How-To turn a fridge into a sub-zero CPU cooler. – Rob Oats
Having used a Global Win FOP32-1 with all sorts of different fans, I managed to get my Duron 800 overclocked to a respectable 1060 MHz. Finally, the noise just got too much for me. I had a continual buzzing in my ears which reminded me of days spent with high power rifles and light machine guns simulating battle situations in my younger days. I really had two choices:
- Stop overclocking (not an optio) or
- Water cooling.
I took the plunge and got myself a Be Cooling jagged edge water block and a Maxijet 750 aquarium pump. I set up a bong type arrangement and achieved working processor temps of around the 30C mark at 1060. I was able to get the system up to 1080 at these kind of temps. I, like most overclockers, am just plain greedy. I should have been plain satisfied with a 35% overclock and no noise. Well I wasn’t.
There was 1100 or even 1200 waiting to be achieved. After a few weeks of squeezing that odd degree out of the system, I decided that I just was not going to get there at these kind of temps. Having read several different articles about CMOS performance enhancement using sub-zero cooling, I decided that I had to find a way of getting there.
The theory is that if you can achieve -50C, then the chances of a 50% overclock are good. I looked at the Pelts; the more I read, the more complex it looked and it was looking like it would be a very expensive option once I started buying big power supplies, etc.
Well my budget doesn’t go to these kind of 3 figure sums just to prove a point. So my thoughts turned to alternative methods. Having done a fair amount of research about phase change and the cooling effects regarding evaporative cooling, I figured that I must have a go at refrigeration.
I didn’t want anything fantastic, but if I could get into the pelt arena and get some temps around the 10C mark, I would be more than happy. British society, not unlike American society I am sure, is very wasteful. We have a local recycling centre nearby where we take all our hazardous waste including fridges [ED Note: Could not agree more! An untapped resource if you have not tried it.]
Off I went, down there on a mission to find an under-counter unit with a freezer box and a fairly powerful compressor. When I arrived, I asked one of the chaps in the office whether he had any and how much they were. He said “stacks around the corner mate and £5 each.” I thought it was Christmas come early. I had a look and sure enough, there were about 40 of them in different shapes and sizes.
Having found a pretty clean looking unit with a 110W compressor and R12 refrigerant, I paid the chap his £5 and loaded it up. Now, if you are going to move a fridge, then the fridge should ideally be moved in the upright position. Liquid refrigerant can cause terminal damage to a compressor if it gets into it.
If you cannot move it upright then it should be moved on its side with the return pipe from the freezer to the compressor at the bottom and the compressor on the top. That way it minimises the chance of refrigerant from the condenser finding its way into the compressor. If these terms are foreign to you, I will be discussing the various components shortly. If the fridge is moved, it should be left to stand for 12 hours before switching it on.
Refrigeration is phase change like a bong. What is meant by phase change is that substance changes physical form. Examples are water changing to vapour, steam or changing to ice. The amount of energy required to change water to vapour is 500 times more than is required to raise the same amount by one degree Celsius.
In the diagram: Phase 1, saturated vapour of refrigerant enters the compressor. In Phase 2, the temperature and pressure is elevated by the compressor and the compressed gas enters the condenser where the heat is dissipated to the air and it becomes a high pressure liquid. In Phase 3, this high pressure liquid passes into an expansion valve where it turns into a low pressure liquid.
In Phase 4, this low pressure liquid is forced into a capillary tube and fed into the evaporator. Because the liquid is fed through a very small tube, the sudden expansion or creation of a sudden partial vacuum causes it to boil and become a gas. This phase requires a great deal of energy and therefore heat has to be absorbed from the surroundings in order to achieve this. The gas is returned to the compressor and the cycle starts again.
Here’s the condenser which you are probably familiar with:
Here’s the expansion valve with the capillary exiting from the bottom and the compressor behind it:
The evaporator is something everyone will be familiar with because that is the icebox/cold plate/freezer found in the inside of every fridge.
Now that everyone is up to speed and know the terminology, I will proceed with the project.
I was very conservative to start. I used a small 10 liter plastic container in the bottom of the fridge with the intention of holding this at around 0C. In order to cool the water coming into the system, I decided to pass it through a bath at -20C in the freezer. I coiled some copper tube and immersed it in a mixture of 40% antifreeze in a basting tray in the freezer. I used the same concentration in the coolant.
Because condensation would occur, I had to insulate the water block and pipes. I took an old mouse mat which is 2mm closed foam and glued an initial layer to the water block on all exposed surfaces. I added a second layer and then taped the whole with electrical insulation tape. I covered the whole area of ZIF socket on the back of the motherboard with a thick layer of twin pack epoxy. This epoxy is non-conductive, chemical resistant, water and sea water resistant and can withstand temperatures from -50C to 180C.
