8
$\begingroup$

Consider a computer CPU consuming electrical energy to perform calculations and consequently emitting heat.

Assumption: That a CPU consuming x Watts of power, emits the same amount of heat as an incandescent heating element which consumes the same amount of power.

The rest of the question assumes that this is true: the energy has to go somewhere, and I can't see that the outcome of the calculations counts as "somewhere".

I've long repeated the claim that "wasteful" appliances such as incandescent lightbulbs, over-full kettles, computers left running, etc. are not wasting power if we make the assumptions:

  • the appliance is in a room which we aim to maintain at above a certain temperature, warmer than the external temperature (cf a home in winter)
  • the room leaks heat to the outside faster than the "wasteful" appliances emit heat.
  • the room contains an electric heater
  • the electric heater is regulated by a thermostat

In these circumstances, I think (?), if a light bulb produces excess heat, its only effect is to cause the thermostat to switch off sooner, so what energy you've "wasted" in the light bulb, is "saved" at the heater.

I don't think that's controversial.

But my question is:

  • If a CPU does useful work, and emits all the input energy as heat as a side effect
  • But a heating element emits the same heat, without doing useful work
  • Is it therefore wasteful to use heating elements? Is it more efficient to (for example) run folding@home on a CPU, to get "free" computation as a side-effect of heating?

One reason to heat a room with a heating element, rather than with CPUs, is that the heating element is much cheaper to produce. But ignoring the initial cost of the hardware, would a "heater" made out of 20 CPUs, each running at full capacity and consuming 50W each, output less thermal heat than a 1KW element that does nothing else useful?

To be useful, of course, the CPUs would need to output the result of their computations, which would account for a tiny amount of energy. Let's assume that's negligible.

I'm sure there's a flaw in there somewhere; probably that the assumption above is false. What have I missed?

$\endgroup$
  • $\begingroup$ My office building lost its heat a few days ago for a couple days; the 4 CPUs in my office could not heat the room but the portable heater we brought in the next day kept it nice and warm. Thus, I don't think your argument about heating your house with a computer is feasible. $\endgroup$ – Kyle Kanos Jan 10 '14 at 16:11
  • 1
    $\begingroup$ @KyleKanos 4 CPUs mostly idling is less than one 1KW heater, yes. 16 dual core Pentiums at 65W each, given a task that uses 100% CPU, would exceed a 1KW heater. $\endgroup$ – slim Jan 10 '14 at 16:19
  • 2
    $\begingroup$ @KyleKanos I can remove the reference to folding@home if necessary. It's there as a well known example of something that maxes out the CPU "indefinitely". You seem to be missing the point here. The question is theoretical. Would it make sense to manufacture "heaters" which do computational work as a side effect? $\endgroup$ – slim Jan 10 '14 at 16:23
  • 1
    $\begingroup$ Have replaced "should we" with "is it more efficient to". $\endgroup$ – slim Jan 10 '14 at 16:38
  • 2
    $\begingroup$ @KyleKanos have replaced "cost the same as" with "use the same amount of power as". In CS we use "cost" to mean that; thought it was the same in physics. $\endgroup$ – slim Jan 10 '14 at 17:09
9
$\begingroup$

There's no flaw in your argument. A computer heats the room just as effectively as an electric heater of the same power and you could use the computer to do something useful (Bitcoin mining?) while it's heating your room.

There are some practical considerations, though I think these have been sufficiently discussed in the comments. Computers would make for noisy and bulky heaters, though anyone who has had to refrigerate a server room will tell you they can be very effective heaters.

If you want to be really, really, pedantic the computer may not be quite as effective as a heater because an infinitesimally small amount of it's power may go into the data on its disk drive. See Is a hard drive heavier when it is full? for a discussion of this, but note that I mention this for its amusement value rather than because I seriously believe it's an issue.

$\endgroup$
2
$\begingroup$

In the physical sense, the computation a CPU does is not useful work. It doesn't produce any force and on average the electrons are raised/lowered evenly in potential. Just like a heater, it is $100\%$ efficient at converting electrical energy to heat. In fact your 20 CPUs will generate more than 1kW because of losses in the power supply, but not too much more. Your computer board will be an expensive way to generate heat compared to a space heater (I think it unfair to ignore that) and will not allow you to direct the heat where you want it, but in terms of heating the whole room it will be as good.

$\endgroup$
2
$\begingroup$

The energy produced by computers in datacenters is often used for heating purposes. Google with "energy from datacenter used for heating". So, on a larger scale, using energy generated by computers for heating purposes does work.

$\endgroup$
  • $\begingroup$ That's pertinent, but the opposite way around. "We're computing so we might as well use the heat" is a common thing (and people do it all the time unintentionally). "We need to produce heat, so let's do some computing" is what I'm asking about. $\endgroup$ – slim Jan 11 '14 at 13:41
  • $\begingroup$ @slim Well, in that case, unless the computing was useful, you'd never pay for the wear and tear on the computing equipment. Also, consider if the alternative was a more efficient form of heat than electric heating (such as gas heating or a heat pump). $\endgroup$ – David Schwartz Mar 12 '17 at 5:49
  • $\begingroup$ I think in 2014 it was already true, but in 2018 with cloud computing being completely mainstream -- companies say "We need to locate a load of computers; that cold part of the world needs to heat a building. Bingo." $\endgroup$ – slim Jan 16 '18 at 16:25

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.