Large-scale water cooling (heat transfer)

[ddrueding]

Now that winter is approaching (it'll hit 38F tonight), it is time to re-evaluate utilizing the waste heat from my computers to heat the bedroom. The following are the reasons I know why this is a bad idea (to discourage anyone else reading this from trying it).

1. Electric resistive heating (what a computer basically is) is one of the least efficient way to heat air.
2. Putting water near computers is a bad idea.
3. Significant thermal losses will occur in the system
4. There typically isn't enough heat coming off a computer to make a difference anyway.
5. This is going to be expensive.

So, here is the plan.

1. Water cooler to act as large reservoir sitting behind equipment rack
2. Basic pond pumps sitting in reservoir
3. Simple, open-ended loops drawing water from the reservoir and dumping it back in.
4. Each loop will have a max of 2-3 water blocks on it.
----Office 1 CPU, Office 1 GPU
----Office 2 CPU, Office 2 GPU
----Office 1 HDD, Office 2 HDD
5. Long-distance run to radiator in bedroom will be made using 1/2" schedule L hard copper pipe.
6. Radiator in bedroom will be 4x120mm with low-speed fans. An additional radiator could be added as needed.
7. Overflow protection in the reservoir routes outside

The thought is that the pumps circulate enough water in the reservoir to make the water roughly uniform in temperature. That water is then brought out and cooled in the bedroom.

So long as the temperature of the reservoir is low enough to keep the parts cool it should be fine. My only worry is that getting the temperature much above 50C would cause significant evaporation.

The biggest downside to this "open ended" design is that all the water heavy lifting has to be done all the time. There is no pressure from the return water to push the supply on. Because of this, the reservoir should be mounted at approximately the height of the equipment shelves that contain the computers.

The advantage to the open-ended system is that individual elements can be maintained separately (the computers can be taken out one at a time, an additional radiator could be added without taking the system down)

More thoughts to come, but barring any major hiccups, I anticipate going through with this in the next few weeks.

Thoughts?

 

[Mercutio]

I'm pretty sure there are easier ways to make stuff warm. That sounds like a bunch of points of failure and I'm somewhat doubtful that it will help that much to actually keep your bedroom warm since heat will be lost through the whole system before it gets where you want it

 

[ddrueding]

Indeed. Last night my wife turned on a 600W electric heater in the bedroom and it cycled on and off all night; I suspect it was on about 30% of the time. According to the power calculators I've been looking at, my overvolted, overclocked CPUs are putting out about 200W each, with the GPUs the same. So even if I have a 50% loss in the system, I'm still putting 400W of heat into the bedroom, while last night 200W was enough to maintain 70F in there. The upside is that any heat lost in the system ends up heating the main room (not a goal, but not bad either) so long as I insulate the two copper runs that will be in the attic.

 

[CougTek]

4. There typically isn't enough heat coming off a computer to make a difference anyway.

Absolutely false. My room has two computers taking each less than 150W and it's by far the hottest room of this floor. Temperatures have dropped below the freezing point for several days during the night here and I've never had to start the electric heater yet. I also have two other 150W computers (one ~100W, the other closer to 200W) working 24x7 (FAH, of course) and the temperature inside has always been comfortable. Overall, my computers take maximum 600W together and I heat an entire floor with them, even when the exterior temperature is at least 15C lower than in your neck of the woods.

 

[Mercutio]

Yeah. You definitely can keep a living area warm enough. I'm more worried about the water and the pumps.

 

[jtr1962]

This isn't a bad idea and I'm wondering if it could be taken a few steps further. I'm thinking along the lines of a hot water heating system for the entire house where the water is preheated first by low-level waste heat from various appliances ( computers, refrigerators, LED lighting ) before being heated further by either resistive or heat pump heating. You could even have a loop where the water is solar heated for good measure. In theory you may even be able to get by entirely on waste/solar heat.

There is one big difference in concept between this and a conventional hot water heating system. In a conventional system, there are obviously only conventional heat sources (boiler, heat pump, resistive heating). And generally the pump only runs when the thermostat calls for heat, not all the time. In truth the pump running all the time isn't a waste. Eventually all the pump power just goes into heating the water further.

