PadinnPlays
Yeah, so following that I'd say around 440 is where my average is.

I'd say that's the average for most of us posting here to be honest with you... Those "spikes" are few and far between. I've had those up to 490+ but then I run two instances of GPU-Z and when observing the current power draw during the same session, I can easily guesstimate my average isn't more than 440-450 when running Time Spy Extreme or any other benchmark I've tried really. Circling back to my earlier comments here, when I had my ZOTAC 3090, it was nearly consistently closer to 500w. Wish I had taken screen shot but I didn't as I was more interested in an EVGA. That said, my scores were not as high as what I'm getting with my EVGA. But then again, I didn't mess around with overclocking that card nearly as much as I've done with my EVGA.

Lastly, I see some people are actually successfully running the OC scanner in PX1 or AB. Yet, some (including yours truly) are getting the error. Someone commented that it's the EVGA 3090 cards but obviously that's not the case if it's working for some... Mine starts out seemingly OK but then it jumps to 254% (nothing much changes according to GPU-Z though) and then it errors out and gives the fail return in PX1 (or error H10 in Afterburner). Any ideas why this is?

I refreshed both bios and retested the OC bios using fills posted by Jacob.

Yeah, so I played around a bit with afterburner while running timespy extreme test. I can clearly see power draw changing when I change powerlimit %. The initial tests are with core voltage % at 0 and no OC.
Here are the averages at various power percentages in afterburner, one loop of the one minute timespy extreme stress test. Temp limit is set to 91c. Ran with max fans.

I cant find where in AB it lets you choose power limit vs temperature priority.

75% = 305w

100% = 398w. Call it 400.

110% = 420w

119% = 426w

119% w/ 100% voltage slider = 431w

119% w/ 100% w +100 core offset = 423w
+25 offset = 435w (this seemed odd)
+50 offset = 425w
+150 offset = test failed

It is entirely possible timespy is a poor choice to test this with. I'll see if I can find something more demanding that is easily repeatable.

PadinnPlays
Yeah, so I played around a bit with afterburner while running timespy extreme test. I can clearly see power draw changing when I change powerlimit %. The initial tests are with core voltage % at 0 and no OC.
Here are the averages at various power percentages in afterburner, one loop of the one minute timespy extreme stress test. Temp limit is set to 91c. Ran with max fans.

I cant find where in AB it lets you choose power limit vs temperature priority.

75% = 305w

100% = 398w. Call it 400.

110% = 420w

119% = 426w

119% w/ 100% voltage slider = 431w

119% w/ 100% w +100 core offset = 423w
+25 offset = 435w (this seemed odd)
+50 offset = 425w
+150 offset = test failed

It is entirely possible timespy is a poor choice to test this with. I'll see if I can find something more demanding that is easily repeatable.

Nice.  I'd go for Furmark 2560x1920 0x MSAA (AA off), that should put you solidly against any power limit. 
 
It's important to stay on perfcap pwr uninterrupted while taking an average power reading.

bavor
I've done experimenting with the beat OC bios and the BIOS that came installed on my FTW3 Ultra.  Sometimes the beta OC BIOS doesn't draw as much power as the stock OC BIOS.  Yet all my highest benchmark results have been achieved with the beta OC BIOS.
 
I have one of the highest AMD CPU results for unigine superposition benchmarks for both 8K optimized and 4K Optimized.  It looks like I have the highest air cooled GPU with AMD CPU results for the 4K optimized superposition benchmark.  My best stock OC BIOS results were all lower.  So obviously the OC BIOS seems to be working.  However GPU-Z isn't reporting a power draw anywhere near 500 watts.  Most of the time its in the 400-420 watts range, even with the beta OC BIOS.  I checked and the version number of the beta OC BIOS is correct on my card after the BIOS flash.  The highest wattage I've seen reported in GPU-Z was a brief jump to 450.1 watts total board power with the beta OC BIOS.
 
I have a 1300 watt 80+ Gold PSU and the 3090 is currently the only GPU in my system.
 
