2008-07-12

Reversible computing

I'm trying to understand reversible computing. My friend Bill made the comment that it may well be of interest to academics, but it has no appreciable effect in the real world. So here's a rough calculation for the power an irreversible computer needs that's operating right at the Landauer limit. For each bit the energy used per switch is T * k * ln 2. Assume a 2 GHz processor, 5 GB of RAM, and room temperature:

power = 293 * 1.38 * 10^-23 * ln 2 * 2 * 10^9 * 5 * 10^9

~ 0.01 W

we can times this by 100 because k*ln 2 is just the theoretical minimum that can never be reached.

~ 1 W

So with Moore's law, this limit will become increasingly significant.

I've no idea if this calculation is correct, I just made it up to prove my own point! Can someone correct me please?

No comments:

Post a Comment