I've been using a Bitcoin miner as a heater in the winter. I figure if I'm going to use resistive heat, I might as well generated some possibly-valuable wooden nickels as a side effect of the conversion from electrical to heat energy.

But I'm curious: does all the energy get turned into heat during computation? In practice, I understand that not all the energy gets truly applied to the computation in the first place, due to power supply inefficiencies and fans and if there are any spinning platters. And also, a little bit of energy leaves the system through LED indicators, EMI radiation, and along network cables, etc.

But does any of the energy actually get "consumed" in the computation itself, turned into information or something? From a thermodynamic perspective, is there any difference between a chip that draws 5W while quietly calculating digits of π inside, versus a length of nichrome wire drawing 5W? Do you get exactly 5W of heat out in both cases?

There's a related question at Is it necessary to consume energy to perform computation?, but I'm sort of asking the opposite: when energy is consumed performing (purely) computation, does every last bit of it end up as heat?

  • $\begingroup$ I mean it doesn't have to; but in practice it pretty much does. You can design a computer where the readout of digits of pi is based on marbles either being at the top of a big Babbage engine or at the bottom, and then some of the energy is stored in the gravitational potential energy of the marbles. Similarly I'm sure some trace energy goes into aligning spins in hard disk drives; it's probably a higher-energy state for the disk to be full of random data than for it to be full of 0s. But it is very small and the key question is "do I have to do this," not "does it happen ever in practice?" $\endgroup$ – CR Drost Dec 21 '16 at 22:47

The answer to this question is a clear "yes", all of the power going into a computing device ultimately turns into heat*. There is no energy equivalent of information. Put another way, energetically speaking information is completely worthless.

*Well, unless you do something fancy such as running a thermodynamic cycle fed by the heat emitted from a processor, say, which could then convert a part of that heat into mechanical work or other forms of energy.

  • $\begingroup$ Actually there is energy in information i.e. entropy. Check the Landauer's principle. $\endgroup$ – user110971 Dec 21 '16 at 23:59
  • $\begingroup$ I know Landauer's principle. Entropy is not the same as energy. $\endgroup$ – Pirx Dec 22 '16 at 0:49
  • $\begingroup$ "Energetically speaking, information is completely worthless." — thanks! You nailed the heart of my question; I do wish I could accept both of these replies as "the answer" since I appreciate them both. $\endgroup$ – natevw Sep 27 '17 at 21:34
  • $\begingroup$ Actually I'm realizing this directly answers my question, while the other answer first focuses on what I said was NOT my question :-) Combining the two, perhaps we can say that the only energy (temporarily) "consumed" by the computation would have been that which was stored to represent information. $\endgroup$ – natevw Sep 27 '17 at 21:53

The short answer is yes, computation does create heat by necessity. That is because the value in a register is often erased and replaced by some new value, and the loss of information (the loss of the old value that was stored) is intrinsically a contribution to entropy. This applies to any irreversible process, like storing a computed number, but NOT to reversible processes, like forming the exclusive-or of a number and a registered value. The exclusive-or of the same number a second time, with the register, will reverse the operation (yield the same result as before the first exclusive-or, i.e. that operation was not irreversible).

The more subtle inquiry, 'does every last bit of it end up as heat', is probably deserving of a 'no' answer. That is because the machine state (the ones and zeros held in memory) includes stored energy (charge variations on internal capacitances), so that with all common electronic technologies, there is nonthermal energy in the computation, produced as well as consumed. The program state of flash chips, for instance, is a kind of charged battery with many years required before that energy finally dissipates and turns into heat.

  • $\begingroup$ Thanks, yes my main question was "does every last bit of it end up as heat?" but "does computation create heat by necessity?" is also an interesting angle. So I'm accepting this as the answer for being a bit more thorough, although I think @Pirx directly answered just my core question. $\endgroup$ – natevw Sep 27 '17 at 21:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.