Another "me to" post. 
Power 1: 117
Power2: 130
Power 3: 34
GPU Load 100%. 
Power Consumption: 87%
PerfCap Reason Pwr
 
I am confused by the support response from earlier.  Are they saying this is not an issue?

MDG73

moadib2k
Another "me to" post. 
Power 1: 117
Power2: 130
Power 3: 34
GPU Load 100%. 
Power Consumption: 87%
PerfCap Reason Pwr
 
I am confused by the support response from earlier.  Are they saying this is not an issue?

That is what support basically told me with my 3070 FTW3. Which I almost do not believe or which is kooky considering I got my daughter a Asus Strix 3060Ti and has proper balancing between the two 8-pin power plugs. I don't understand why both wouldn't be used equally if other board manufacturers can do it, why can't EVGA?

If I am characterizing this correctly, my GPU is at 100% and so should be drawing full power.  Also curious by GPUz would say there is a performance cap.  I am a bit new to this so it could also be my understanding.
 

I guess I'll live with my 430w power draw and 100w plug 3 limit :) It works pretty well in games anyway. Maybe EVGA will find a way to fix it in a bios update or something one day.
 
 

ClowReed

bmwm3smg
Any ideas on what is going on with the 8-Pin #3 Power draw on my new 3090 FTW3? These were the max numbers on my first Port Royal run on my card. Default settings (no OC), running the OC Bios (same results with the Normal Bios). Power Supply is a 1300 G2. Please let me know if I missed any pertinent info that you guys need.

Thanks!
 
Lester

Man, sorry for crashing in... Maybe this is old news already, but after seeing your post I made some tests here with my 3090... In Furmark the 3rd 8 pin cable stayed at around 70W and total board power around 320... Cables 1 and 2 was getting 120W give or take. Ok... I installed OCCT and used their stability test for vga... well... now things got a lot more interesting... Total card power raised to spikes of 450W (I'm using the stock "OC Bios" switch) and all three 8 pin power were using full 150W. So... Maybe it's by design... The third one is the last one requested, just in case of extreme power usage.


Just my 2 cents... Sorry if this was already mentioned. I hope it helps.

Just tested your theory and I still get a discrepancy of 30 to 40w difference even in OCCT with Perfcap of PWR. Power draw on pin #1 12.0v on pin #2 11.9v. wonder why the dip on pin 2?
Damn, I was hoping that was the case for me.

Hot take: If the power balancing works properly, it should be able to balance throughout the entire range of loads and applications.  I realize that it doesn't really matter until you reach the max on any individual input, but it is silly if it can't power balance at lower loads either. Observing that it doesn't power balance properly in Furmark but can power balance properly in OCCT is kind of silly.  This is just another indicator of power balancing which isn't effective.  Like I say, I realize that it doesn't really matter until you reach a max where a power limit is being imposed, but, you see this is an indication of an implementation limitation.

3090 KPE here, could this be the reason that I sometimes see power limit when the card is drawing only ~400W?
 

I don't think they'll rma it for that. Yours doesn't look that bad. Has anybody tried with different PSU's and see if there's a difference?

This issue exists on the kpe 3090 as well - however we've managed to overcome the limit by increasing the msvdd with classified tool.  This increases the internal clocks of the cache (as far as I know the higher voltage it runs increases the frequency with an internal v/f curve) which is finally able to feed the rendering pipeline with less latency per clock.  With a low msvdd - the card even at 2200mhz will not post scores consistent with other brands of card, but increasing msvdd by a small amount (have to play with to find the right number for certain external gpu clocks) - results in substantial performance increases - and also solves the power balancing issue.  Pre-msvdd tweaking - the card will run to the limit at 430w.  After msvdd - 540w and massive increase in performance.   I suspect that msvdd is power throttling the card far sooner than the global power limit would suggest.

