Cooling System Engineers: Computer in a Refrigerator?

Factual Questions
1

[sup]Anthracite: I do mean engineer in the sense of someone who can do the math around how/why a fridge works, not just someone to move the fridge.[/sup]
I’m moving into a house where there is plenty of room in the garage for me to set up my computer equipment. This would be ideal for things like DSL/Cable modems, firewalls, and any servers I might want to use, since they aren’t particularly quiet or attractive.

The problem is, in the summer the garage can get up to about 120 degrees F. The inside of the house stays nice and cool, but the garage is not so well insulated. I started thinking about crazy things like putting the CPUs in a small box with an air conditioning unit facing into the box and radiating into the garage.

Then it occured to me that I could take an old refrigerator, cut a small hole in the side, thread cables through, and stick CPUs in the fridge to keep them from overheating. I don’t need them to be at 40 degrees F, just some level below ambient, say to a maximum of 80 degrees. Ideally to keep them around 60-70 degrees or so. If I wanted to get fancy I could put a sensor in the fridge to monitor temperatures, and if they get above a certain level, shut down the CPUs ( in case of fridge failure.)

Anyway, I’ve searched the web and found a site dedicated to overclocking that has a dead link that claims to be about a computer in a refrigerator, and silly things like this:

This is not a valid concern, because things only draw condensation when you take them out of the refrigerator (assuming modern frost-free designs.) Also, a hot CPU in a cold refrigerator is not going to draw condensation - a cold CPU cooler in a hot computer case will. As long as I don’t open the door, no condensation. Also, if I’m running at 60 degrees instead of 40, I’ll have less worries if I do have to open the door.

What this all boils down to is a few simple questions that I can’t seem to find the answer for on the web.
[ul]
[li]Has anyone successfully done this before? There are dead links on the web that hint at others trying, but no solid data (i.e. “anecdotes”) I could find.[/li][li]Refrigerators are designed to pump out only the heat that has ‘leaked’ in. What is their heat dissipation capacity? Ideally this would be some figure in Watts I could compare directly to the power consumption of a computer to get a rough idea of whether the fridge could keep the computer cool. I understand that putting a heat source in the fridge would increase the duty cycle of the fridge, I just want to get a feel for what that increase might be, and how much I could shove into it before it hit an unsustainable level.[/li][li]Which brings us to… What is a maximum reasonable duty cycle for your typical fridge? My fridge seems to kick on for 5 minutes every hour or so, but that’s just a guess on my part - no solid data. Would the thing burn out in a day if I ran it at 50% duty? I understand that if I run it at 10 times the normal duty cycle I’ll get reduced life, but I want a duty cycle that will not directly cause immediate failure that is not due to wear and tear (i.e. overheated compressors, bearings drying out, plutonium core meltdowns, etc.)[/li][li]Does the ‘coolness’ setting dial adjust temperature or duty cycle in most fridges? In other words, is there a thermostat in the fridge, or does it just time itself and depend on ambient not changing a whole lot? I’d imagine there are some of each. Duty cycle would be OK for me because it would be acceptable for the temperature of the CPU to range up to 80 degrees or so.[/li][li]Does asking this increase my geek factor?[/li][/ul]
For example, I’d love to find out that the Kenmore Belchfire 197 Fridge, model year 1992, can dissipate 500 watts when the compressor is running and typically is used at a 5% duty cycle, and is adjustable up to 50% duty cycle (no thermostat.) and will last about 50% as long when run at high duties as when run at low duties. I would thus assume I could shove about 200-250 watts worth of computer gunk in the fridge and run it at 50% duty cycle and get similar temperatures to my kitchen fridge.

I would not pursue this if I find that a fridge dissipates 50 Watts when running, usually runs at 1% duty cycle, and will cause every molecule in my house to explode outward at the speed of light if run over 20% duty cycle. That would be bad.

Any help out there?

2

It’s been a damn long time since I took thermodynamics, so I really can’t help much there. I would say that you could probably put a hydrometer, and perhaps even a small dehumidifier in the thing if necessary to prevent condensation. Hopefully, someone will be around to help in the rest of it soon. I’m still wondering how much stuff you have that it would be 250 watts.

Also, it would probably not be hard to rig the refrigerator to run at whatever duty cycle you want. I would think that a freezer, especially one of those big floor models, would have a pretty sturdy compressor.

3

not qualified in this field, but have a mate who’s done this. everything ran an absolute dream! He simply cut a piece of the magnetic strip off the door so the cables could exit, turned it up/down (whichever one was the warmest!) and left it there night and day! Had no problems with it in the least!

