d88-cwe@nada.kth.se (07/03/90)
Hello! I would like some input on some thoughts of mine; Could one say that to gain information in computers, you would have to lose time. This is some kind of parallell to the second thermodynamic law. The reasoning: Since, for a given kind, brand etc of computer, a certain energy- consumption per timeunit is consumed, energyresources is limited. (I disregard secondary storagemedia etc, I only mean memory and cpu in this context) Then, to gain information, or lower entropy in other words, on would have to convert a certain amount of ordered energy to a less ordered form, eg heat. Thus, for a given brand there should exist a theoretical limit on entropy-gain per time unit. My assumptions: I know that the controversy about if shannon-entropy and thermo-entrpy is exchangeable. I assume, obvoiusly, that they are. Would the reasoning hold even if it wasn't exchangeable? (Does it hold in the first place?) Sorry, the first sentence in the last section should be; I know that the controversy about if shannon-entropy and thermodynamic entropy is exchangeable is far from settled. (I am working directly, through telnet against the nntp-daemon. Our system is only a week old, so they have not installed the newssystem on our local machine. telnet have no editing-facilities. Please don't give up already. ;-)) I also see that the ratio is far from the true maximum in nature. Just look at animals. Also, most of the energy is wasted in just empty cycles, and thus is not used efficiently. If this reasoning is faulty, please tell me where. Could it be modified somehow, so it applies under other assumptions? My intuition tells me that energy isn't a problem in computers, but time is. If you want to do a big calculation, you dont get exhausted because of energyproblems in computers, but of the huge amount of time. Is that intuition wrong? Of course you can improve our current records of the entropy/time-ratio, and we are far from the absolute maximum. To avoid those problems, I only take within a given computer of a specific brand etc. Yes, I under- stand that a CM is faster than an IBM PC! I don't wan't this to be a brand-performance competition. Thank you of any feedback. Hope you have the patience to read this catastrophy, as I said, I cannot edit anything else than the current row. (I have to tell them to install the news-system right away) Christian Wettergren, d88-cwe@nada.kth.se PS. Use the above address, if somehow the header would screw up...
sestoft@rimfaxe.diku.dk (Peter Sestoft) (07/04/90)
d88-cwe@nada.kth.se writes: >I would like some input on some thoughts of mine; >Could one say that to gain information in computers, you would >have to lose time. This is some kind of parallell to the second >thermodynamic law. >The reasoning: >Since, for a given kind, brand etc of computer, a certain energy- >consumption per timeunit is consumed, energyresources is limited. >(I disregard secondary storagemedia etc, I only mean memory and >cpu in this context) >Then, to gain information, or lower entropy in other words, on would >have to convert a certain amount of ordered energy to a less ordered >form, eg heat. >Thus, for a given brand there should exist a theoretical limit on >entropy-gain per time unit. Do you know the following reference: Rolf Landauer: Dissipation and noise immunity in computation and communication. Nature vol. 335 (27 October 1988) 779-784. Excerpt from the abstract: "Reversible computers which carry out each step without discarding information can, in principle, dissipate arbitrarily small amounts of energy per step if the computation is carried out sufficiently slowly. ..." This to me sounds quite relevant to your suggestion. Not being a physicist, I am not in a position to say whether the arguments put forward in the article make sense. (But the referees of Nature must think so). Peter Sestoft * sestoft@diku.dk * DIKU, Department of Computer Science University of Copenhagen, Universitetsparken 1, DK-2100 Copenhagen O, Denmark Tel: +45 31 39 64 66 * Direct: +45 31 39 33 11/406 * Fax: +45 31 39 02 21