So I'm encountering several issues with PX1 and my new Kingpin. I duplicated multiple times.
In PX1 8000mhz memory is not close to stable at stock voltage, while I can run 8000mhz memory in AB all day long and is perfectly stable.
Also best efficiency with my example with stock voltage is 2190/8100.
For some reason I can not get anything beyond stock voltage stable. (2220 - 1.15v)I've tried with evbot and classy tool and no matter what, seems like it will not overvolt stable beyond stock voltage with either tool. Even with chilled air and keeping temps below 30*c.
I'm hoping I'm missing something.or there is a bug to be fixed. I'm also noticing some discrepancy between set voltage and reported on card. 1.093v set reports up to 1.15v on card.
Next step was to throw a universal on for better temps on chilled loop. But, if I cant get card to like overvolt, not sure if it's going to make a difference as I will only see some.minimal scaling on stock volts.

Yes the memory issue makes absolutely no sense,I agree. But having installed uninstalled px1 and ab half a dozen times and getting the exact same result....
Also this would be the first GPU I've owned that does not respond to additional voltage for increase in clock versus stock voltage, And that's a lot, looking at the 2 dozen hawks, lightning, matrix, and kpe on my shelf.
I'll keep playing with it report back.
Appreciate your response.

I wanted to clarify, because based on response, which i do appreciate, need to.
 
At stock voltage my card tops out about 2190-2205 core and 8000-8100 memory. Pretty good! Additional voltage on core, up to 1.15v, has absolutely no effect and cannot clock higher. My temperatures are very good and highest i've seen full load is 45*. I have been able to keep idle down to 15*c and load to 30*c with chilled air. Not being able to get any response from additional voltage with temps and clocks like that seems very odd to me, in my experience. If there was a temp issue or additional voltage unavailable it may seem reasonable.
I am going to go ahead and drop in my chilled loop and will feel better about making some other adjustments and bump voltage a bit more more and see what it will do. i cant do much else at this point. Hopefully sliders will make some difference and do remember previous KPE cards needed this to shine. Really dont want to put a pot on it until i can confirm some clock gains with increased voltage, otherwise it will feel like im benching a Txp and solely depending on cold scaling.

Did you try different LLC settings? Not sure what Stock level is, but using LLC 5 helped to stabilize higher voltage settings during runs. Even with a beefy VRM, these GPU pull enough current to drag down the load voltage. Stock at 1.093v forced thru AB curve (Ctrl+L to lock a curve point) it would sag to 1.029v during GT2 TimeSpy near the beginning when your turn right and look down the hallway. (Highest current draw during the bench). Using 1.15v NVVDD LLC 5 it would sag to 1.058-1.062v and allow for higher clocks. LLC 10/11 had more sag.

This is using AB 4.60 final. (Curve editor is where AB shines)

This is measured with a Fluke 187 at the probe-it points.

There's also the Beta 3.15 out. Try loading it up as there's updates thru the X1 software.

AHowes
Wow complicated card.. is there any guide on how or what all these other voltage options and software do?

https://xdevs.com/guide/2080ti_kpe

I'm "only"  hitting about 2145 at stock voltages so you seem to be doing better than most I've seen on here. I got past 2200 this morning by waiting until it was pretty cold (probably low 60s) and over volting to 1.22v and setting llc to 11 to get 2220 stable enough just to get halfway through superposition. 1.15v gets me to 2160 or 2175 but any higher and it requires signficantly higher voltage and great ambient temps. I live in South Carolina which is really humid and I left my window open to get temps down in the room. Not the greatest idea as when I was benching this morning I noticed a droplet of water on my monitors bezel (I assume due to the high humidity creating condensation) so I killed everything out of fear of killing something. It's kinda weird as the spot the water droplet was wasn't even warm so maybe something else happened there.
 
That being said, by the time I got it to a voltage that wouldn't instantly crash at 2220 the core temp was at low 40s but the closer it got to 50 the lower the clocks went. 

