A little update on that, and not sure this helps much, but it looks like the KPE 1080 Ti comes with a bundled "EPOWER V.5" sort of power board. This helps understand what the hell "vdroop" in SW100 and SW101 are.
- Thread on EPOWER V
- PDF version of the guide above (or the other way around)
in particular, pages 24 and 25.
The differences are:
- SW3401 switch is like VOFS, but comes in increments of 25mV in KPE 1080Ti instead of 80mV from EPower V.
- there's no "force" switch pair, instead there's the memory's "Vdroop" switch (SW100), that's not present in EPower V
- the GPU "Vdroop" switch seems to be the same, so SW101 relates to DROOP
Basically, by what I could understand from the guides (my assumptions base also on reading the KPE 1080Ti uncorking guide
), the "Vdroop" switches are used to determine where power to the memory or GPU are sensed; switching only one dip kind of balances sensing between the power board and GPU (via a passthru 0-ohm resistor). This can be used to "compensate" loss of voltage in the circuit that negatively affects measurements from the whole power module.
The usefulness of this is pretty clear in the EPower board, especially when you may de-solder those zero-impedance resistors and pull a twisted pair of wires straight to the board to make the measure closer to the GPU (remote sensing). But I suppose KPE should already be wired to go that close, stock (without changes/soldering)? So I'm not sure this should be very helpful here.
I am trying the setting at 50% Vdroop for memory now (just one switch turned on), going a little more aggressive on the o/c, and see what happens. Once it lights up the 3909, I may try it at 100%. I really don't see much point in figuring out if KPE 1080Ti has equivalent RFP/RFN to solder a bridge ("remote sensing" from EPower guides), as the board is all bundled and its "on" position probably has good rails straight to the points of interest.
I am also not sure all this vdroop thing should be needed in "light overclock" that's not LN2 extremes...