Supercooler
For years hardware hackers have been working to get more speed out of their computer, one of the most popular methods has been overclocking the CPU. If a chip is designed to operate at say 500 MHz at 100^F, you may be able to get it to run at a little faster if you can cool the chip to say 80^ F. This is partly because substances become better conductors of electricity the closer they come to absolute 0 (-459.67^ F, -273.15^ C). In fact even rubber becomes a perfect conductor at absolute 0.
CMOS chips likewise can run (be clocked) faster at lower temperatures. In fact if you cool a CPU down to 120 C you can safely double the speed of the chip. I hope to be able to overclock a PIII 500 MHz CPU to 1 GHz, to achieve such speeds a custom supercooler needs to be built that will cool the system and keep it cooled for an extended period of time. Below you can check out where I am in this project, if you have any comments or suggestions please drop me an email.
Phase 1
Ok, this is Phase 1, so don’t expect much. 
I have been doing a lot of research trying to come up with a good design. I looked at some people that were using phase change (liquid to a gas) systems to cool a liquid and then circulate that liquid around the CPU. My goal is to cool the CPU down to -120^ C and I could not find a pump that would pump a liquid that cold. So, what I am going to try to do is have the phase change occur right on the chip.
I now needed to pick a refrigerant, I looked at R-134a, R-22, R-502, and R-404a. I ended up going with R-404a this was chosen for several reasons, it is easily obtainable, has a low boiling point at 0 PSIg (-52^ F), and is environmentally safe. Don’t get me wrong on the last one, it is not that I am a hug a tree guy or that I believe CFCs are a big part of the “greenhouse effect”, but why contribute to it. Anyway, enough with the commercial, lets get back to the project
Perhaps the most important part of my project is the evaporator. Most evaporators are fairly large, I wanted the phase change to occur on the CPU so my evaporator had to be small and yet still be able to provide proper heat absorption, I came up with the following design.
Above you can see the design for the evaporator. I was tried to turn my drill press into a milling machine, but that did not work so well. I ended up building the evaporator using 1/2 soft copper, 1/4 soft copper, few end caps, and 1/16 plate copper. I soldered it all together, I am sure it is not efficient as my design, but it would do for testing. The main 1/2 flare connection connects to a Constant Pressure Expansion (CPE) Valve. The two outputs connects back to the suction side of the compressor. Two hoses were needed because there is move volume that needs to be removed after the liquid flashes to a gas in the evaporator.
To save time I used a off the shelf condenser unit instead of building one, I selected a Copeland condenser unit. The unite comes with a compressor, condenser coil, liquid receiver, and is charged with R-404a.
The system is very straightforward, the compressor compresses the gas into a liquid. The liquid is then cooled off in the condenser and sent to a liquid receiver. The liquid then enters the CPE valve that is set to maintain a evaporator pressure of around 10 Hg. As the liquid flashes to a gas it absorbs heat from the evaporator, it is then discharged to the compressor to start the process all over again. To lower the CPU temp farther I decided to use Thermo Electric Coolers (TECs) to drop the temp further. The evaporator is designed to fit on top of two Intel Celeron CPUs.
As you can see I was able to cool a sensor down to -73^ F with a suction pressure of about 5 PSIg, not bad for a first crack at it, but that was without a load. I ran into a big problem when I tried to put it on a real CPU. The problem is that the 500 MHz CPUs that I wanted to use have the clocks locked, so the only way to up the CPU speed is to up the buss speed. I wanted to reach 1GHz, to do that on a 100 MHz system buss I would need to overclock the buss to 200 MHz. Even if I could cool the CPU, I would not be cooling the memory, video card, etc.
Phase 2
Well Intel through a loop on the last phase. They lock the internal clock multiplier on all chips except the engineering samples so that people like me can not overclock the chip (If anyone knows where to get engineering samples please let me know). This is a big problem because it means that to 2X the speed of the CPU I can not just tell the chip to run at X faster internal clock.
I thought about just ditching the Intel platform and going with AMD (they do not lock their chips), but I have not found a dual processor AMD board and I very much want 2 CPUs clicking along at 1GHz. So, back to the Intel boards, the only way I know to overclock the chip is to increase the speed of the system buss of my ABit motherboard. This should work because the clock multiplier is locked to system buss. The downside of such approach is I have no idea if the motherboard, memory, video card, etc. can be clocked to 200 MHz even if it is cooled to -120^ C, but I am going to give it a try.
The plan is to submerse the motherboard in a high dielectric liquid and then chill the liquid down to -120^ C. I found a 3M product called 3M Fluorinert, It has high dielectric strength and an extremely low pour point of -101^ C. The problem is 5 gallons of this stuff is over $2500. It is not that I am not prepared to spend that, but I don’t even know if this would work. I started search for a liquid that I could use as a substitute, I found several that had high dialectic strength, but they also had high port points. One day I was looking through my Grainger catalog and ran into the Mobil compressor oil section. I ran into Rarus824 it has a pour point of -60^ C, sure it is no where near -120^ C, but it will get me started. I also figured that I may be able to make up the ~55^ C with TECs.
Of course I was not able to use the same evaporator to chill the liquid, so I had to come up with something else. I did not want to spend a lot of time building one so I looked for something off the shelf. I started looking at transmission coolers. I found one from Flex-a-lite, put on some new connectors and added a CPE valve.
This is what everything looks like put together.
Well I am having problems with the suction pressure on my condenser. I need it to be about 20 Hg, but currently it is about 3 PSI. Anyone out there have some suggestion on getting a lower suction pressure without building a two stage system?