So a little update. I've been playing with LLC and getting same weird results. What i have found is during idle card will show set voltage in comp to llc level, however during load, voltage drops to 1.0v and stays their until idle between tests. Something is not right here. DMM reading and card reading is so different. Card polling will show 1.04v during load, but DMM shows consistent 1.0 v throughout test, regardless of llc setting. Feeling like something is off with voltage controller or firmware. OC and LN2 bios shows no difference. Its not a PL problem or a Temp problem. Unfortunately i can not post pics or video straight to forum.  Not going to breakdown card for chilled or LN2 at this point as it would be a waste and/or give reason for issues I'm encountering. Can someone post some realtime DMM info for comparison. This card should be a monster, but without real voltage getting to card.....
Tspy for comparison. Efficiency in AB is not a problem. Only running a  7820x @ 4.8ghz.
https://hwbot.org/submission/4139430_
 
 

I finally got it to show load correctly by DMM. Flipping the 25mv switch seems to be only way i am seeing "right" voltage via DMM. 
What info do you need Tin? Set voltage via classy tool was 1.093 up to 1.125v showing or lacking voltage on dmm during load benching Tspy. Ran multiple times, reboots, ect., ect.
 
https://youtu.be/ZNgOrNlmKGE - Here is video showing 1.07v load on led while dmm shows .9v during Tspy.

LLC should have no effect on report of loading on stock voltages without tweeks save clocks.. 1.043 - 1.07v is typical report via dmm on any 2080ti. Nothing wrong with dips switches or contacts.I've only tried +25mv for function quality and reset to off.
While I've tried functionality of classy tool, evbot, and now xoc bios, I've gone back to stock voltages and trying to work out why reporting by dmm is off.
I can say that xoc bios is showing correct or expected reporting via dmm. Led is still high with peaks of 1.12v though. But, change of LLC level is not shown in change log so....
I'll try rechecking oc and stock bios to see if this somehow carries over or what.
I guess my question is are the led and probe it taking voltage from same spot??

are the OLED and probe it taking voltage from same spot??

No, they show different voltages. ProbeIt shows GPU die voltage, almost exactly what chip see.
OLED (that is with GPU NVVDD sensor shown) display what power controller outputs. These voltages are not the same, as there is always some power loss between VRM and GPU due to PCB, components, etc. With different loading you will see difference in voltages as well. I wouldn't focus too much on trying to have all voltages be same, it's not really meaningful. What matters is the relative readings. So typical is choose one point as you reference, be it OLED NVVDD reading or your DMM reading, and use that as a baseline to tweak and play with different settings.
 
And yes, Loadline settings applied for any voltage, even default.

BIOS have no mechanism to affect loadline. But loadline is always in play. There should be no difference on any of the BIOSes, if all settings are same.
And first I don't even know what reading you referencing now, you going by precision x1 voltage and DMM reading from ProbeIt? 
Comparing card A to other card B directly is not going to work, because there is no "default" loadline setting, it's up to each design to take that choice (given the result is still meeting GPU specs). Also reading point is different. I don't think many 3rd party (if any) cards report you real voltage anyway. Reference cards and other typical cards (even FTW3) go by NVAPI voltage, which is just preprogrammed VID, and not the actual real measured value.

Are you thinking you are able to somehow finesse something in the classified tool, px1 or on a pcb switch that will get you higher clocks that 2200?

Judging by all the literature, 2200 is the golden zone on the AIO, 2300 maybe on hydro copper, and more on ln2. To me, at 2200mhz, at stock voltages, it sounds like you win the lotto. Perhaps you can get a few MHz more with some fine tweaking. But why do you want more voltage, if the clock won't see a benefit? That's like scratching a winning number off the lotto ticket. The whole beauty is that it's so efficient it doesn't need the voltage.

That being said, I think tin was saying that others cards don't provide accurate measurements in regards to either voltage or watts, because they aren't designed to do so accurately. In addition to that dissimilarity between kpe and the masses, is that you actuslly have the ability to set llc and voltages on a granular level to eek out those last MHz. Or go to town on ln2.

Just in case you weren't catching his drift. Sidenote, I didn't think matrix's were out yet?