changboy
For those who think there card are bad and dosen't perform well i have this to tell you :
 
If you score 200 even 500 lower on port royal benchmark you can barely notice this in game at 4k, its translate maybe between 1 to 3 fps maxxxx.

Kylearan
The only common thing I could find is that Memory (FBVDD) power is not exceeding 98 watts and the lowest I saw with the power limit flag is 93 watts.
I'm 100% sure it's either memory power or PCIE slot power that is limiting you.
 
It can't be anything else.

willm1224

MNFirstBlood
Maybe this is not the right place or the right time but I have been messing around with the beta oc bios and the original default oc bios. I am getting more power with the beta oc bios (at default clocks) but the card is still downclocking due to power limit as soon as I hit 450 or higher. If this has already been discussed I apologize. I didn't read all 69 pages of this thread. Would that explain why the performance is actually worse because the bios even though it's rated for 500w now, it still is adjusting clocks and vrel at 450w. Like maybe it down clocks further the higher past the PL?  I have gotten 490w in timespy extreme no overclock just max power slider but as soon as I hit 450 and beyond it starts downclocking and GPU-z also saying power limit has been reached. I am probably way wrong but it seems odd it would be reporting power limit when I have never reached 500w. Another thing I have found is that if I apply any overclock or voltage adjustment in AB it never goes beyond 450w. Only on default settings with just the power target set to 119% does it go above 450w.

yea like 75% or more of people on here can't get 500 w w/XOC bios and are having similar issues. I just sent my card in for RMA they will have a sample, they know this XOC problem is with my card, EVGA has been lurking and monitoring this thread, hopefully when they get my card tomorrow they can look at it more closely. I sort of feel like they don't have a sample of a card that doesn't work with XOC (well they can't just open all their factory sealed merch can they)

Kylearan
So it looks like the 8 pins should be all drawing 140 watts before relegating to the PCIE slot.
But for some reason, the third 8 pin isn't pulling its weight in power.  140 * 3 + 75 is exactly 500 watts, which is probably how this bios is supposed to have its power draw partitioned.
But if the stock bios doesn't draw from the three 8 pins in an even way, then you're going to have problems.
 
The big question you guys need to figure out is this:

Is the 140 watt 8 pin limit signaling a power limit rail?
Or is the 75W PCIE slot limit signaling a power limit rail?
You guys need to realize there are multiple rails that determine power limits. 
The only one the TDP handles is Total Board Power, which is equal to all 8 pins + Slot power.
 
I know on the 3090 FE card, the 8 pin power caps out at either 170W or 174W, but I do not know if this is affected by TDP--it probably is *NOT*.  PCIE Slot can go up to 85 (Unconfirmed).
GPU Chip power caps out at 300W.
 
There's also individual 8 pin power rails, Slot Power rail, GPU Chip Power rail, MVDDC Power rail, PWR_SRC Power rail, and possibly one other.
 
The one thing I haven't seen *ANYONE* here post:

What is the difference between your TDP% and your TDP Normalized %?  GPU-Z does not show this.  HWinfo64 DOES show this.
 
If TDP Normalized is appearing much higher than your TDP% and your TDP% is appearing much lower than the "cap" you set , when you hit the power limit, it means an internal rail is throttling you.

Kylearan
>stuff<

Kylearan
There are ways to test this.
 
1) Underclock the memory as far as reasonable.  Try -500 mhz on the memory.
Then increase the power limit as high as possible and see if you can exceed 450W now.
if FBVDD actually responds to memory, it should be possible to reach 470-480W this way.
 
2) check FBVDD at lower power limits and lower memory settings then increase the power limit and check where FVBDD is when you start getting hard throttled.

Kylearan

MNFirstBlood

Kylearan
There are ways to test this.
 
1) Underclock the memory as far as reasonable.  Try -500 mhz on the memory.
Then increase the power limit as high as possible and see if you can exceed 450W now.
if FBVDD actually responds to memory, it should be possible to reach 470-480W this way.
 
2) check FBVDD at lower power limits and lower memory settings then increase the power limit and check where FVBDD is when you start getting hard throttled.

Okay I will try this. That would be called just FB in afterburner right?

