Hi lukeebee,
I'm sorry for forgetting to do an update! It's been some time, but I can let you know how it went. We did end up building a cooling device that worked quite well, but we had to change a few things around.
We did some calculations for heat transfer, and found that adding insulation would prevent condensation from forming on the hose if we ran the system at -20C (-4F). Unfortunately, the heatsink would be so cold that condensation would form on the BACK of the motherboard as well. Again, we could insulate for that, but the amount of work necessary was much more than we initially thought. We were designing this so that somebody could install it on their computer as easily as a water cooling setup, so this problem forced us to look at alternatives. In addition, the amount of insulation required on the hose would have made it incompatible with most holes already available on computer cases, which was another mark against the -20C build.
We decided to instead go with what we called a "dual-loop" setup. The dual-loop setup consisted of a water loop connected to a refrigeration loop. The water loop was similar to other water cooling setups on the market. We had a pump, a reservoir, and a CPU heatsink. This worked quite well, as it made our system compatible with any other custom CPU, GPU, or RAM heatsink on the market. It was exactly like hooking up a water cooling setup, except that our system was in its own box that sat beside the computer.
We used a refrigeration system to cool the water. The water reservoir and refrigeration system sat inside the "cooling box". We designed a heat exchanger to transfer heat from the water reservoir into the refrigeration loop. We used a refrigeration compressor that we selected due to its quiet operation and small size. After testing the setup, we found that it was capable of cooling up to 600 Watts, based on the duty cycle of the compressor. This meant that it would have been very easy to connect a GPU heatsink in series with the CPU heatsink.
We were still limited by condensation, but now we were also limited by the freezing point of water... Our refrigeration coils would hit -16C (-3.2F), so of course the water would freeze. Our solution to the condensation was to tune the system so that the compressor would maintain the temperature of the water slightly above the point where condensation would actually form (the dew point of the air). For a room with air at 20C (68F) and 50% relative humidity, this was about 9C (48.2F). This might not seem like a big improvement over water loops, but water loops have to run at temperatures that are hotter than the room they are in. If the room is 20C, the water loop may run at 30-40C. Our system was able to maintain the water at 9C very consistently.
We solved the freezing issue by adding propylene glycol to the water reservoir. Propylene glycol (PG) is a "food safe" antifreeze. It's used in fog machines and e-cigs as well. The good think about it was that it was also "plastic-safe", meaning we could use it with the combination of materials that were present in our system without risking corrosion. The green colour of the PG also gave the water loop a really awesome look, especially as the fluid spun around our helical-shaped heat exchanger. I think we ended up using 40%PG and 60% water. We had some slight icing in our heat exchanger, but it didn't accumulate beyond the first bit and it didn't affect the performance overall.
So how did it work? Very well. I can't remember the exact motherboard that we used. I think it was an ASUS Rampage. It was the expensive ASUS board that was available about three years ago. We ended up paying about $270 CAD for the board. We changed the overvoltage jumper on the motherboard to allow for the highest voltage possible, and proceeded to overclock an AMD 965 II Phenom X4 BE. I'm not a master at overclocking, unfortunately, but we were able to achieve some pretty respectable speeds. We almost beat the water cooling record posted on HWbot.org. Our limitation wasn't temperature, it was my inexperience at overclocking and possibly the affinity to overclocking that our specific CPU had. I've heard that some batches of CPUs are more overclock-able than other. Throughout all of this, the highest CPU temperature that we measured was 40C (104F).
All in all, it was a very fun project that performed better than we had hoped. Unfortunately, the cost of the components and assembly was too high for the system to be considered marketable. It was ultra-enthusiast and would cost more than a dual-780 rig, even after factoring in favourable production and component costs.
Feel free to ask any questions. As I found out today, I still get email alerts for this thread.