Yes, there are some who report that they could not get anymore than 50Mhz extra boost clock on their Classified, leaving them asking why they paid extra money when even a reference PCB could accomplish the same results on air.
The extra power phases only come into play on subzero cooling.
The "stock boost" is 1860mhz for a 1080 Classified. I can get it completely stable at 2152mhz and semi-stable (enough to get a 3dmark validation) at 2202mhz with a low ambient temperature (8c).
Considering these GPU's were only designed to run at a boost of 1733mhz that is a MASSIVE overclock (+467mhz overclock).
An updated 1080 Classified voltage controller came out last week too. Which supports voltage increases for Core, Memory & PCI-E.
Here's a graph with some memory benching I done which compares results between memory at stock 1.37v (stock) and 1.5v: (60 benches total to create this graph)
1080 MEMORY BENCHMARKING
EDIT:
*Don't pay too much attention to 1.5v at 675 (I had to do +675 the following day as I missed them) so "different day" factors probably changed results).
Whats interesting though; is before I performed these benches (and put it into graphical form) I assumed my max stable memory O/C was around 500 - 650. Yet this graph still shows a very small, but steady increase all the way up to +925. Artifacts/Crashing begun at +950.
Another interesting factor is performance seems higher every 50mhz, so:
725,
drop,
775,
drop,
825,
drop,
875, drop,
925.
Think it must be caused by the timing "synchronises" with the clock which causes this. Having it "synchronise" better by choosing a number with a better ratio also seems to give stability improvements.
Also not had a chance to test a more "medium" memory voltage boost yet (1.46v for example) but that's my project for this weekend.
Also: I ran 2 benches for each +offset, or 3 if I got an unexpected result.
Its a pitty there isn't software that could automiatically run benches and change the MSI-AB settings as it goes. Imagine waking up in the morning to all your numbers
Anyway more importantly:
Kingpin says voltage doesn't do much past 1.2v on Pascal.
But the nvidia limit is 1.093v. So That's still just over 100mv headroom.
I'm waiting for my waterblock to arrive (this weekend) before I begin playing with voltage on the core.