Everything that I know about physics (conservation of energy & mass etc) is telling me that no power is needed to do maths... so I don't think there is.
Problem is "doing maths" isn't really a physical concept - the key is how you do the maths. However the implication there is that to do maths you need to do work to get the result, and doing work certainly does require power in some form or another (or at least a transfer of energy.)
100% in terms of efficiency is theoretically possible using classical mechanics laws, but practically it'll never happen, or ever get close (and it's a long time since I've looked at it but I'm pretty sure effects such as quantum tunelling will prevent 100% efficiency even in theory on the level processors work at.)
Instead if you're looking at the next huge leap in processor design, check out quantum processors. In the next few decades or so I reckon it'll go mainstream and give us an order of magnitude better processing capability than we have at present (it's actually a lot closer than people realise.) Thing is unlike computers at present you'll never really understand it unless you've got a pretty good grasp at vector maths well beyond 3 dimensions and a solid grasp of quantum theory. The potential however, is huge!