A NUMERICAL STUDY ON MULTI-OBJECTIVE DESIGN OPTIMIZATION OF HEAT SINKS FOR FORCED AND NATURAL CONVECTION COOLING OF IMMERSION-COOLED SERVERS
Satyam Saini, Tushar Wagh, Pratik Bansode, Pardeep Shahi, Joseph B. Herring, Jacob Lamotte-Dawaghreh, Jimil M. Shah, Dereje Agonafer
Abstract
Rapidly rising computational, storage and networking demands have brought about an increment in both the number and energy density of modern data centers. Typical air-cooled high-performance servers require low air-supply temperatures as well as higher air flowrates making air-cooling inefficient above certain thermal design power values. Single-phase immersion cooling solves this issue by offering significantly higher thermal mass and known reliability enhancements. When choosing to implement immersion cooling for server hardware that is designed for air-cooling, immersion-specific optimized heat sinks should be used. This study presents an in-depth investigation of multi-objective and multi-design variable optimization of heat sinks for immersion-cooled servers for a fixed pumping power using CFD. The optimization is conducted for both copper and aluminum heat sinks by using pressure drop and thermal resistance minimization as objective functions. Differences in the optimized values of the heat sink geometry were compared and quantified for natural as well as forced convection cooling. The results show that the total effect values of heat sink geometric parameters vary significantly between forced and natural convection optimized heat sinks.