Dabadger84
The 1000W BIOS really doesn't "fix" the issue, just an FYI, all it does is make the card draw more power through the PCIe slot, which is dangerous & could lead to card failure later on down the line, or sooner, depending on what your computer feels like doing about it.  PCIe spec as was discussed a while back in this thread, is 75W.  The 1000W BIOS from what I've seen can get upwards of 90W draw from the PCIe slot (which is why it's giving you "better" power draw than the "500W" broken XOC BIOS).  I would not recommend using that BIOS myself, it isn't really fixing the issue, it's just creating another one.
The XC3 BIOS also didn't work for me, clocks didn't go any higher, weren't anymore stable, and scores were lower at the same clocks, so it looks like this card will be headed the way of the RMA department once I get my Kingpin card in.  Can't RMA right now because I don't have a viable backup card.
 
Also, if you're seeing 84C, your ambient is either too hot, or something else weird is going on like needing to repaste indeed, because I've never seen temps that high, not on the 3080 FTW3 Ultra I had before (topped out at around 65C under prolonged heavy load) or my 3090 FTW3 Hybrid now (tops out at 53-57C depending on the load & ambient temps).  Anything over 60C on a Hybrid card is too high imo, unless your ambient is greater than 80F.  That's an indication of either needing a remount/repaste, or horrible airflow.

Dabadger84

Kylearan
Does the new hotfix driver fix the "6500C" thermal throttle temp flag bug that's been present in every single driver since the very first 457.xx branch?

I can confirm the 461.09s does NOT fix that issue, I'm running the newest driver & had it happen on 2 sensors during a Cyberpunk 2077 session yesterday.  Haven't tried the HotFix ones yet.

Clayman31
 
How much faster is a strix? Isn't it 480w with worse warranty service?

arestavo

eXp67

pellaaa

eXp67

arestavo
Press F to pay respects.

Please post back here when your system crisps, as others would benefit from hearing about it. (People aren't joking about the no protections - the VRMs, the VRAM (this one seems the most likely, with the non-cooled VRAM stack on the back of the card), and everything else that there might be - the card will not stop until something gives out with the unlocked, warranty voiding 1000W Kingpin VBIOS)

You mean that temp that i reported was too much ?
I'll inevitably kill my card with the 1000W on aircooled ?
From 500W XOC to 550W unlocked, i see not so much difference with temp?
I never see the card bottlenecking because of too hot with the VRAM on XOC.
 
If you confirm at 100% that i'll kill my card, i'm doing a rollback now hehe.
 
Best regards,

its not killing the card actually, but if in future something goes wrong and temp get really high the card won't shutdown till it start burning because it doesnt have any protetction on

Okay.
I'll be aware of the temperature.
It is sure that with that setup, the GPU will not like a burning at 100% during long minutes/hours.
 
Will buy a EK ASAP and waterblock it. Will be my first watercooling EK :)
I needed to watercool the CPU (5950X) anyway which has good potential at stable 5ghz :)

Just make sure EVERYTHING that currently has TIM/pads on it has EXTREMELY good contact - and the back side VRAM will likely need a fan blowing on the backplate if you don't have water running through it (I saw 1 waterblock with active cooling on the backplate). Hot spots where icx sensors aren't at will kill that component, and again - with a warranty void VBIOS on the card, you have a $1800+ door stop when EVGA laughs at you for trying to RMA with that on it. Normal VBIOSs have limits on temperature as well that a user cannot see (each VRAM chip for sure, perhaps other things) - that's why using that unlocked VBIOS as a daily driver is insane (unless you literally have money to burn).
 
As to any guarantee that your card will or won't fry itself - I guarantee nothing. You're the one playing with fire, so you do you.

P4R4DiSi4C

eXp67

arestavo

eXp67

pellaaa

eXp67

arestavo
Press F to pay respects.

Please post back here when your system crisps, as others would benefit from hearing about it. (People aren't joking about the no protections - the VRMs, the VRAM (this one seems the most likely, with the non-cooled VRAM stack on the back of the card), and everything else that there might be - the card will not stop until something gives out with the unlocked, warranty voiding 1000W Kingpin VBIOS)

You mean that temp that i reported was too much ?
I'll inevitably kill my card with the 1000W on aircooled ?
From 500W XOC to 550W unlocked, i see not so much difference with temp?
I never see the card bottlenecking because of too hot with the VRAM on XOC.
 
