The InnovationQ application will be updated on Sunday, May 31st from 10am-noon ET. You may experience brief service interruptions during that time.
Browse Prior Art Database

Method for Dissipating a Chips Thermal Load

IP.com Disclosure Number: IPCOM000015338D
Original Publication Date: 2002-May-08
Included in the Prior Art Database: 2003-Jun-20

Publishing Venue



Disclosed is a method for dissipating a semiconductor chips thermal load utilizing the refrigeration effect provided by the adiabatic decompression of a gas . As circuit densities and clock speeds increase, integrated circuits are generating problematic amounts of heat which must be removed to prevent the chip from being destroyed. At present, the common method for dealing with this is to place a large metal heat sink on the chip and use sometimes noisy fans to move air past the heatsink to carry the heat away. Thermoelectric devices have been used but they still need a heat sink to carry the exhaust heat away from the system. Other systems have used chilled water (old system 370s) or Freon refrigeration (current zSeries machines). Using heatsinks is cheap but not very efficient which limits the number of CPUs that can be placed in a given area. Chilled water and Freon are very efficient at removing heat but are expensive and cumbersome. In the case of chilled water, a leak in the plumbing gets very messy and has the potential to be destructive to the electronics. The limitations on CPU density that these thermal problems have created are a problem for those who need to use large numbers of CPUs in a small area. (small being relative). This would include ISPs, and companies running server farms for web hosting, application hosting, etc. This is also an issue for customers running large scientific/engineering/life sciences clusters used for solving massively complex problems. This disclosure documents the use of decompressing gas, preferably air, to absorb the thermal load from the chip and carry the waste heat away. While most refrigeration methods use decompression plus a phase change in the refrigerant to move heat (because of it greater efficiency) decompression alone works and avoids the plumbing nightmare of trying to use something like Freon to chill hundreds or thousands of CPUs. If this system leaks a little bit, no one cares because there are no flammable, toxic, or costly materials being lost. This avoids the bulkiness of present day heatsinks using an ambient air stream to absorb a chips heat load and the cost and complexity of other solutions. Because of it's efficiency and compactness it also allows very high CPU densities. A useful embodiment of this disclosure would employ a small heatsink that looked like a solid block of aluminum. It would have ports for exhausting air and a fixture that would accept a high pressure air line. It would be thermally bonded to the chip being cooled. The heatsink would have no external fins but internally would have a series of passageways for compressed air to pass through. The aerodynamic design of the internal passageways would allow the compressed air to enter and gradually decompress as it passed through those passageways. As it decompresses, it would chill allowing it to absorb substantial amounts of heat along the way. When the air stream is decompressed to atmospheric pressure it would be exhausted into the interior of the server or could be ducted out of the server enclosure, as appropriate. The system could be tuned such that the exhausting air stream could be cold enough to aid in the cooling of other components in the server. 1