Landauer's Principle is often presented as a fundamental limit of efficiency for classical computing. It states that in order to erase one bit of information, at least the following amount of heat has to be released:

$$ E = k_BT\ln 2 $$

It's very interesting as a bound of how efficient we can ever hope to get (similarly to other principles like Bremermann's Limit), but it is also given as a reason for developing reverersible logic gates, "reversible computers," etc, which is what my question is about: how good of an argument for reversible computing is it, really?

I'm talking here only about "reversible classical computing," i.e., excluding quantum computers which are of course also reversible and clearly useful if realized.

First, the reasoning for reversible computing as above strikes me as a bit unintuitive: by changing our, say, AND gate from one output to two outputs, we're probably adding more circuitry to whatever our one-output gate would be (and at least aren't removing any). Therefore, it seems to me that we haven't improved our situation at all: we still have our wires with electricity flowing through them and losing some energy in the process (and likely more due to the added complexity). Overall, it seems that without major updates to our tech (like superconductrs), we can't hope to improve our situation with things like reversible computing.

Second, my understanding of the principle is that it is more or less the formalization of the following intuitive argument: "according to known laws of physics no information is ever lost, therefore our classical, non-reversible computers need to release some of it as heat whenever 'erasing' data."

This argument suggests to me that in order for the limit of the principle to really become relevant, we'd have to have devices which don't bleed information to the outside world. However, this is the exact difficulty of building quantum computers! Therefore, it seems to me like that it's going to take forever (perhaps literally), before we even begin to approach this kind of limit.

Hence, do we really expect our classical computers made to "quantum standard" to ever be able to approach the efficiency of our classical computers? And if they do, are we then not likely to also have quantum computers as fast as the classical ones, rendering the latter obsolete?



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.