This is correct, but the _amount_ of energy required to run to completion isn't correlated with the number of steps in the computation. Bennett shows that the computation runs more-or-less forward (using a very small push) until a few steps from the end, at which point you need to add a few extra kT's to get it to _stay_ at completion. At 06:41 PM 12/10/2013, Eugene Salamin wrote:

A reversible computer can run without an energy source, but which way does it run? You need some entropy increase to drive it in the desired direction.