I also sealed the inside edges of the ZIF socket with the same epoxy and then made a central pad of the same mouse mat material to go inside the socket. I made a mat to go over the outside of the ZIF socket on the top of the motherboard. I used dielectric grease on the socket and also covered the top of the processor with it.
Having installed the processor and the water block, I jammed small wedges of foam into the gaps between the water block and the processor to make it air tight. Having lagged the inlet and outlet pipe with ordinary closed foam pipe lagging from the local hardware store, I was ready to go. I let everything cool off overnight and was ready to go in the morning.
Rob Oats – Southampton, UK
The main tank temperature was at 1C and the freezer tank was at -22C.
I fired it up and all was fine for about 15 minutes and the temperatures started to creep up slowly at the rate of about 2C/hour. So while I was able to achieve a processor temperature of around 12C initially, it slowly rose over the hours. This wasn’t working.
The problem lay in the heat transfer in the freezer – there just wasn’t sufficient cooling capacity there.
The bath was heating up faster than the evaporator could cool it off. The reason was the area I was trying to use. The bath was sitting on only 40% of the available cooling area and there was the problem of multiple heat transfers – from coolant to pipe to solution to container to evaporator. I reasoned the only way forward was to shorten the path and get straight coolant to evaporator transfer.
So off to the hardware store again and I bought a 20L plastic storage container of the correct dimensions. I then unhooked the freezer from its mounts and trimmed it with a pair of tin shears so that it was able to fit comfortably in the container. I then gently dropped it into the container.
The freezer is a U shape and this was mounted with the legs downwards and the majority of the evaporator coils would be immersed in the coolant. I have tried it on its side since then, but the layer of coolant above the leg of the evaporator is too shallow, causing it to freeze, and I was losing cooling efficiency.
I sited the pump between the two legs and put a T piece on the return pipe so that returned coolant flows past the outer sides of the legs and then comes back along the inside into the pump intake, thereby maximising the evaporator surface area (pic below). The bags at the back contain polystyrene chips. Under normal conditions, the whole tank is covered in these bags to minimise transfer of heat from the air inside the fridge.
Because I was going to drop the temperatures even lower, I decided to add additional insulation to the processor. I took a length of the pipe lagging and cut 4 vertical sections. These I was able to squeeze in between the motherboard and the edge of the plastic clamp so there was a nice snug fit and air is unable to creep in through the bottom. I made corner sections to seal off the 4 corners. Below is a picture of the finished product:
So with everything connected the system looks like this:
Then the real testing started. With 16L of 40% antifreeze solution in the tank, I again allowed it to cool overnight. The system actually cools at the rate of 2C/hour with no extraneous heat being added. It takes about 48 hours for temperatures to stabilise properly. I was keen to get going next morning so with tank temperature at -16C I got started.
I was really impressed: The temperature dropped slowly until it was -14C and then the compressor switched on and it started recovering at the rate of 1C/hour. Within a couple of hours, the temperature stabilised at around the -17C mark. I noticed that the compressor gets pretty hot, so I have cooled it with a 80mm fan which handles the job very well. Below is the tank thermometer reading.
While the compressor runs most of the time, it holds the temperature at around the -17C quite comfortably with the system running under full load. The processor is giving a reading of -2C to -3C under full load. The irony is that at these low temperatures, the water block is starting to contract and the pressure exerted on the processor tends to decrease; therefore temperatures are lower while working than when idle.
When the system is put into idle the temperature remains unchanged for a few minutes and then generally climbs 1C. Below is a screen capture of the Motherboard Monitor dashboard:
I am still tinkering with the system looking for another 5C. The one thing I have noticed is that sufficient antifreeze must be added to prevent ice forming around the evaporator. Once this occurs, there is a total disruption of the cooling effect. As long as the coolant is kept fluid, it cools very well.
Definitely! I got 1100 and some change. I reckon with another 5C, I may even get close to 1200. The system is posting and going into the operating system at 1150, but is giving errors in Prime ’95, so it’s not stable. I have had no crashes at all since putting the sub-zero cooling in; as far as stability is concerned, it is far more stable than with air.
Here is a composite of SiSandra benchmarks – I am so impressed with this little Duron! It just keeps delivering the goods no matter what and just appears to get better and better. Not bad for a £45 processor.
The final cost of the project £15 or about US$21.
Am I satisfied? No, not ever, I want more. I am busy doing some serious research to find out how I can convert to direct on die refrigeration and just maybe at -50C I’ll get 1600.
For those interested in going this route, I found a site with on-line calculators to determine what the refrigeration requirements would be for any given setup in terms of compressor power and coolant flow rates:
A basic requirement would be to know the delta of coolant entering and leaving the water block. Armed with that, go to this site and look for calculators at the top. Another window is opened and you will see a number of different and useful on-line calculators, but go down to water chiller. By entering the various figures, like the delta and the flow rates, one can calculate the compressor power required to cool your rig. I have entered my info and the power is spot on for my setup.
Rob Oats – Southampton, UK