arestavo

Kylearan

Someone said that increasing MSVDD and MVDDC voltages with the classified tool (I believe it was to 1.125v and 1.40v) helped raise the internal power limits that were triggering this throttling on Kingpin cards.

This won't help on your card as you don't have access to these voltages.  The only thing that will address this is an eVGA Bios update or a shunt mod.  Your card is drawing a lot from the PCIE slot, relative to the 8 pins, but it still isn't 100% clear if it's just the PCIE draw limiting you or if it's another power rail.  I know that the Founder's Edition will throttle at 79.9W PCIE Slot (you would only reach this if you shunt modded WITHOUT shunt modding the PCIE slot), even though the vbios dump says the PCIE limit is 86.2W...

The "good" thing is that your TDP and TDP Normalized are very close to each other at least, so the rails "think" they are balanced.

Theoretically, you can flash the Kingpin VBIOS onto the 3090 FTW3 and have access to those voltages via the classified tool. Both cards allegedly use the same analog voltage controller. Edit: I tried, and it indeed does unlock those voltages on a FTW3 card. However, it didn't make a difference to the power limit (380 to 400W in Valley with perfcap power, and furmark maxed at 450W and most of the time was at 430 (Kingpin 520W VBIOS)). 1.12 nvvdd, 1.4 fbvdd, & 1.2 msvdd and was able to change both load lines to level 1 and disable both ocp. The NV and MS frequencies were stuck at 400 Hz.

cerealkeller
Has anyone been able to update Precision X using the XC3 bios without being asked to update firmware? I think I’m on 1.1.1 still
Regarding clocks, I’ve been running +70/+900 on the XC3 BIOS with whatever voltage it gives me since the curve doesn’t work and I don’t have the voltage turned up. I think it’s 2040 after thermal throttling 2070 without, I think, it may actually be 2100, but it never actually runs there. Maybe once I get my water block. Soon I hope.
It’s been very stable though, for Red Dead 2 and Cyberpunk 2077, what I’ve been playing lately.

Step 1: Install the latest PX1, but don't open it yet. Step 2: Navigate to the PX1 installation folder and delete everything in the Firmware folder. Step 3: Profit. Found out that this doesn't work for PX 1.1.4. Maybe it would work on an older version, but I just use AB instead.

My card would also drop to 2040MHz or so on air because it would get up to 78C while playing CP2077. The Hybrid kit keeps it to mostly 62C, with a max of 64C now, and the clocks stay around 2100 with 2130 being the max with my gaming OC.

How do you validate the voltage settings you set actually applied?
 

drunknfoo
How do you validate the voltage settings you set actually applied?
 

The tool kept the settings after I closed and opened it, except for the Hz ratings which were set to 400 instead of max. Don't think there's any other way to check it.

Well I think I have evidence beyond a reasonable doubt that the primary problem causing the Power Limit issues for the 3090 FTW3 Ultra/Hybrid cards on the BIOS or board settings indeed has to do with the vRAM.
 
I set to stock, run Port Royal, tickles power limit.  I begin revalidating my vRAM OC settings by setting my core to 2040MHz @ 981mV (which results in NO power limit readouts during Port Royal), start with +500 vRAM, no power limit.  +750 tickles power limit... +1000-+1250 solid power limit (at 450W, mind you, not "500W").  Score doesn't go up after +1100, I back off to +1000 for a nice even number this time.
I let the core back to "stock" so it's running the stupid voltage with horribad clocks, still +1000 vRAM, solid power limit through the test.  I think about it, and figure why not see if I can break my record...
I set to +210/+1000 - solid power limit, the core only actually hits 2100+ for a few seconds through the entire test.  Only hit 14.6K, then it hits me, every time I see Power Limit on GPUz, the Core Clock is going down to keep the card "under" the limit, which is why I'm not seeing 2100+ throughout the test, I think to myself, there's 24GBs of vRAM on this thing, maybe try underclocking the vRAM & see what happens...
Sure as the sky is blue.  At -250 vRAM, +210 Core, in Port Royal, Power Limit is still dinged in GPUz, BUT... the Core Clock is hitting 2070-2085-2100-2115 the whole test.
 
So something to do specifically with the voltage or power draw or a combination of the two, of the GDDR6X, is what's actually causing the issue on these cards of not being able to hit clocks or the actual 500W power limit, as others have suggested in this thread, from the looks of it.
 