No.  GPU-Z calls it MVVDD.   HWinfo64 calls it FBVDD.   MSI Afterburner doesn't call it anything.
Just use the memory clock slider in MSI AB or precisionX.

Attached Image(s)

Kylearan

MNFirstBlood

Kylearan

MNFirstBlood

Kylearan
There are ways to test this.
 
1) Underclock the memory as far as reasonable.  Try -500 mhz on the memory.
Then increase the power limit as high as possible and see if you can exceed 450W now.
if FBVDD actually responds to memory, it should be possible to reach 470-480W this way.
 
2) check FBVDD at lower power limits and lower memory settings then increase the power limit and check where FVBDD is when you start getting hard throttled.

Okay I will try this. That would be called just FB in afterburner right?

No.  GPU-Z calls it MVVDD.   HWinfo64 calls it FBVDD.   MSI Afterburner doesn't call it anything.
Just use the memory clock slider in MSI AB or precisionX.

 
 Does this tell you anything? What is the MVVDD suppose to be at?
 

You have no attached image.
All I can see is that it seems to throttle at 93-98W.  So it's either that or PCI Express slot W causing the throttle.
So if I'm correct, forcing MVDCC to be lowered should increase the wattage headroom of everything else.

Attached Image(s)

arestavo
RMA finally approved for the 78 to 82W average PCIE slot power draw (with over 86W spikes) on my Taiwan manufactured 3090 FTW3 Ultra (no overclock, stock (shipped with) OC VBIOS, S/N starts with 20141, advanced RMA started on Saturday). Had to call EVGA twice - first time today it was they've got to get management to approve any EARs (6 minutes of hold time), and the second took just a couple minutes on the phone after a 30 minute call back timer close to the end their regular office hours (which is when, I assume, management is available). I'm guessing that the gentleman that I spoke with, James, is a manager? He didn't say, but I did mention needing approval from a manager for the advanced RMA in the call back message. He was courteous even as he mentioned that PCIE power draw should be fine up to 85W, and I didn't receive push back after mentioning that 75W is the spec, and that I wasn't overclocking at all with those readings (taken from the latest GPU-Z).
 
For those that didn't see my previous posts being concerned with this: Another member who had similar PCIE power slot draw from his 2080 Ti (I assume overclocked, while mine is stock for those readings) had his motherboard die on him when his GPU kicked the bucket after ~6 months of use - something that scares the hell out of me because my other add-in cards are expensive, and the CPU is hard to find at retail prices (even though both motherboard and GPU are covered via EVGA warranty, the RAID card, dual 10GbE card, and Optane PCIE drive probably wouldn't be - not to mention data/time lost!). It's something that I'd rather not have to ever deal with with a top-tier, and pricey, GPU - especially since overclocking (which I do!) tends to increase power draw (averages and spikes).
 
Now the wait to see if the replacement card is out of spec out of the box, is a Taiwan or Chinese card, and (per the thread topic!) can actually leverage the 500W VBIOS. The latter of which is likely not happening with cards like mine that draw up to 82W average (depending on the game/benchmark/resolution) without any overclocks from the PCIE slot and likely holding the card back from drawing more power, e.g. that 500W average that some people can get on their cards - however, that is just a theory on my part and I could be completely wrong. I'm not an GPU engineer, afterall - just an enthusiast.
 
Oh, and thankfully I chose EVGA - otherwise I might have been sent a free Mooncake, or would be waiting a couple weeks to hear back.
 

bavor

Gonzhilla
I'm just very disappointed that I can't even sustain 60fps in Assassin's Creed Valhalla in 4k when it's advertised as an 8k card. Like why did I spend 2000 dollars on this

 
That's an ubisoft game with poor programming and a lack of optimization.  Something Ubisoft is known for.
 
 
 
It looks like the 3090 does great at 4K gaming!  Most of those games are on "High" "Ultra" or Epic" graphics settings.

 

MaxTheOwl
Hello,
 
I've been following this thread since it started and decided to do some testing. Mine is a 2014XXX card. Ordered on launch day.
All the test are with 100% at the fans and STOCK BIOS.
I like to here what you are thinking about the results.
Thank you.
 
Here are my results. Everything you need to know is in the image discription.
 

 
Edit: My default power draw seems to be too low, right?