An optimal replacement policy for cooling of 3D stacked mesh array

This paper addresses reconfigurations and the cooling of a 3D stacked mesh array by considering thermo-radiation. We propose a heuristic replacement policy to avoid defects on a wafer by moving hot processing elements (PEs) toward edge of the wafer to cool efficiently. Introducing a thermo-radiation model for the 3D stacked complex, the reconfiguration performance and thermo-radiation of the 3D stacked mesh array are evaluated. Three cooling schemes are proposed and evaluated. The placement of spare PEs, moving hot PEs and heuristic replacement are discussed from the viewpoint of system yield and cooling performance. The heuristic replacement policy uses a repeating reconfiguration procedure, and estimates the cooling efficiency based on the thermo-radiation model for each possible replacement, then the policy adopts the best replacement. Comparing system yields and maximum temperatures by thermo-radiation simulations on a Cray-T3E supercomputer, these cooling approaches maintain the 3D stacked mesh array at a lower temperature while keeping system yield high.