GPU-Z BIOS info for my card after the flash.

you don't have the highest air-cooled gpu/amd cpu 4k superposition score.

kevinc313

PadinnPlays
Yeah, so I played around a bit with afterburner while running timespy extreme test. I can clearly see power draw changing when I change powerlimit %. The initial tests are with core voltage % at 0 and no OC.
Here are the averages at various power percentages in afterburner, one loop of the one minute timespy extreme stress test. Temp limit is set to 91c. Ran with max fans.

I cant find where in AB it lets you choose power limit vs temperature priority.

75% = 305w

100% = 398w. Call it 400.

110% = 420w

119% = 426w

119% w/ 100% voltage slider = 431w

119% w/ 100% w +100 core offset = 423w
+25 offset = 435w (this seemed odd)
+50 offset = 425w
+150 offset = test failed

It is entirely possible timespy is a poor choice to test this with. I'll see if I can find something more demanding that is easily repeatable.

Nice.  I'd go for Furmark 2560x1920 0x MSAA (AA off), that should put you solidly against any power limit. 
 
It's important to stay on perfcap pwr uninterrupted while taking an average power reading.

Using these settings, I hit right around 440w. GPU load is sitting at around 85%. This is with no OC and +100% on voltage slider.

PadinnPlays

kevinc313

PadinnPlays
Yeah, so I played around a bit with afterburner while running timespy extreme test. I can clearly see power draw changing when I change powerlimit %. The initial tests are with core voltage % at 0 and no OC.
Here are the averages at various power percentages in afterburner, one loop of the one minute timespy extreme stress test. Temp limit is set to 91c. Ran with max fans.

I cant find where in AB it lets you choose power limit vs temperature priority.

75% = 305w

100% = 398w. Call it 400.

110% = 420w

119% = 426w

119% w/ 100% voltage slider = 431w

119% w/ 100% w +100 core offset = 423w
+25 offset = 435w (this seemed odd)
+50 offset = 425w
+150 offset = test failed

It is entirely possible timespy is a poor choice to test this with. I'll see if I can find something more demanding that is easily repeatable.

Nice.  I'd go for Furmark 2560x1920 0x MSAA (AA off), that should put you solidly against any power limit. 
 
It's important to stay on perfcap pwr uninterrupted while taking an average power reading.

Using these settings, I hit right around 440w. GPU load is sitting at around 85%. This is with no OC and +100% on voltage slider.

But is your perfcap staying on pwr the whole time?

evgatorator

PadinnPlays

kevinc313

PadinnPlays
Yeah, so I played around a bit with afterburner while running timespy extreme test. I can clearly see power draw changing when I change powerlimit %. The initial tests are with core voltage % at 0 and no OC.
Here are the averages at various power percentages in afterburner, one loop of the one minute timespy extreme stress test. Temp limit is set to 91c. Ran with max fans.

I cant find where in AB it lets you choose power limit vs temperature priority.

75% = 305w

100% = 398w. Call it 400.

110% = 420w

119% = 426w

119% w/ 100% voltage slider = 431w

119% w/ 100% w +100 core offset = 423w
+25 offset = 435w (this seemed odd)
+50 offset = 425w
+150 offset = test failed

It is entirely possible timespy is a poor choice to test this with. I'll see if I can find something more demanding that is easily repeatable.

Nice.  I'd go for Furmark 2560x1920 0x MSAA (AA off), that should put you solidly against any power limit. 
 
It's important to stay on perfcap pwr uninterrupted while taking an average power reading.

Using these settings, I hit right around 440w. GPU load is sitting at around 85%. This is with no OC and +100% on voltage slider.

But is your perfcap staying on pwr the whole time?

Yes. I'm making sure to let it go, not running it for an hour, but its stable for the couple minutes I let it go.

PadinnPlays

evgatorator

PadinnPlays

kevinc313

PadinnPlays
Yeah, so I played around a bit with afterburner while running timespy extreme test. I can clearly see power draw changing when I change powerlimit %. The initial tests are with core voltage % at 0 and no OC.
Here are the averages at various power percentages in afterburner, one loop of the one minute timespy extreme stress test. Temp limit is set to 91c. Ran with max fans.

I cant find where in AB it lets you choose power limit vs temperature priority.

75% = 305w

100% = 398w. Call it 400.

110% = 420w

119% = 426w

119% w/ 100% voltage slider = 431w

119% w/ 100% w +100 core offset = 423w
+25 offset = 435w (this seemed odd)
+50 offset = 425w
+150 offset = test failed

It is entirely possible timespy is a poor choice to test this with. I'll see if I can find something more demanding that is easily repeatable.