The other big difference is the water temperature. In a normal hot water system, it must be high enough so the radiators in each room transfer enough heat to the room via convection. Figure then maybe 70°C. With your idea the water temperature will be much lower. That's not a problem so long as it's hotter than the rooms being heating, but you will need forced circulation via fans (as you figured out already).

I'm a big fan of taking advantage of low-level waste heat. In some cases the savings goes further than the heat you recover. If you liquid cool your refrigerator, the compressor will run less, saving money/wear as a result. And PCs running cooler is always a good thing.

 

[ddrueding]

I'm more worried about the water and the pumps.

As am I. There are a few ways to mitigate this, and I think the easiest one is a thermal sensor on a waterblock that kills power to the machine if it exceeds a threshold. No need for software or anything fancy, just a sensor and a relay. I should start looking for those, but these machines have 1200W PSUs, so it should be something substantial.

 

[ddrueding]

Thinking about this further, the reservoir could be above the devices to be cooled without increasing the energy required to move the water, correct?

 

[BingBangBop]

#1: I do not understand why electric resistive heating is any less efficient than any other form of heating. A watt is a watt. Is the process of converting electricity to heat energy via computer somehow creating some other form of energy like light that is significantly robbing its efficiently? The law conservation of energy still applies, even to computers.

#2: Water is only bad if it comes in direct contact with the electricity. Are you planning on doing that?

#3: What difference does it make if you have thermal losses in the system. The whole goal is to heat and what is thermal loss other than a conversion to heat so it is only a loss if that heat goes somewhere you do not want to heat.

#4: The amount of heat generated by a computer depends a lot upon what it is doing, what its designed to use and of course how it was OC'ed. Of course a computer that is turned off will generate little to no heat. An i7 will generate less heat than a c2d. An OC'ed CPU/GPU will generate much more heat than non-OC'ed. If you try, I'm sure you can easily generate 300-500W per computer and if there are multiple high end modern GPU's it can be a lot more that that.

#5: Is definitely true but only if you are actually buying the computer as a heater. The original purchase price of a computer is much more expensive than an inexpensive electric heater and maintenance is also more expensive.

That being said, the price of electricity for a given amount of heat should be very close to the same, if not identical to a cheap baseboard or electric space heater. Really, the heating cost just depends upon your electricity costs assuming you already have the computers and if electricity is expensive where you live (compared to other energy forms like natural gas) then yes it will be relatively expensive.

Why an open loop? There is the problem of evaporation. There is the possibility that the water does not get hot enough for the radiator to efficiently collect and dissipate much heat rather the evaporative process may end up being the primary cooling process. You'll have to manually add water to the system regularly. The possibility of to much humidity for comfort. If you add enough water to the air then condensation on any cool surfaces may become a problem. If you have a tight house, high humidity can cause mold problems because the water has no place to go other than to sit on surfaces creating a good environment for growing mold.

 

[jtr1962]

As am I. There are a few ways to mitigate this, and I think the easiest one is a thermal sensor on a waterblock that kills power to the machine if it exceeds a threshold. No need for software or anything fancy, just a sensor and a relay. I should start looking for those, but these machines have 1200W PSUs, so it should be something substantial.

These are about as simple as you can get. I'm actually using something similar on the water block of my thermoelectric temperature chamber as a failsafe in case water flow is cut off. Mine came with some surplus TEC assemblies I bought.

 

[BingBangBop]

Thinking about this further, the reservoir could be above the devices to be cooled without increasing the energy required to move the water, correct?

At the same level is best! With any upward section of the tubing the pump has to work against gravity expending more energy therefore needing a bigger pump to compensate. You don't get that energy returned (to the pump) on the downward sections of the tubing.

 

[LunarMist]

Oh for Pete's sake, just hire an engineer. This is not the pioneer days or MacGyver. In most areas electric heat is very expensive compared to gas or even a heat pump (which works well if the outdoor temperatures don't get below ~freezing).