If you confirm at 100% that i'll kill my card, i'm doing a rollback now hehe.
 
Best regards,

its not killing the card actually, but if in future something goes wrong and temp get really high the card won't shutdown till it start burning because it doesnt have any protetction on

Okay.
I'll be aware of the temperature.
It is sure that with that setup, the GPU will not like a burning at 100% during long minutes/hours.
 
Will buy a EK ASAP and waterblock it. Will be my first watercooling EK :)
I needed to watercool the CPU (5950X) anyway which has good potential at stable 5ghz :)

Just make sure EVERYTHING that currently has TIM/pads on it has EXTREMELY good contact - and the back side VRAM will likely need a fan blowing on the backplate if you don't have water running through it (I saw 1 waterblock with active cooling on the backplate). Hot spots where icx sensors aren't at will kill that component, and again - with a warranty void VBIOS on the card, you have a $1800+ door stop when EVGA laughs at you for trying to RMA with that on it. Normal VBIOSs have limits on temperature as well that a user cannot see (each VRAM chip for sure, perhaps other things) - that's why using that unlocked VBIOS as a daily driver is insane (unless you literally have money to burn).
 
As to any guarantee that your card will or won't fry itself - I guarantee nothing. You're the one playing with fire, so you do you.

Thanks you for all advise.
I just confirm (lol) you that the 1000W Bios in absolutly not designed for an aircooled card.
I was very attentive to the temperature and i cannot stay on it, doesn't want to break my card.
 
the XOC bios let me go to 500W with a 430-450W avg. It's cool OK.
I tired the XC3 bios, but i get lower score in anybenchmark than with XOC.
Is this normal ? It possible than for some ppl the XOC bios perform better than the XC3 on a FTW3 Ultra ?
 
The XC3 bios doesn't seems at all to consume the power it need. 8PIN1 ~ 90W 8PIN2 ~150W , PCI-E @ 66>70W.
Correct power is not 8PIN1 : 150W / 8PIN2 : 150W ?
 
Maybe you have idea why XC3 perform less than XOC on my FTW3 Ultra ?
I followed your tutorial and cleanning right trace with DDU. No curve modding....
 
Best regards,

I get the exact opposite as you have...
XOC used more power but my score was way lower than XC3...
XC3 power numbers are all *****d up but my scores are way higher...Im talking like going from 12.7k to 14k on Port Royale from XOC => XC3
 
Strange indeed...

god503
FYI FOR THE OPTIMUS HATERS....  I see ek just released its water block for  the 3090?  340 bucks!!!!   
 

exlink
Decided to try out a 3090 FTW3 to see how it compares to my 3090 XC3. Pretty sure the FTW3 I got is a complete dud. I max out the power limit on the XOC BIOS and the thing maxes out at 430W but it’s average power draw is 390W...

Not only that but my 3090 XC3 maintains better clock speeds at a much lower average power draw of 360W. I’m returning this FTW3. It spikes to 85W power draw from the PCI-E slot while the first two 8 pins max out at 120W and the third one at 100W. My XC3 never pulls more than 75W from the PCI-E slot and both 8-pins pull as much as 156W.

The 3090 FTW3 definitely has issues with power delivery and load balancing.

Clayman31

MastahFR
I was sick on waiting for a bios to correct the issue (a bios that will probably never exist) and because I am just going to get a Strix O24G that is on it's way, I will just sell the GPU.
I wish a good luck everyone for getting either an RMA or a bios to fix the issue.
 
For me after 4 months and no official message from evga since a long time is not a good sign.
That was my first EVGA card (always wanted one but each time the cards were out of stock on each generation) and will be most likely my last from EVGA.

How much faster is a strix? Isn't it 480w with worse warranty service?

eXp67

P4R4DiSi4C

eXp67

arestavo

eXp67

pellaaa

eXp67

arestavo
Press F to pay respects.

Please post back here when your system crisps, as others would benefit from hearing about it. (People aren't joking about the no protections - the VRMs, the VRAM (this one seems the most likely, with the non-cooled VRAM stack on the back of the card), and everything else that there might be - the card will not stop until something gives out with the unlocked, warranty voiding 1000W Kingpin VBIOS)

You mean that temp that i reported was too much ?
I'll inevitably kill my card with the 1000W on aircooled ?
From 500W XOC to 550W unlocked, i see not so much difference with temp?
I never see the card bottlenecking because of too hot with the VRAM on XOC.
 
