Adequate CPU Cooling

Anyone who remembers the days of the Celeron 300A will no doubt recall that extreme cooling referred almost exclusively to home built watercooling rigs. It was not long thereafter that manufacturers of after market coolers began to use ‘extreme’ to refer to anything other than the stock cooling system supplied with the CPU. In so doing they have stripped the word of most of its original meaning.

The word ‘extreme’, even if we take it to be synonymous with ‘water cooled’, seems somehow to be an inadequate description for some cooling alternatives. Certain cooling practices have proven so effective that they cool the processor to the point where it encounters a phenomenon known as the ‘cold bug’. This is the point where a processor is cooled so much that the quantum mechanical processes that make the device run are no longer able to take place.

If ‘extreme’ is inadequate to describe this type of cooling, it stands to reason that we need a different word. Thus I propose we now term any cooling system capable of reducing the temperature of an operating processor so low that it encounters the cold bug, ‘adequate’. Unlike ‘extreme’, ‘adequate’ has a rigid definition impervious to overzealous marketers.

Since the last article I wrote, what little overclocking time I have had has been spent developing such a system. Before designing anything, I began with a list of desiderata for an adequate cooling system. Like any ideal solution, it should cost nothing, weigh nothing, take up no space, and have no associated risks. My solution will not satisfy any of these in any way shape or form. There are, however, some criteria which any cooling system must satisfy in order to be in anyway practical:

• It must not damage your other hardware in any way. Pot type dry ice coolers have a tendancy to spit acetone onto the motherboard. This dissolves most of the plastics and coatings, leading to an early death for the PC.
• If it is not meant to operate continuously, it should be easy to switch to and from a more conventional cooler. Again, LN2 and CO2 pot coolers require the motherboard to be completely removed from the system.
• It should be able to operate continuously, without user attention, throughout the duration of a reasonably long game (~1 hour is the goal I set).

• The user should not be restricted by the half-life of the coolant in storage. The amount of LN2 or dry ice typically used for bench testing a computer will boil/sublimate in a day if not used.

My solution is to make dry ice from liquid CO2 at the computer when it is needed. The process is essentially the same as what happens inside a CO2 fire extinguisher. My solution is a little more elegant, less prone to clogging, and will run off any gas, liquefied or compressed including CO2 LN2 and LHe. At this point, out of full disclosure, I should mention I am related to Tyler Anderson of Iceburg Dry Ice who supplies all three. For the sake of his in box, I should also note that LHe costs $20/L or ~ $80 per US gallon, dependent on the cost of energy at the time it is made, and is not available in small quantities.

Before you try blasting the inside of you PC with a fire extinguisher, there are a few caveats you should be aware of:

• First, make sure it is a CO2 extinguisher. Foam and dry chemical will likely end up destroying your PC. If not then they will certainly decrease the cooling performance of your system and leave a respectable mess.
• Second, the dry ice and gaseous CO2 are incredibly dry and move very fast through the hose and horn. This means it will pick up a static electric charge off the inside of the plastic horn. This charge will be more than enough to destroy any PC components. You will need to ground the gas stream before it hits any components.
• Third, you will need to insulate the back of the motherboard to prevent condensation. One of the great advantages of using a dry gas to cool you system is that there is simply no water present to condense on the front of the board.
• Fourth and most important is to recognize how much CO2 you are putting in the atmosphere. Discharging the contents of a 20 lbs fire extinguisher into a 10′ x 10′ x 8′ room will raise the CO2 levels of the room well beyond OSHA approved levels.
• Fifth, the resulting dry ice and gas will be extremely cold and will damage any skin it comes in contact with.

• Sixth, if pressure is allowed to build up anywhere in the system it could result in a catastrophic failure. It is possible (though difficult) to store enough energy with liquid CO2 to equal the explosive force of a medium sized rifle cartridge. This is what will happen if you pump liquid CO2 into a conventional waterblock.

So, you have opened the doors and windows, removed any pets and small children from the room, and are aware of the dangers of handling the dry ice and the cold gas that will result from what you are about to do? You are now ready to cool you computer with liquid CO2. Here is what you will need:

• A high pressure siphon tank filled with liquid CO2. These are normally used to provide carbonation in soda fountains (on left). As previously mentioned, you could also use a CO2 fire extinguisher.
• An insulated high pressure hose (attached to tank, also included in extinguisher). The cooling effect begins at the valve so you should insulate the hose between the tank and the ‘horn’ for the sake of CO2 efficiency.
  

• A ‘horn’ included in extinguisher, custom unit shown. This is the heart of the system. This is a tube specially designed to allow dry ice to precipitate from the liquid and come to thermodynamic equilibrium. In other words, this device allows the CO2 to drop all the way from room temperature to dry ice temperature. There is one on the end of all CO2 extinguishers.

  I chose to design my own and fabricate it from aluminum. This one is designed to provide higher energy efficiency and ground the gas stream. I also redesigned the end cap to accelerate the ice in order to prevent it from clogging the heatsink, causing hydraulic lock and a catastrophic failure.

  The gauge on the opposite end is a safety feature. If the pressure on the gauge rises significantly, then the system has clogged. The gauge also provides a weak link and will break and prevent catastrophic failure if you cannot turn off the system in time. Exactly how to design and construct a horn that will provide maximum cooling without clogging the system with snow is the product of fluid and thermo dynamics calculations that are well beyond the scope of this article.

• A conventional, non-heatpipe, heatsink. At these temperatures the contents of the heatpipe will freeze, negating the effect. Ideally you want widely spaced fins to avoid clogging as much as possible.

The following two photos are to show operation and condensation respectively. Note that after ~15 s of operation it has developed enough frost from the air to completely obscure the aluminum:

Ian Anderson