Cooling server farms and suchlike things.

Latest Articles

Cooling server farms and suchlike things.

Downsizer Forum Index -> Energy Efficiency and Construction/Major Projects

Author

Message

Hairyloon

Joined: 20 Nov 2008
Posts: 15441
Location: Today I are mostly being in Yorkshire.

Posted: Mon Feb 24, 25 4:33 pm Post subject: Cooling server farms and suchlike things.

I was recently moved to think about the cooling systems for server farms and other high intensity IT applications and wondered if anyone anywhere has tried recovering some of the energy through heat engines to make electricity.

The instinctive response is that it goes against the laws of thermodynamics, but I think that it does not because of the amount of energy being pumped into the system.

As a thought experiment, if you connect a heat engine to a dynamo and drive a heat engine with an electric fire, then will the dynamo produce electricity?
If the engine, dynamo and heat source are suitably matched, then of course it will. It would be horrendously inefficient, but that's not the point because instead of the electric fire, we have a bank of processors doing useful work.

As is, they do that work and generate heat which is typically thrown away, although in some cases it is used for heating...

Which brings me to another thought: where people are using electricity for heating, why not put that electricity through a bank of processors first and do some calculations with it?

Leaving aside the question of what calculations we might do, as far as I can work out, we would get the same amount of heat either way, so the only reason to not do it is the cost of the hardware, and how many obsolete graphics cards are there sat tucked away in drawers?

Slim

Joined: 05 Mar 2006
Posts: 6634
Location: New England (In the US of A)

Posted: Mon Feb 24, 25 7:21 pm Post subject:
Pairing processing with heat generation has certainly been done with crypto miners https://time.com/7017395/bitcoin-data-center-heat-bathhouses/

Mistress Rose

Joined: 21 Jul 2011
Posts: 16105

Posted: Tue Feb 25, 25 8:23 am Post subject:

It isn't against thermodynamics because the electricity being used produces heat but has done some 'work'. There will be less energy produced as heat than is put in in electrical power.

Of course it is far better to power these places with power from renewable resources.

In the early days of electronics cooling was a major problem and I don't know how valve operations were cooled, but by the time of transistors it was quite possible to build a radio in a cigarette box that would spontaneously ignite. Even early ICs had enormous (relative to the size) cooling fins on them. Modern chips use far less power, and thus heat, but are a lot better than they were.

Hairyloon

Joined: 20 Nov 2008
Posts: 15441
Location: Today I are mostly being in Yorkshire.

Posted: Tue Feb 25, 25 10:53 pm Post subject:

Slim wrote:

Pairing processing with heat generation has certainly been done with crypto miners https://time.com/7017395/bitcoin-data-center-heat-bathhouses/

I'd heard about that, but my thought there was on the domestic scale: instead of heating the room with a little fan heater, heat the room with a little fan heater that's doing loads of calculations.

Slim

Joined: 05 Mar 2006
Posts: 6634
Location: New England (In the US of A)

Posted: Wed Feb 26, 25 11:19 am Post subject:
Yes, those also exist. This is the one that popped up first when I googled, but I remember seeing a different one previously https://heatbit.com/

Hairyloon

Joined: 20 Nov 2008
Posts: 15441
Location: Today I are mostly being in Yorkshire.

Posted: Wed Feb 26, 25 12:34 pm Post subject:

Mistress Rose wrote:

It isn't against thermodynamics because the electricity being used produces heat but has done some 'work'. There will be less energy produced as heat than is put in in electrical power.

It's not clear which point you're addressing here, but I think that the work done by a processor to do calculations is not work in the thermodynamic sense, so it should all come out as heat.

dpack

Joined: 02 Jul 2005
Posts: 46369
Location: yes

Posted: Wed Feb 26, 25 1:12 pm Post subject:

iirc there is a pilot scheme of server cooling to domestic energy harvest by popping the boxes in many homes in a suburb and using the cooled heat in that home, might be dutch?

even if it worked it is only appropriate for a new build project mated to a server array

iirc encryption avoided the data loss potential of having a distributed bit crunching set up and the homes had a lot of their domestic energy needs from heat that needed to be removed from the pooter bits

as a very relevant aside, one of my home built pooters(old kit able to render DV or run a 32 track dig studio rig for live and pootered on the triggers) needed about 1.5 Kw removed from the box using several fans
i did not need room heating when it was running on full

ps i cooked the clocked chip in seconds when a fan failed

lots of low temp heat to be had from pooter kit, that might change, but if the investment could be returned in say less than a decade it should be a goer

Mistress Rose

Joined: 21 Jul 2011
Posts: 16105

Posted: Thu Feb 27, 25 8:30 am Post subject:

Hairyloon, the electricity is being used, therefore it is a form of work, whether it is to run a processor or turned into light in a light bulb. To my mind, which is technical rather than purely scientific, that is still work, so thermodynamics apply. I tend to look at the practical side rather than the theory. In theory a bumble bee shouldn't be able to fly, but it does, therefore the theory must have a flaw somewhere; same applies here.

Hairyloon

Joined: 20 Nov 2008
Posts: 15441
Location: Today I are mostly being in Yorkshire.

Posted: Thu Feb 27, 25 6:21 pm Post subject:

Mistress Rose wrote:

Hairyloon, the electricity is being used, therefore it is a form of work...

We need to be specific about what we mean by "Work".
In the thermodynamic context, work means no more or less than a transfer of energy.

Quote:

whether it is to run a processor or turned into light in a light bulb.

With a light bulb, it is easy to see where the energy is transferred, but where is the energy transfer in a silicon chip?
Work is done to change a 0 into a 1. That work may be stored energy if the 1 is a higher energy state than 0, but that's released when it goes back to 0.
All of that work comes out as heat.

"Work" in the more general sense means getting things done, and if those 1s and 0s convey useful information, then that is getting things done, but there is no transfer of energy.

Quote:

In theory a bumble bee shouldn't be able to fly, but it does, therefore the theory must have a flaw somewhere; same applies here.

I believe your information is out of date.
The laws of aerodynamics used to say that a bumble bee could not fly, but those laws failed to take into consideration vortexes and edge effects, or something.
Modern understanding of aerodynamics do allow a bumblebee's flight.

	Downsizer Forum Index -> Energy Efficiency and Construction/Major Projects	All times are GMT
Page 1 of 1

View Latest Posts