and can you link a gpuz showing the 500w hit with ideally a running graph of the power consumption so we can see how long you hit it during your benchmark?

Updated the OC bios on the OC switch. Precision sucks I use Afterburner. Here is a recent run with Firestrike Ultra. Can hit 500w peak with Timespy usually. 
 

ehabash1
I want to test with timespy and HWinfo but my wife was annoyed im spending all day on the comp so ill have to test tmrw

.

uninstall/install px1 ,seems to raise a little watt
but still not reach 500w

Tried the afterburner, but result is same with px1 on me
when run time spy even worse than px1,only get the half scores.
 
 
 

Hi Jacob, I get an ~6.5% performance increase in 3Dmark using Afterburner instead of Precision X1... Same settings. Jayz2cents also reports this in one of his videos. Can your team please look into this? I would like to use your program, but, a performance hit like that is not worth it. 
 

Can't wait for the non ultra version!

Chamidorix
Also want to point out that for shunting, you can literally just hotglue another resister on top of the existing one. No need to solder like all the techtubers driverlessly repeat. As long as there is some metal contact on both ends of the shunt that is more than enough for the entire current to flow through. You don't need full surface conductivity like soldering would provide for shunts to work. So shunting isn't that bad and reversible. 
 
The fact that EVGA hasn't replied at all to the widespread problems after getting some proof that the pci-e limit is a problem is very telling. Give it another day or two for them otherwise and I'm off to sell the FTW.
 
Glad I could provoke Framechasers to do the shunt and give us this good hard data. 

As long as you can clamp them all in place between the heat sink/PCB/back plate... Otherwise, I'd make sure to use a high temp liquid phase hot glue so they don't potentially move when the card heats up. Even then, the hot glue will expand when it heats up potentially creating a gap between the conductors... I would definitely clamp them in place with non-conductive thermal pads if using hot glue.

Chamidorix

vexorian87
Having no issues with mine, peak I've seen is 497W.
 
" />
 

Just want to point out this is another example of a non ftw3 ultra board working fine with the 500W bios.

What PSU do you use ? What are the readings per pcie power pin and pci slot power etc ? Which driver for Nvidia ? How did you apply the bios, first ultra then xoc utility right ?

I will try to replicate your settings on my 10980xe 5ghz/seasonic 1000w psu.

tayback longleg
 
One of the comments on that video mentions something very important I think. Something along the lines of testing performance of stock BIOS using the shunt mod to verify what is actually happening. What if this shunt mod reaps the same exact results regardless of BIOS?
 

 
 
 
Exactly. From what I saw in frame chasers video, I'm thinking you can get better power balancing between the PCIe slot and 8 pins by modding all of the shunts. He claims to be drawing around 120W from the PCIe slot... modding only the one resistor. If you replace all of the 005 shunts with 004 shunts you should end up with a max power draw of  525W @ 100%, with a draw on the PCIe slot of about 94W. Of course, this calculation needs to be validated with real life measurements. I would definitely not recommend drawing 120W from the PCIe slot on a regular basis, that is 60% over spec. Speaking of that... Anyone know what the safety factor for current draw on a PCIe slot is?
 
 
 

sky5600 
Exactly. From what I saw in frame chasers video, I'm thinking you can get better power balancing between the PCIe slot and 8 pins by modding all of the shunts. He claims to be drawing around 120W from the PCIe slot... modding only the one resistor. If you replace all of the 005 shunts with 004 shunts you should end up with a max power draw of  525W @ 100%, with a draw on the PCIe slot of about 94W. Of course, this calculation needs to be validated with real life measurements. I would definitely not recommend drawing 120W from the PCIe slot on a regular basis, that is 60% over spec. Speaking of that... Anyone know what the safety factor for current draw on a PCIe slot is?

I think its obvious a full card shunt will give you the best results, as the whole point with that is to remove power limitations regardless of vBios limits. The point of just shunting the PCI-E is to show why just upping the total board power limit in software isn't of much practical use. 
 
You can observe that the 8Pin power draw actually goes down when shunting the PCI-E. This is because the card still has shunts reading on each of the 3 main power rails: core, uncore, mem. So the board still knows it is hitting 500W total power and stops more power draw from 8pins. The issue is that the 8 pins are locked into a fixed ratio with the PCI-E power draw. This is what we are mad at EVGA for. I think it is purposefully done to forcefully segment the 3090 FTW from the 3090 Kingpin, otherwise you could just flash the KPE bios and get most of the performance advantage. This is what has been done with past generations and I'm sure they are not a fan. 
 
To truly unlock the card you have to shunt the 8 pins, so total board power is uncapped, then shunt the pci-express, so the 8pins don't drop down to their fixed ratio at 75W on pci-e, and then also shunt the 3 main power rails since on unlimited board power draw you will very easily hit the individual rail power limits. 

Interesting. Can someone who has this working tell us which bios they had prior to to upgrading with the exe file ?

My update path :

94.02.26.80.D0 to 94.02.26.48.F8

Is there any other ultra bios prior to the exe update that is different than D0?

I just wanted to come here and thank @EVGA_JacobF.

The 500W BIOS works great!... on my MSI 3090 Gaming X Trio.

:)