Microchannel Cooling in Computing Platforms: Performance Needs and Challenges in Implementation

Abstract

Remote cooling is the established cooling scheme in notebook computers, and increasingly, other computing sectors like desktops and servers are evaluating this approach as an option for cooling future platforms. While remote cooling facilitates a larger heat exchanger than the space directly over the processor would allow, it introduces an additional thermal resistance, in particular, θp-f (plate to fluid resistance) — the resistance in getting the heat from the cold plate to the fluid. For any remote cooling system, this resistance needs to be carefully evaluated and minimized. Pumped fluid loops incorporating microchannel heat exchangers are a viable option to achieve low plate-to-fluid resistances. In this paper we will identify a reasonable target for θp-f and subsequently describe two similar but fundamentally different thermal systems to accomplish this target performance: single-phase and two-phase pumped loops. Although two phase flows are traditionally thought of as the way to accomplish the highest heat transfer coefficients and thus the lowest resistances, with microchannel heat sinks the contrast is not so acute. We will present results from our experimental work on single- and two-phase heat transfer from microchannel heat sinks and demonstrate a transition where single-phase performance matches that of two-phase operation. This will be followed by the analysis methods used to predict the heat transfer and the pressure drop data. Moreover, we will discuss system level issues and other hurdles that need to be overcome in commercialization of microchannel technology for cooling computer systems.