Keeping computer hardware cool is a common problem sometimes requiring an uncommon solution. Traditional methods of heat sinks and air flow fail on two fronts when pushing equipment to the extreme. First they become very loud and second they do not take full advantage of the integrated circuit operating characteristic's. Higher operating frequencies are possible when chips are cooled below 0c.
My solution was to make a chilled air system which would pump air to each computer, reducing the number of fans required to one. The original plan would have the chilled air ducted to each computer on our LAN.
This photo shows a set up used to see if the idea would work. The 'humidifier' was filled with cold items which ranged from ice to chilled metal. It was determined that pumping chilled air had many advantages so I began learning how to design my own A/C unit. During the design of this system (which would have chilled the air and the CPU) the budget was reduced to a few hundred dollars so a custom made unit became out of the question.
The Kyrotech unit, which has garnished so much attention, is not suitable because it chilled only the CPU. All other parts still required conventional heatsink/fans. In fact the condenser/fan arrangement in the bottom of the Kyrotech case seemed to promise more noise, so even if it was available it would not address my main concern. The expected $500 minimum cost of the Kyrotech unit was now beyond my budget anyway.
The solution has been to take a compact (also called a mini-bar) refrigerator and use it as a computer case. First job was to get a compact fridge that would be suitable. Of the designs available I chose one with the evaporator completely exposed. Other designs did not allow such easy access to the evaporator as the Danby designer compact refrigerator model DCR080WEY-7. With about 1.7cu/ft of internal space I have enough room to install all required parts.
This is a photo of the first set up. The door has been removed and is not used. This was used to take basic temperature measurements, concentrating on the CPU heat sink. The internal parts include the power supply and hard drive. The power supply and hard drive has been attached to the bottom of the motherboard mount. Pieces of sheet metal are used for legs to position the CPU at the correct height. The inside of the case/fridge has been separated by a piece of Plexiglas and 1" Styrofoam. A piece of 1" Styrofoam was duct taped to the top 2/3rds leaving the CD-ROM and diskette drive at the bottom exposed.
The photo on the right is a close up of the CPU to evaporator connection. A chunk of 1/8" 6061 T6 Aluminum was screwed to the PII Thermal Plate and clamped on to the vertical part of the evaporator. The purpose of this set up was to collect data required to design a proper heat sink for the CPU.
This arrangement resulted in a CPU temp of -25c with the computer off. Once the computer was started the temp would begin to rise to above 0c. The -25c CPU temp did enable the CPU to boot at 374MHz. not bad for a PII 266 with multipliers no higher than 4.5.
Normally the CPU heat sink has to transfer heat from the CPU to the air. In this design the heat had to go from the CPU to the evaporator. With the the evaporator at -25c I needed a material that would conduct heat rather than transfer it to the air. Although Aluminum has a good Coefficient of Heat Radiation I needed something with a better Coefficient of Heat Transmission. It turns out that Copper is about twice as good as Aluminum at transmitting heat so the next generation has been built of copper. Silver would have been even better but I couldn't find any solid silver plates around. I wonder why?
I am now on the 5th generation of this setup. I have tried a couple of different designs for the CPU to evap connection and have settled on a straight connection, using 1/4" x 4" copper bus bar. This final design (final for this stage) has resulted in a very usable system.
This image shows the 3rd generation set up. In the upper right hand corner there is a temperature probe taped to the evap. This measures the exit temp of the gas. Just below that is a black rectangle. This is the boxed PII heat sink and fan. I was using it to cool the internal air temp. The large green square in the centre is an AWE 64 ISA card.
In the lower left hand corner you can see a tray of light blue desiccant. Actually the tray is a plate for eating boiled corn, don't worry I washed off some of the melted butter before using it;) There is about 250ml of Tell-Tale Silica Gel Desiccant spec: Mil-D-3716, Grade H, Type IV. Please contact me if you know where I can get small quantities of this. I would like to get it in $5 packages rather than 50lb barrels @ $60. This is great stuff as the color will tell you the state of activation. When it is completely activated and ready for use it is a deep dark blue color. As it gets used it becomes a light blue as it is in the picture, pink and white indicate complete use. Heat is used to reactivate. Bake at 80c for about 1hr or until all desiccant is a dark blue color, do not add butter:)
The purpose of the desiccant is to keep the internal air as dry as possible. The amount of moisture air can hold is directly related to the temperature of the air. As the air temperature falls so does its ability to hold moisture. At some point the relative humidity will reach 100% and the moisture will condense or precipitate. This will occur on the evaporator because it is the coldest object inside the case. Since the evaporator is about -25c this water does not stay in its liquid state for long. The real problem occurs when the case is opened up or when the case temp rises above 0c. Then this frost build up on the evap will melt and drip on to the motherboard.
