DashwoodFoxe
So I should just return my FTW3 Ultra? Kinda makes my ridiculous water cooling setup a moot point if the GPU won't be able to utilize the enhanced cooling it has to offer.

Chamidorix
No we are talking about the PCI-E slot power (regarding the 10amp fuse). The point is that EVGA is artificially gating power draw in a hardware provable way. It is not causing the issues in this thread. The separate issue of voltage controller firmware locking 8 pin power draw at a stupid low ratio is the only cause of preventing us from drawing 500W.  But the PCI-E slot fuse being so low rated vs other vendors is hard evidence EVGA is artificially segmenting. It shows EVGA is actively trying to lock the card down so people don't just use it instead of the Kingpin, even in the hardmodding XOC crowd. It shows they are not acting in good faith. You cannot prove that EVGA will not at some point adjust the voltage controller firmware ratio at some point to be more favorable to the consumer, but you can prove they have limited draw via fuses vs other vendors and infer they are not optimizing for the consumer. 
 
The explain the voltage controller firmware enforced ratio, it is simply the 8 pins are not allowed to pull more than 1.66*3 of what the PCI-E pulls. Other vendors are 2*3 of what the PCI-E pulls, for the full 150W per 8 pin at 75W on PCI-E. EVGA, on the other hand, caps the 8 pin total draw at 375 W when the PCI-E is measured at 75W. This can be spread over all 3 8 pins, and prefers to load the first 2 8pins at 150W and the 3rd at whatever is left. The complicated thing is the fact that the VC doesn't simply read the PCI-E shunts for draws, it also compares them vs the gpu core 1.1V power rail shunt to determine what the actual draw is. This is why in applications that pull heavier on cache or memory you can exceed 450W, as it spikes up the power consumption without increasing the PCI-E draw in the VC. 
 
With 10 amp fuse, you cannot draw more than 120W (input is 12V) without the fuses blowing. This means with shunting your theoretical max power draw on a FTW3 is (with the enforced voltage regulator ratio of 1:1.66*3) 120W PCI + 3 x 200W = 720W. While an MSI Trio with a 20 amp fuse with the non gimped ratio of 1:2:2:2 could simply pull the max amp allowed by fuse, for 240+240+240+240 = 960W. The Asus Stix actually doesn't even have fuses after looking more closely over the PCB, so you can literally draw as much power as you want before the wires melt (VRMs can theoretically handle thousands of watts if kept cool, they will just be extremely inefficient at higher wattages and dissipate gargantuan amounts of heat). The kingpin will similarly not feature fuses so that it can pull as many watts as it would like, when XOCers have upped to voltage to 1.5+V and cool with LN. But the FTW3 is fused at such a low 10 amperage and has such a poor pci-e:8pin ratio that it will be heavily restricted when shunting + water-cooling for 24/7 daily usage, vs other cards. 
 
The end point is not that the FTW3 is a garbage card, it is that it is objectively worse then the Asus Strix at the same price and worse even then the MSI Trio at a lower price. For some the marginal performance differences for daily air cooled usage are vastly overshadowed by EVGA's good customer support rep, and so the FTW3 remains the best option. But for those seeking to get the best performance and flash unlimited VBioses, shunt, watercool etc, for the first time the top end non-ln2 card is actually gated to the point that performance suffers vs competitors even when not volt modded or ln2 cooled. 
 
I'm guessing less than 10% will actually take the team to read and really understand this, but now at least the full picture and explanation should be visible to fellow enthusiasts.

joshjd93
This thread needs to be locked. So much meaningless speculation. If we really couldn't draw the power at a hardware level, why was the FTW3 card used to break records on LN2 and draw 800w+ in those benches. Can we wait and see what EVGA figures out? 50w under load extra has already been proven to not show a huge gain. Maybe 1-2%. I keep checking this thread to see if EVGA updates anything but instead see the same echo chamber comments that are just speculation and nothing more.

xrb936

joshjd93
This thread needs to be locked. So much meaningless speculation. If we really couldn't draw the power at a hardware level, why was the FTW3 card used to break records on LN2 and draw 800w+ in those benches. Can we wait and see what EVGA figures out? 50w under load extra has already been proven to not show a huge gain. Maybe 1-2%. I keep checking this thread to see if EVGA updates anything but instead see the same echo chamber comments that are just speculation and nothing more.

SO WHAT? The only thing we can do before EVGA makes any comments is speculating. That's not out fault. I deserve to know what my card is DOA.

DarthYodi

joshjd93
This thread needs to be locked. So much meaningless speculation. If we really couldn't draw the power at a hardware level, why was the FTW3 card used to break records on LN2 and draw 800w+ in those benches. Can we wait and see what EVGA figures out? 50w under load extra has already been proven to not show a huge gain. Maybe 1-2%. I keep checking this thread to see if EVGA updates anything but instead see the same echo chamber comments that are just speculation and nothing more.

All of what he said is provable, how are you saying this is meaningless speculation? I guess you are part of the percentage he said would not read through his post. Also link to your source on the power draw above 800W please?
 
In addition, interestingly GN in this video: https://youtu.be/S1EiRjZH8Xk?t=185 at the time stamp specifically mentions an issue involving fuses on the EVGA card. Is this the same issue we are speaking about here or unrelated to this thread's commentary?