 

[BingBangBop]

Oh for Pete's sake, just hire an engineer. This is not the pioneer days or MacGyver. In most areas electric heat is very expensive compared to gas or even a heat pump (which works well if the outdoor temperatures don't get below ~freezing).

Hiring an engineer will successfully make it expensive, even if it wasn't beforehand.

I thought Ddrueding was an alias of MacGyver's.

 

[jtr1962]

Oh for Pete's sake, just hire an engineer. This is not the pioneer days or MacGyver. In most areas electric heat is very expensive compared to gas or even a heat pump (which works well if the outdoor temperatures don't get below ~freezing).

The goal of this project seems to be to recover the heat from machines which will be running regardless. Obviously if heating is your only goal it's easiest to just use a conventional heating set up.

Incidentally, the one PC I have on in my bedroom 24/7 seems to make that room about 5°F warmer than the rest of the house. Power usage for this machine is only in the 100 watt area. The house normally holds about 10 to 15° F over ambient without supplemental heating, so my bedroom remains comfortable without heat until temps outside get under about 45-50°F.

 

[sechs]

This seems like a really Rube Goldberg way of moving heat. You might as well set up a system to use the heat from the computers to warm bricks, and then move them to the bedroom at night; it would be far more efficient.

Why not just circulate the warm air?

 

[Mercutio]

a computer that is turned off

I recognized all these words but they make no sense in this order.

 

[LunarMist]

The goal of this project seems to be to recover the heat from machines which will be running regardless. Obviously if heating is your only goal it's easiest to just use a conventional heating set up.

It is opposite of how I run projects. I would establish the objective to be: Make the internal home environment comfortable for the human inhabitants. Then I would define the user requirements, functional requirements, etc. I could go into the usual 45-50 minute design control speech, but the point here is that the method of heat transfer should not be determining the objective up front.

 

[ddrueding]

#1: I do not understand why electric resistive heating is any less efficient than any other form of heating. A watt is a watt. Is the process of converting electricity to heat energy via computer somehow creating some other form of energy like light that is significantly robbing its efficiently? The law conservation of energy still applies, even to computers.

One watt/hour of electricity put into a computer, resistive heater, or lightbulb produces one watt of heat. One watt put into a heatpump produces (technically moves) several watts of heat. If I wasn't already using the electricity to do something else, I'd be using a heat pump.

Why an open loop? There is the problem of evaporation.

Thinking on this more, the reservoir should be sealed. Not pressurized, but vapor sealed to reduce evaporation.

 

[ddrueding]

These are about as simple as you can get. I'm actually using something similar on the water block of my thermoelectric temperature chamber as a failsafe in case water flow is cut off. Mine came with some surplus TEC assemblies I bought.

Nice tip! I'll look into those.

 

[ddrueding]

At the same level is best! With any upward section of the tubing the pump has to work against gravity expending more energy therefore needing a bigger pump to compensate. You don't get that energy returned (to the pump) on the downward sections of the tubing.

If the reservoir were at the highest point, the mass of the water on the downward leg would be pressing upon the water on the upward leg. No?

 

[ddrueding]

Oh for Pete's sake, just hire an engineer. This is not the pioneer days or MacGyver. In most areas electric heat is very expensive compared to gas or even a heat pump (which works well if the outdoor temperatures don't get below ~freezing).

That wouldn't be much fun, and would be outside my budget. I would absolutely use a heatpump if I weren't already using the electricity on the computers.

 

[ddrueding]

This seems like a really Rube Goldberg way of moving heat. You might as well set up a system to use the heat from the computers to warm bricks, and then move them to the bedroom at night; it would be far more efficient.

Yes, but it is hard to sleep while moving bricks. I suppose with training....

Why not just circulate the warm air?

That is what I'm currently doing, though at night there is no need to heat the main room (5000+cu.ft.). And while the computers aren't enough heat to do the whole house, I'm sure it would be enough for what we need (the bedroom).

 

[ddrueding]

It is opposite of how I run projects. I would establish the objective to be: Make the internal home environment comfortable for the human inhabitants. Then I would define the user requirements, functional requirements, etc. I could go into the usual 45-50 minute design control speech, but the point here is that the method of heat transfer should not be determining the objective up front.

