It is true that reversible computers can solve the issue of necessarily losing heat due to entropy decreases, but current computers are still far too inefficient for entropy loss to be a concern.
I am not so sure, all of the heat being dissipated by your typical CMOS transistor is due to the transistor state going through a linear region as it switches from fully saturated to fully depleted or fully depleted to fully saturated.
As I see it, if you're transistors did not dissipate that energy as heat, and instead 'stashed' it so that it could be re-used. Then you're 4 Ghz CPU would no longer need a heat sink.
That’s the point. The entropy cost thermodynamically inherent to doing a calculation is infintesimal compared to the energy cost of current transistors.
From Wikipedia:
> At 20 °C (room temperature, or 293.15 K), the Landauer limit represents an energy of approximately 0.0172 eV, or 2.75 zJ. Theoretically, room‑temperature computer memory operating at the Landauer limit could be changed at a rate of one billion bits per second with energy being converted to heat in the memory media at the rate of only 2.85 trillionths of a watt (that is, at a rate of only 2.85 pJ/s). Modern computers use millions of times as much energy.
My knowledge of quantum general relativity isn’t good enough to even know for sure that the search for answer to that question is exactly why Hawking et al have been concerned with the black hole information paradox.
As I see it, if you're transistors did not dissipate that energy as heat, and instead 'stashed' it so that it could be re-used. Then you're 4 Ghz CPU would no longer need a heat sink.