DaBit

04-06-2003, 10:12 AM

Not sure if it belongs here or in the Overclockers section, but I want to reach the Prof, so I will try it here. Feel free to move it if appropriate.

We overclockers like to chill our CPU, which is basically a surface measuring a quarter of a square inch, putting out ~100W of heat. Designing an evaporator which is able to efficiently remove the heat is a hell of a job due to the enormous energy densities and space restriction: an evaporator made for mounting on the processor cannot be much larger than 2"x2.5"x2" (WxLxH).

Now, many creative designs exists. The spiral or maze-like channel machined into a block of copper is very popular. Gary's idea of a sunburst fin pattern with injection of the refrigerant at the edge (multiple injection points) of the block and suction at the center also has potential.

Just one thing bothers me. How to design the channels? We need to pack a lot of surface into a tiny package. Maximum channel cross-section is set by minimum refrigerant speed required.

But what about minimum channel size? There must be a limit on the minimal channel diameter. Take a 1/4" pipe as an example. We can split it in two parallel pipes with their cross-section area each half of the cross-section area of the 1/4" pipe. We can split it into 100 parallel pipes, the cross-section area adding up to the area of the 1/4" pipe. Maybe add a bit more to compensate the added friction.

The 100-pipes example provides much more refrigerant->metal surface area than the 1-pipe example. Refrigerant speed is also equal, assuming 100% perfect distribution. Pressure drop is a bit higher due to added friction.

But where is the limit? Can I make millions of channels, each having a few square microns cross-section area? If not, what limits the minimal channel size?

We overclockers like to chill our CPU, which is basically a surface measuring a quarter of a square inch, putting out ~100W of heat. Designing an evaporator which is able to efficiently remove the heat is a hell of a job due to the enormous energy densities and space restriction: an evaporator made for mounting on the processor cannot be much larger than 2"x2.5"x2" (WxLxH).

Now, many creative designs exists. The spiral or maze-like channel machined into a block of copper is very popular. Gary's idea of a sunburst fin pattern with injection of the refrigerant at the edge (multiple injection points) of the block and suction at the center also has potential.

Just one thing bothers me. How to design the channels? We need to pack a lot of surface into a tiny package. Maximum channel cross-section is set by minimum refrigerant speed required.

But what about minimum channel size? There must be a limit on the minimal channel diameter. Take a 1/4" pipe as an example. We can split it in two parallel pipes with their cross-section area each half of the cross-section area of the 1/4" pipe. We can split it into 100 parallel pipes, the cross-section area adding up to the area of the 1/4" pipe. Maybe add a bit more to compensate the added friction.

The 100-pipes example provides much more refrigerant->metal surface area than the 1-pipe example. Refrigerant speed is also equal, assuming 100% perfect distribution. Pressure drop is a bit higher due to added friction.

But where is the limit? Can I make millions of channels, each having a few square microns cross-section area? If not, what limits the minimal channel size?