I was tempted to not even include the water cooling in the first post, and start with my objective:

"My computers produce some heat that I would like to use to heat my bedroom, and I can't move my computers".

Water just seemed like the best way to do it, though sechs' bricks might get a second thought (trained pets?).

 

[BingBangBop]

This seems like a really Rube Goldberg way of moving heat. You might as well set up a system to use the heat from the computers to warm bricks, and then move them to the bedroom at night; it would be far more efficient.

Have you considered the benefits of installing a Pizza Oven in your bedroom as your heating method? Seems to me that eating a fresh pizza in bed outweighs the benefits of computers (or warm bricks) in your bedroom and both will add heat.

A secondary heating method that may be desirable involves hot and steamy (assuming you wish to add humidity to the air) sex :wink:. Again a much more enjoyable heating method than running computers 24x7.

 

[Mercutio]

And while the computers aren't enough heat to do the whole house, I'm sure it would be enough for what we need (the bedroom).

Something something something make your own heat something something bedroom something.

 

[Howell]

Compost piles put off lots of heat as they are decomposing. Have you considered stashing a compost pile under the bed?

Sorry this doesn't help with the heat removal aspect.

 

[Bozo]

Add a little anti-freeze to the water. It increases the thermal transfer and helps mitigate evaporation.

You could use a conveyor belt for the bricks. :smile:

 

[Howell]

Add a little anti-freeze to the water. It increases the thermal transfer and helps mitigate evaporation.

You could use a conveyor belt for the bricks. :smile:

Anti-freeze in no way improves heat transfer rates. Water has a much, much lower viscosity so you'd waste energy at the pump; and I wouldn't think there would be any danger of boiling or freezing. Also, in a closed system evaporation is either not possible or irrelevant.


Good idea about the conveyor though. I was thinking about the possibility of reprogramming an army of Roombas.

 

[LunarMist]

Antifreeze has less heat capacity too. Depending on the system it has other useful properties though.

 

[Stereodude]

I don't have any specific ideas, but I must admit this thread and all the responses contained herein made me chuckle this morning.

 

[ddrueding]

Yup, whether it happens or not, it is a fun thought experiment with plenty of humor

 

[Bozo]

Anti Freeze. 'Other uses' second paragraph.

 

[Howell]

Anti Freeze. 'Other uses' second paragraph.

Even including that paragraph being poorly written from a scientific standpoint it backs me up. The first para describes that ethylene glycol and propylene glycol are both used and then goes on to describe a downside of ethylene glycol. The second para describes in what specific ways ethylene glycol has advantages over propylene glycol despite the fact that its toxic. But that doesn't change the fact that water is 17 times less viscous than ethylene glycol and 50 times less viscous than propylene glycol.

I'm not trying to get into a petulant pedantic pissing contest , maybe you could describe better what you mean by "increases thermal transfer".

 

[time]

No wonder you don't have an avatar - your head wouldn't fit in the frame.

He's right, Bozo. Here is a clearer source of information.

 

[CougTek]

Oh well...

 

[LunarMist]

Who is that supposed to be? Is it a Halloween costume?

 

[ddrueding]

Looks like this is getting scrapped. I was putting the budget together for it, and was informed that that money is to be redirected to a proper hydronic underfloor system.

A more sane option, no doubt, but less exciting. If anyone here knows anything about underfloor heating, I'd love to hear about it in a new thread in the B&G.

 

[CougTek]

Who is that supposed to be? Is it a Halloween costume?

First sentence , two posts above yours.

It was a joke. It's no longer a joke when you have to explain it.

 

[Howell]

No wonder you don't have an avatar - your head wouldn't fit in the frame.

Just for you; its all that would fit in the frame.

 

[Howell]

Who is that supposed to be? Is it a Halloween costume?

My halloween costume was actually a picture of the back half of a donkey package taped to my t-shirt. Untucked t-shirt and other dishevelment included. I knew everyone else was probably going to be putting significant effort in and I don't really like costumes.


It was my half-assed costume.