Nice.  I'd go for Furmark 2560x1920 0x MSAA (AA off), that should put you solidly against any power limit. 

It's important to stay on perfcap pwr uninterrupted while taking an average power reading.

Using these settings, I hit right around 440w. GPU load is sitting at around 85%. This is with no OC and +100% on voltage slider.

But is your perfcap staying on pwr the whole time?

Yes. I'm making sure to let it go, not running it for an hour, but its stable for the couple minutes I let it go.

Very interesting.  Kind of surprising it's not at near 100% load.  Could be nvidia driver settings or something else capping frames, or maybe it's just the power limit having a dramatic effect.  But being on percap pwr is what's important to measure power limit.
 
Are you averaging like 437w, or exactly 440w, or a bit more?

PadinnPlays
I should have taken a screen shot, I think it was 437.5.

Superposition 8k was locked at 420.

OK. You are definitely running against a 440w power limit.
 
Not sure about the superposition power draw and what it means.  You were also on perfcap pwr the whole time? Could also be on Vrel.
 
You know, SP 8K uses more Vram than any other bench I can think of.  Hmm.

kring
Port Royal will NOT max your power... end of story.  You have to use a real game i.e. Doom Eternal to test your card/bios power draw limits.
 
Please stop using a benchmark that is totally incapable of drawing power as a test of “maximum power draw“ and then complaining about low power draw ?!?!? 

Yeah I'm going to fire up Doom Eternal in a bit. Was hoping it had a repeatable benchmark but it doesn't. Got a recommendation for a specific level to play?
 
 

SladeX
Picked up a FTW3 Ultra today and did some tests. Out of the box, stock bios, I hit a peak 475W on the normal BIOS profile with fans maxed. I was dumb to update both normal and oc bios to the files in this thread, now the highest peak I can get is 461W. What is even more confusing is that when in OC bios switch mode, I have even less wiggle room for oc settings.
 
Reverting back to the "stock" files in this thread doesn't get me back any oc room.
 
I also cannot exceed 103% power limit, I can set 119%, but both px1 and gpu-z say otherwise. GPU-Z reports I am hitting pwr limit when trying to push the card.
 

Same here, without flashing the oc bios, i was getting much better results, idk why but this is really not  cool :(

Have some of these just in case the bios is not fixed buy the time you put a water block on and soldering skills , Yikes !  
ERJ-M1WSF8M0U

zogge it was sustained for 10 seconds at a time but it kept shooting over it so the card would throttle down.  Interesting thing to note out of no where my pc started acting funny it wouldnt show the splash screen to allow me to get into bios before start and if i tried to restart into ufei it would freeze.  A post said change the gpu see if it fixes it and it did.  So i googled that.  Apparently the card could sense a voltage issue when going to bios so it was freezing up to stop damage.  Reason why...it became partially unseated from the pcie slot.  I have my mobo chip set heat sinks that came on the mother board right behind my pcie slots like most mobos.  I realized comparing it with a zotac 3080 and a couple other card I have that on the ftw3 ultra 3090 the pcie connector is a few milimeters shorter.  Hell one of the slots I cant get it to click in locked.  What the heck did EVGA do because I compared it to a 2080 super from evga and the pcie connector is normal length on it. anyone else notice this?

I have hit near 495 watts when playing Shadow of the Tomb Raider and even Horizon Zero Dawn also yielded near the 500 watt mark.

TheAngryIrishman
He said he called EVGA and they said there was a bug in the the new precision x 1 they told him to enter his numbers. Apply and save and completely close precision and then benchmark.  That fixed it.  I tried it and I got higher scores.  I am still getting higher scores than that tho with after burner I actually hit the 119%  threshold and hit 501 watts several times.  Clocks were running at 2241.  So their oc software is messed up it appears. You have to get the new beta from guru3ds website tho as they incorporated ampre support in it if you don't have it then it wont work.
 

 
I tried that and when I closed PX1 my settings weren't saving.  PX1 is buggy.
 
The latest MSI afterburner won't allow me to adjust the voltage on the 3090 FTW3 Ultra.