This is not as bad as it may sound because it is almost pure water, drying off the board will result in very little residue. The water will react with contacts, like a DIMM connector, which is a problem and even though it is an electrical insulator it can cause short circuits. If you are doing this without desiccant it may be possible to place a tray under the evap to catch the water. I would think the tricky part would be preventing water running down the heat sink into the slot 1. Note: pure water is an insulator but it takes few impurities to make it very conductive. Do not get electrocuted in your bath tub while using your computer and blame me.
This image is meant to show the CPU to evap arrangement. There is one vertical piece of copper which extends up from the thermal plate to the evaporator. A horizontal piece which is under the evap makes a T connection to the vertical piece. This was meant to increase the surface contact but it failed to result in lower CPU temps. In fact the lower horizontal piece warmed the heat sink due to Radiation loses. This was significant once I started trying to cool the internal air, due to the increased air movement.
The insulation attached to the evap and the heat sink was by far the best modification. This reduces losses to the air and increases the amount of heat sink kept cold. To reduce this further I could mount the board closer to the evap but there is a 64MB DIMM sticking up. 2-32MB SDRAM DIMMS would allow for a closer mount.
This photo shows a fan inside the evap to cool and circulate the internal air. The fan and Pentium Pro heat sink are mounted to a 1.25cm thick X 10cm wide X 20cm long piece of aluminum. This arrangement rests on the evap, the duct tape was to prevent it from moving too much.
At this point the ability of the compressor to keep things cool was exceeded. With the fan running the evap temp rises quickly. This results in the CPU temp rising. It is great for cooling the unit down as an internal air temp of -20c can be reached using this fan/heat sink arrangement. It is not that useful while the computer is in operation.
While I write this the fan is off and the CPU has a temp of -9.6c, a top heat sink temp of -15.3, an evap exit temp of -12.0c, internal air temp is +5.2, video card heat sink temp of +17.9 and a HDD surface temp of +10.5. Running time is over 2hrs. I'll turn this fan on now and let it run for 3 min. Ok after 3 min I have a CPU temp of -8.0c, top heat sink -14.0c, evap exit -4.0c, internal air +2.0, video card +6.5 and HDD temp of +7.6. The air temps drop but the evap goes up because the compressor can not keep up. If I leave this fan on all temps start to approach equality. I just use this fan when the computer is off.
This image shows the final heat sink which is currently being used. A pretty good chunk of copper. It requires support as the slot 1 (rugged as it is) can not handle heat sinks measured in kilograms. Although the copper is attached using machine screws they are not tight, just snug. You can tell by the angle of the copper that there is considerable movement possible between the heat sink and the thermal plate. Though out the installation, expansion and contraction due to thermal changes means very few devices are attached tightly. Room for movement is build into the design. The CPU heat sink is snug but the holes through the copper are larger than the 6-32 screws would normally require.
So what is it like to use? Well it is quiet and as a result listening to music is much more enjoyable. The fridge takes up about as much room as an end table but not as much as some servers. I am still working out the bugs, one of which occurs on boot. It seems that WinNT wants something to be warmed up a bit first. I can boot into Win95 right away but NT takes 2 or 3 attempts before it will start with out restarting the computer. Also it has on occasion just shut down and rebooted even after running for some time. I suspect a poor connection but getting at the computer involves breaking the air tight seal and reactivating the desiccant. I don't run at 374MHz except to show off or do tests. Heating is still a problem and crashing the OS hard gets real annoying after a while. It is just easier and less time consuming to run at 333MHz (which is completely stable), at least until I complete my next cooling project.
If you do this people will think you're odd. In fact this will prove to many that your a bit 'out there'. It is a lot of fun and is the cheapest way to completely solve your current cooling problems. Even the cheapest heat sink will work when it has 0c air to work with. So if your thinking of it then , GO FOR IT. You have nothing to lose except your neighbors respect, your motherboard, your CPU, your computer and worst of all your beer cooler. Keep Kool!
Check out the tech rambling page for more info.