I'm genuinely curious now if I set those clocks in Time Spy if I'll actually see 500W of draw instead of hitting a hard wall at 480-493W.  Going to test that next.
 
This gives me the sense that this is part of the problem Kingpin owners are running in to as well, since even on those, you can only adjust the voltage on the GDDR6X so much... Guess I'll find out when I get mine.
This really is a bummer because this card really seems like it would be a great overclocker if I wasn't being totally blocked from doing so by having to choose between core clock & vRAM clock since ones goes down if the other one goes up regardless of what I set.
 
The real question is, is this even something they can fix with BIOS updates or is this a hardware-side issue?  
 
Is anyone else experiencing that specifically, where you set a GPU core clock & it doesn't actually hit & stay at it, if you're hitting power limit? 

Sure as the sky is blue folks:
 

 
466W draw, only tickling power limit under PerfCap occasionally, GPU core clock was more consistent, and also worthy of note, look at the individual power draws, PCIe slot is only at 78W, and the 8-pins are at 129-129-134, so much more even than I have seen before trying this.

arestavo

drunknfoo
How do you validate the voltage settings you set actually applied?
 

The tool kept the settings after I closed and opened it, except for the Hz ratings which were set to 400 instead of max. Don't think there's any other way to check it.

ya, it's not applied.... i remember maxing out values when attempting to play with it after realizing they're not actually applied...
your statement made it sound like it was 100% functioning... so i had to ask how you validated...

drunknfoo

arestavo

drunknfoo
How do you validate the voltage settings you set actually applied?

The tool kept the settings after I closed and opened it, except for the Hz ratings which were set to 400 instead of max. Don't think there's any other way to check it.

ya, it's not applied.... i remember maxing out values when attempting to play with it after realizing they're not actually applied...
your statement made it sound like it was 100% functioning... so i had to ask how you validated...

How did you validate that the settings weren't applied?

Because the card didn't explode when you literally maxed out all voltages?

arestavo

drunknfoo

arestavo

drunknfoo
How do you validate the voltage settings you set actually applied?

The tool kept the settings after I closed and opened it, except for the Hz ratings which were set to 400 instead of max. Don't think there's any other way to check it.

ya, it's not applied.... i remember maxing out values when attempting to play with it after realizing they're not actually applied...
your statement made it sound like it was 100% functioning... so i had to ask how you validated...

How did you validate that the settings weren't applied?

Because the card didn't explode when you literally maxed out all voltages?

pretty much lol
 
set my clock and mem to what i know is borderline stable, playing around with the sliders while running benchmarks, didn't cause any instability

Just ran Port Royal with only the -250MHz vRAM applied, nothing else, and now I'm seeing 465W power draw during the test (used to max out at 450-457W at the very most), and my core is clocking higher at stock... 
 

Dabadger84
Just ran Port Royal with only the -250MHz vRAM applied, nothing else, and now I'm seeing 465W power draw during the test (used to max out at 450-457W at the very most), and my core is clocking higher at stock... 

Someone in the kingpin power draw problem thread mentioned that there are three power limits that the FTW3/kingpin cards can hit very early at low overall power draw (390W in my case) - PCIE slot, VRAM, and one other that I forget. They also mentioned that the power limit for the VRAM is set lower on these cards than other manufacturers set, even the FE cards.


Don't ask me exactly how a 2 power pin XC3 VBIOS corrects that issue, but it does.

arestavo

Dabadger84
Just ran Port Royal with only the -250MHz vRAM applied, nothing else, and now I'm seeing 465W power draw during the test (used to max out at 450-457W at the very most), and my core is clocking higher at stock... 

Someone in the kingpin power draw problem thread mentioned that there are three power limits that the FTW3/kingpin cards can hit very early at low overall power draw (390W in my case) - PCIE slot, VRAM, and one other that I forget. They also mentioned that the power limit for the VRAM is set lower on these cards than other manufacturers set, even the FE cards.


Don't ask me exactly how a 2 power pin XC3 VBIOS corrects that issue, but it does.

I really want to give it a try but I'm genuinely worried about bricking the card...
 
If I were to brick a BIOS, I can just boot up with the BIOS switch in the other position, switch it back to the broken one & reflash, correct?