NBTI and Power Reduction Using a Workload-Aware Supply Voltage Assignment Approach

Abstract

Supply voltage assignment (SVA) can alleviate the performance aging induced by the negative bias temperature instability (NBTI) effect. However, due to the random characteristic of an actual system workload, it is difficult to estimate the aging rate and control the supply voltage reasonably. To solve this problem, we present a workload-aware SVA method (WSVA) that encapsulates the workload change into the aging estimation using an LUT-based approach. Moreover, an NBTI and leakage co-optimization strategy based on an integer linear programming (ILP) approach is proposed to obtain the optimal input vector in standby mode. Simulation experiments on multiple benchmark circuits demonstrate that the LUT-based approach can track the dynamic change of the workload online and provide an accurate aging estimate for SVA with little computation cost. Compared with the SVA method without considering the workload, the proposed aging estimation approach and the optimal input vector selection strategy in the WSVA framework can enable the CMOS circuit conserve additional power dissipation while guaranteeing the performance requirements.