I suspect that we can derive the greatest amount fo computational
power per mass by converting as much mass as we can to neutonium.
By the tine we know how to do that, we'll also know how to build
nurtonium nanotech. The reason that neutronium computers are faster
is that they are a lot smaller, so the internal signals don't
have to travel as far. The speed-of light delay has always been
an important constraint on supercomcuter speeds, and is now
becoming important on workstations. Other things being equal(!)
a neutonium computer should be thousands to millions of times
faster than a computer of the same mass made of normal atoms.
Of course, we still need to use some of the mass for power,
and we'll want to minimize the wastage so as to maximize the
amount of neutronium we'll have available, but these are
trivial engineering details for our sucessor civilization
to worry about.