IBM uses glue to cool chips

Willy Higinbotham · Post by **Willy Higinbotham** » Fri Mar 23, 2007 8:56 am

The technique, developed by a team of scientists at the IBM Zurich Research Laboratory in cooperation with Momentive Performance Materials, formerly GE Advanced Materials, overcomes a barrier in chip cooling by improving the application of the "glue" that binds chips to their cooling systems. The new technology could allow faster computer chips to be cooled more efficiently

http://www.zurich.ibm.com/news/07/cooling.html

Source: IBM Zurich

MC FLMJIG · Post by **MC FLMJIG** » Fri Mar 23, 2007 9:23 am

how crazy is that!

J. Sparrow · Post by **J. Sparrow** » Sat Mar 24, 2007 6:28 am

scientists noticed that a cross formed in the paste as large amounts of particles piled up, inhibiting the layers of glue from spreading out evenly. The scientists were able to trace the cause of this back to the flow behavior of the paste, which simply follows the path of least resistance. Along the diagonals, the particles are pulled in opposite directions and, as a result, they remain where they are. As the squeezing process continues, however, the particles begin to pile up, forming what scientists call a "magic cross".

Thus if you just spread your paste instead of squeezing it, no magic crosses and no need for fancy-branchy IBM solutions ?

Elixer · Post by **Elixer** » Sat Mar 24, 2007 9:55 am

While I salute IBM's attempt to improve cooling in this manner, I would say that the thermal interface is not the major problem, at least with custom system builders like us. This is shown by the fact that different thermal pastes yield the exact same results. However for your typical OEM install with the crappy thermal wax pad with a stock heatsink, I can understand how this would improve cooling greatly if less wax was used. This is also probably good news for extreme overclockers though as the thermal interface become s more important with better cooling.

Gojira-X · Post by **Gojira-X** » Sat Mar 24, 2007 9:59 am

IBM has taken the notion of minimal TIM application to the extreme with the diagonal crosses. They have figured that having diagonal crosses in the middle of the heat sink will increase the rate of heat transfer by creating Hi and low spots the TIM.
These combined with the slightly increases surface area will (according to IBM) speed up heat transfer in contrast to the accepted practice of lapping (i.e. polishing the surface to a mirror like shine).

I think the method suggested will not really make any head roads in the OEM and consumer heat sink markets, but will be advantageous in the packaging of CPU's (between the CPU die and the heat spreader) and the space conscious Embedded computing market.