Thermal placement optimization of MCMs using a force-directed based algorithm

The minimum system temperature design of MCMs containing a number of chips of equal power for design in natural convection is pursued through the optimal thermal placement of chips. For dealing with the thermal optimization problems, a simple but effective thermal design methodology that integrates a modified force-directed (FD) algorithm and a finite element (FE) modeling technique is proposed. In the FD algorithm, novel force functions that include attractive and repulsive forces are proposed, which are built up, as a function of location and power of chips, based on the modeled actual temperature information through FE analysis, a response surface method (RSM) and an interpolation scheme. The sum of these forces within the system, approximating the total system chip junction temperature, is considered as the objective of the optimization problem, and non-overlapping conditions between chips are imposed as geometry constraints. The optimum locations of these chips are sought through an iterative search of the entire design space based on two proposed searching strategies. To demonstrate the effectiveness of the current design optimization methodology, several design case studies involving direct chip attached (DCA) typed MCMs containing a number of chips of equal power are performed. The obtained results are extensively compared against the published data.