Another issue on the non kingpin bios is the SRC Power limit linked to the 8 pin rail.
While all of the 8 pins have a "global" power limit, which is only one value which is related to TDP (8 pin + PCIE Slot power limit= TDP),
Each 8 pin has a "SRC" rail linked to it, which is usually 150W default, 175W max, and this limits the max 8 pin power draw (the 8 pin power limit doesn't, the SRC power limit for that rail does).
So there are 3 SRC's for three 8 pin cards.

Appreciate all the great info here guys (the ones that actually provided some info vs the "kids n trolls"!
My 3080 FTW3 Ultra exact same thing with gimped power from PCI-E #3 plug.
Power is about 1/2 from Plug #3 vs #1 & #2.
 
With XOC 450w bios only can reaxh 395w!
 
btw psu Corsair AX1600i.

add me to the list now of uneven power draw. I can't even play games now. I load COD and from the FPS capped selection screen the power draw goes stupid and my entire computer crashes. Now to RMA and remove my block and setup **** all over agian. FUN TIMES.

Intoxicus
 
Are you using 3 separate power cables or are you using any of those splitters?

I am curious about this as well.  I've always heard that three separate power cables should be used and a splitter should never be used for these cards.  And also have a power supply with enough connectors on it so that three separate cables could be connected without the need for a splitter.

So has anyone tried playing with the classified tool on the ftw3 3080 and 3090's too see what happens? (or is it locked to Kingpin cards only? Has anyone been able to get around that? I guess maybe on a 3090 you could flash the bios but not a 3080?)

Clearly this is just another "issue" in a string of problems with these cards that just seem to be cost cut hard, for whats ment to be one of their top tier cards, they're not much better than a bottom end tier brand.
As Buildzoid points out, it seems to just be a reference card overspec'ed in the wrong areas and woefully underspec'ed in others (like a good cooler, great power delivery but gimped by an cheaper caps and an analog voltage controller).

Ughh, i know in gaming it probably only nets very little "gain" but it still miff's me that the card was sold with the PWR limit of 420w and mine, thats watercooled (after umming and ahhh'ing over the whole Bricking fiasco when you remove the cooler/fit a water block), running nice and cool and ready to overclock can't even hit that with the 450w bios (EXCEPT when it put under OCCT, where it'll draw 100w on that final plug #3 and hit 420w, in games you bet that im hitting the same values, sitting at 380w and gimped.

Then you just know theres others who have no intention to OC or do anything other than plug it in and play (which is also perfectly fine) who probably have cards with no power issue)

I'm very near just throwing caution to the wind and running the hacky bios to get full power (i mean you can always keep the OC bios from the switch on the eVGA bios and switch back if it fails....), hell if there was a decent hardware modification that'd fix this problem (such as maybe all these cards have a bad batch shunt resistors or something stupid that causes the voltage reg to read/report incorrectly?) that were removable (maybe hotglue shunt mod like Frame Chasers likes to do), i'd even consider that option

Anyway, guess i'll put my name down as another Gimped 3080 FTW3 (Ultra.....) that probably won't get "fixed", just swept under the rug (i mean Jacob said stay tuned for news in a couple weeks on this issue....that was in February no?)....i wouldn't have been so worried about my card when waterblocking it if i had have known about this before....probably would have risked an EK block over a Bykski too, if it died its a decent reason to RMA it.....

Only saving grace is i got this card as proper Australia MSRP ($1250 delivered) and not the $1750ish they're asking now (or higher for other brands), i guess now i know why it was so cheap?!?!?

Prior anyone starts claiming it is faulty product keep in mind cart itself was designed as 350W.
Now we are getting it unlocked to 450W normal and 500W OC. There was never any statement that those cards will reach 500W.
Mine usually ranges from 485W to 510W during heavy bench on OC but I usually run it on default on 450W vBIOS on standard 420W target.
 
 

Nothing has been said by EVGA yet about the power balancing.  Terrible so many have this same issue.