If you confirm at 100% that i'll kill my card, i'm doing a rollback now hehe.
 
Best regards,

its not killing the card actually, but if in future something goes wrong and temp get really high the card won't shutdown till it start burning because it doesnt have any protetction on

Okay.
I'll be aware of the temperature.
It is sure that with that setup, the GPU will not like a burning at 100% during long minutes/hours.
 
Will buy a EK ASAP and waterblock it. Will be my first watercooling EK :)
I needed to watercool the CPU (5950X) anyway which has good potential at stable 5ghz :)

Just make sure EVERYTHING that currently has TIM/pads on it has EXTREMELY good contact - and the back side VRAM will likely need a fan blowing on the backplate if you don't have water running through it (I saw 1 waterblock with active cooling on the backplate). Hot spots where icx sensors aren't at will kill that component, and again - with a warranty void VBIOS on the card, you have a $1800+ door stop when EVGA laughs at you for trying to RMA with that on it. Normal VBIOSs have limits on temperature as well that a user cannot see (each VRAM chip for sure, perhaps other things) - that's why using that unlocked VBIOS as a daily driver is insane (unless you literally have money to burn).
 
As to any guarantee that your card will or won't fry itself - I guarantee nothing. You're the one playing with fire, so you do you.

Thanks you for all advise.
I just confirm (lol) you that the 1000W Bios in absolutly not designed for an aircooled card.
I was very attentive to the temperature and i cannot stay on it, doesn't want to break my card.
 
the XOC bios let me go to 500W with a 430-450W avg. It's cool OK.
I tired the XC3 bios, but i get lower score in anybenchmark than with XOC.
Is this normal ? It possible than for some ppl the XOC bios perform better than the XC3 on a FTW3 Ultra ?
 
The XC3 bios doesn't seems at all to consume the power it need. 8PIN1 ~ 90W 8PIN2 ~150W , PCI-E @ 66>70W.
Correct power is not 8PIN1 : 150W / 8PIN2 : 150W ?
 
Maybe you have idea why XC3 perform less than XOC on my FTW3 Ultra ?
I followed your tutorial and cleanning right trace with DDU. No curve modding....
 
Best regards,

I get the exact opposite as you have...
XOC used more power but my score was way lower than XC3...
XC3 power numbers are all *****d up but my scores are way higher...Im talking like going from 12.7k to 14k on Port Royale from XOC => XC3
 
Strange indeed...

Yep i absolutly don't understand and i'm not skilled so much with these 3090 like some of ppl here :)
All i can say, with XOC > PR @ 14K, XC3 > PR @ 13.2K ?
 
With XC3, the freq seems to got a little higher than XOC, but not stable at all.
Perfcap reason is most of the time in PWR mode with XOC/XC3.
With 1000W set at 550W, i had rarely PWR cap reason ... a little normal you'll say lol
 
I think really that if you have a watercooled card, you mod the RAM timing, not afraid about 90W PCI-E, the kingpin 1000W is the gral with a 600W target :)
 
Regards

H-Emmanuel

arestavo
Like I've written several time already:
 
3090 FTW3 cards that don't have the power limit issue very likely won't show any benefit at all on the XC3 VBIOS. Thanks for sharing your results that confirm my suspicions on that.

My FTW3 Ultra has a power limit issue around 460-470W depending on the app. The XC3 BIOS was completely useless, throttling at a reported 296W while the PCIe maxed out at 63W. It was a lot worse than the XOC 500W at 119%.
 
I can't believe how we're all getting different results. For this variance to occur, Evga must have been inconsistent in its sourcing of components during manufacturing. We haven't really been able to completely correlate this power problem to one lip style, serial # or place of manufacture. I'm picturing a bin filled with "interchangeable" components manufactured by different companies that are grabbed randomly by a tech on the assembly line lol.
 
In all seriousness, I think Evga shopped around for a given component(s), either for cost savings or due to sourcing problems without realizing they were introducing functional differences within the same product line, and maybe even within a given batch. I wouldn't be surprised if even they couldn't trace it back to a certain serial # or batch # and that would explain why they still haven't gotten back to us, they can't come up with a one-size-fits-all BIOS.