4

Knew I’d seen something about this somewhere, try this site.

Somehow I doubt this procedure is recommended, though!

5

Thanks, Steve. They don’t seem to have run the computer for extended periods of time, just long enough to overheat the processor. But, perhaps I’ll try to get in touch with them and see if they did any non-overclocked endurance runs to see the effect on the freezer temperature.

I’m beginning to like this idea more and more. Hopefully it’s not impractical…

I was thinking two systems, a pentium 120 MHz and a Sparc20. The Sparc is pretty power hungry. I don’t have exact figures, I was just speculating, but I’ll do some measurements before I jump into this with both feet. I’m already thinking about wiring extra circuits into the garage to handle the power demands, what’s one more 15 amp circuit to handle an overloaded fridge?

Anyone have some real data on fridge performance?

6

I doubt a normal refrigerator could provide enough cooling. My 20 cu.in. Kenmore fridge at home only draws about 200 Watts when the compressor is running, of which maybe (WAG!) 50% goes to actual cooling. It tends to run at 40% duty cycle or more.

On the other hand, you can count on pretty much all of your computer power going to heat, so if your PCs draw 200 Watts total, it’s all going to heat.

I’d guess that you’d drastically reduce the life by running it at full duty cycle- a typical Sears Kenmore fridge (www.sears.com) is rated for about 1000 kWh/year, which is an average of 114 Watts.

Maybe there’s something you could rig up with a small air conditioner?

Arjuna34

7

To find out how much heat a refrigerator can remove, multiply the EER by the motor rating. The EER will be on the Energy Usage tag stuck to the fridge when you buy it. It should be around 8 or 9. The units of the EER are BTU/watt-hour. This preposterous unit was saddled on us by the Federal Government back in the ‘70’s, during the first energy crisis.

If you prefer a true ratio, divide the EER by 3.414 w-hr/BTU. This gives you the coefficient of performance (COP), which is what refrigeration engineers actually use. An added advantage is it will give you the maximum heat load in watts, which is what you want. Your COP should be around 2.5, so if your compressor motor draws 200 watts, your refrigerator can handle a 500 watt internal load, less leakage, which should be slight anyway in your case because your thermostat will be set to “warmest”. Yes, a refrigerator moves more heat than it uses in the process.

With a 200 watt internal load the fridge should run about 50% of the time, not a bad service factor for a compressor. Don’t worry about melting down the neighborhood and such.

Once you decide a fridge will do the job, go to a beer supply store and get a keg hose pass-thru. Install it in the side of the fridge, and run your conduit thru it. Happy surfing.

Ruth

8

Here’s one more option for you to consider. You said just a little cooler than room temperature would be fine. Some of those new electric coolers(Like those Coleman thermoelectric coolers with a heat pump.) will drop the temperature 40 degrees below ambient. They also have an internal fan designed to keep circulating the cool air around. They also seem to work fine running for days on end(umm although the car battery it was attached to didn’t :slight_smile: ) I have no idea if they could deal with something actively creating heat, or if they are just designed to slowly counteract heat that seeps in, but it’s an idea If a little cooler is all you need.

9

Excellent info, TNTruth. Obviously it’s been awhile since my last thermo class :slight_smile: For some reason I was thinking that the refrigerator couldn’t move more heat than it consumes :rolleyes:

Arjuna34

10

Ruth, you are dynamite! That’s exactly what I needed to know. Even now that you have provided it for me I can’t find listings of EERs or COPs for refrigerators, but at least I have a ballpark figure and I can do some experimentation.

I’ll have to scrounge up a refrigerator at a garage sale or something and work on making my computer shelter. I really do think I need to add a temperature sensor/shutdown mechanism, as a computer in a non-operative refrigerator will rapidly cook itself, I believe.

Also, when the weather is cool I can just leave the door open, or remove it, and turn the fridge off. Yes, I do believe I’m going to have something to keep me busy when I move into the new place.

11

Yes. Posing this question has increased the geek factor 100%

12

Go to http://www.3dcool.com and see all the selection they have to freeze you CPU really good. They have some amazing devices that will allow you system to run frostie without putting it in a fridge.

13

Putting a constant 200 Watt heat load in a fridge is kinda wastefull in terms of energy. A standard residential fridge, the type that you would get at a garage sale, might have a COP of 2 or so, depending on it’s vintage and the outside air temp. Any rating, such as EER (which I thought was just for A/C), is based on a standard temp, and it’s not 120 F. Humidity also could be a problem due to air leakage, and if you do open the door during cold times it will be more of a problem. You can minimize the humidity with baking soda or something else to absorb moisture, but you can’t really eliminate it.