Chamidorix
No we are talking about the PCI-E slot power (regarding the 10amp fuse). The point is that EVGA is artificially gating power draw in a hardware provable way. It is not causing the issues in this thread. The separate issue of voltage controller firmware locking 8 pin power draw at a stupid low ratio is the only cause of preventing us from drawing 500W.  But the PCI-E slot fuse being so low rated vs other vendors is hard evidence EVGA is artificially segmenting. It shows EVGA is actively trying to lock the card down so people don't just use it instead of the Kingpin, even in the hardmodding XOC crowd. It shows they are not acting in good faith. You cannot prove that EVGA will not at some point adjust the voltage controller firmware ratio at some point to be more favorable to the consumer, but you can prove they have limited draw via fuses vs other vendors and infer they are not optimizing for the consumer. 
 
The explain the voltage controller firmware enforced ratio, it is simply the 8 pins are not allowed to pull more than 1.66*3 of what the PCI-E pulls. Other vendors are 2*3 of what the PCI-E pulls, for the full 150W per 8 pin at 75W on PCI-E. EVGA, on the other hand, caps the 8 pin total draw at 375 W when the PCI-E is measured at 75W. This can be spread over all 3 8 pins, and prefers to load the first 2 8pins at 150W and the 3rd at whatever is left. The complicated thing is the fact that the VC doesn't simply read the PCI-E shunts for draws, it also compares them vs the gpu core 1.1V power rail shunt to determine what the actual draw is. This is why in applications that pull heavier on cache or memory you can exceed 450W, as it spikes up the power consumption without increasing the PCI-E draw in the VC. 
 
With 10 amp fuse, you cannot draw more than 120W (input is 12V) without the fuses blowing. This means with shunting your theoretical max power draw on a FTW3 is (with the enforced voltage regulator ratio of 1:1.66*3) 120W PCI + 3 x 200W = 720W. While an MSI Trio with a 20 amp fuse with the non gimped ratio of 1:2:2:2 could simply pull the max amp allowed by fuse, for 240+240+240+240 = 960W. The Asus Stix actually doesn't even have fuses after looking more closely over the PCB, so you can literally draw as much power as you want before the wires melt (VRMs can theoretically handle thousands of watts if kept cool, they will just be extremely inefficient at higher wattages and dissipate gargantuan amounts of heat). The kingpin will similarly not feature fuses so that it can pull as many watts as it would like, when XOCers have upped to voltage to 1.5+V and cool with LN. But the FTW3 is fused at such a low 10 amperage and has such a poor pci-e:8pin ratio that it will be heavily restricted when shunting + water-cooling for 24/7 daily usage, vs other cards. 
 
The end point is not that the FTW3 is a garbage card, it is that it is objectively worse then the Asus Strix at the same price and worse even then the MSI Trio at a lower price. For some the marginal performance differences for daily air cooled usage are vastly overshadowed by EVGA's good customer support rep, and so the FTW3 remains the best option. But for those seeking to get the best performance and flash unlimited VBioses, shunt, watercool etc, for the first time the top end non-ln2 card is actually gated to the point that performance suffers vs competitors even when not volt modded or ln2 cooled. 
 
I'm guessing less than 10% will actually take the team to read and really understand this, but now at least the full picture and explanation should be visible to fellow enthusiasts.

zogge
I think it is related to silicon lottery honestly.

Either way, I picked up a 3090 Gaming trio x today and returned my EVGA card and canceled my optimus block order. Good luck guys and I hope thinks will work out for you eventually.

Over and out.

sirien
Is there no response from Jacob yet? Why is the evga card drawing 80w from the pcie slot while the strix is drawing 50

jdmwrxpower

Bilc0

DashwoodFoxe
So I should just return my FTW3 Ultra? Kinda makes my ridiculous water cooling setup a moot point if the GPU won't be able to utilize the enhanced cooling it has to offer.

I would do it and get a strix imo.... I got lucky(or unlucky) and got one off amazon 2 hours after my FTW3 shipped - but I hadn't known it at the time.  I was staring at the delivery date for about a week while following this thread hoping eVGA would say something... anything.  People have mentioned a 'we're looking into it,' but I haven't seen that comment related to the current issues with the bios.  As far as I am concered silence is deafening here.   
 
Well I am glad I decided to break the seal and bench the Strix.... it beats the pants off the FTW3 I had in terms of clocks and at a lower power limit while being at HIGHER temps... so when this baby gets under water it should really sing.
 
This company is great at supporting you when the hardware fails, but it seems when it comes to actually admitting fault they are silent for whatever reason - protecting percieved reputation or protecting current sales who knows.  All I know is I was drawn to this for the higher bios PL but they failed to deliver in a 3 week period, so I'm out.

That's just it. 95% of us are going to block our cards and if it isn't worth putting the ftw3 under water with no gains than what's the point of the extra premium cost? My ultra is coming in Tuesday next week as it shipped from evga yesterday and to be honest I think I am just gonna sell it at whatever the going rate put a few hundred in my pocket and shoot for a strix. My asus tuf 3090 went to my brother yesterday and I had zero issues with the card. On air it was 70c 350-360 watts. The fact EVGA is silent on the whole topic is a bit concerning.