If the house stays cool, couldn’t you put the cases inside on a rack and run the various cables into the garage? Put them in an out of the way place and put a small fan on them or something. An air conditioner has a much higher COP (and EER) than most refrigerators because it would normally works with a lower change in evaporating and condensing temperatures, and therefore refrigerant pressures.

14

That will keep the actual CPUs cool ( I mean the processors ) but I think douglips would rather be assured that all the components are cool. I can’t give you formulas to help, but I can at least relate my experience ( this will increase my geek factor by about 200% )

I used to overclock a Celeron 433 to 645 stable using a peltier cooler, a homemade aluminum waterblock and 120 gph pump, tranny hose and a tranny oil cooler with fans mounted on it.
The water temperature would of course rise over time and so would the temp of my CPU. I therefore had the bright idea that I could put my water reservoir in a bar fridge to keep it cool.
After running all day the water was at a pleasant 3 C, but it would take only about 15 minutes to heat up to about 30 C. The fridge at this point was completely useless and couldn’t even cool the air any more. Granted a 78W peltier cooler puts off some serious heat, but the complete failure of my little experiment made me think that your consumer-grade fridge won’t do the trick. You’ll have a bigger fridge, but you’ll be putting the entire box in it, so you’ll have the heat of the CPU as well as the power supply and all the heat-disspating components on the cards and motherboards.

In the end I coiled up 25 feet of soft-copper tubing, put it in the freezer part of the bar fridge, sealed up the freezer, then moved the whole rig so I could hang the tranny oil cooler in front of the air conditioner with 2 100 cfm fans sucking the air through the fins. Geek overload! I think you should consider a chest freezer if anything.

15

I agree it will be using more energy than if I had the machines in the house. However, I don’t foresee needing to use the refrigerator more than about 10-15% of the time - most of the year the temperature is just fine in the garage, and during the summer it is OK at night. All this might be more trouble than it’s worth, though.

Yeah, I will have to worry about it, but I think it will be OK because I live in a dry climate, and I won’t have the inside of the fridge cold, just cooler than ambient. I may have to think about a dewpoint sensor, though, to determine if there is a problem. The procedure for opening the fridge could be to power it down and wait for the temperature inside to rise to near ambient before opening. In any case, the inside of the CPU will be warmer than the rest of the fridge, so I’m not worried about condensation inside the computer - I may need to put a ‘tent’ over the computer to direct drips away from it, but as long as I don’t get drips I’m not too worried.

That’s definitely the smart way to go. The question is, can I find such an out of the way place that will be satisfactory to my wife? I do have some ideas, my current set-up is to have them hiding under desks and such. I monitored them for a while and convection seems to be OK. The problem with this is if they are in a room that would benefit from silence once in a while, they will become a problem. I currently have to power down my systems when we have overnight guests, because of the noise.

But, if I can dedicate some space to ‘office’ functionality, I can probably stick the machines in there under my desk. I’ve got the networking infrastructure in the house so that it will work fine, but I do need to consider some additional power circuits as the house is old and about 4 rooms and the garage are on one 15A circuit, so when the laundry is running the lights dim. But, I need the power circuits anyway, as the garage has the same problem, and adding a fridge just adds to the problem.

I’m not convinced I’ll do the fridge thing, but I’d rather do that than suffer marital friction. Actually, I’d rather just stick the machines in the garage and let them overheat than suffer marital friction. My wife doesn’t deserve headaches.

Keep in mind that you were overclocking, which increases the power consumed/heat dissipated by the CPU by a large amount. A typical CPU power supply is rated at 200 W, and that has to support power-up peak consumption of all the hardware in the box. My guess is that while running the entire computer is consuming no more than 75 W, though I don’t have a cite to back me up. Perhaps I’ll measure it. Your 78 W peltier is giving that much heat to the refrigerator by itself, add to that the insane power requirements of an overclocked CPU, and you’re probably trying to put about 125-150 W into a little tiny fridge. If I have two computers in my fridge I’ll be using about the same amount of power as I’m guessing you were, and it’ll be a bigger fridge AND I’ll be cooling it less (20-25 C instead of 3 C) although my ambient might be higher (32 C instead of 20ish C)

The upshot of all this is I remain optimistic that this concept could work, but I will probably try to avoid putting the boxes in the garage for simplicity’s sake.

I’ll let everyone know how it works out.