Early Design Phase Power Performance Trade-Offs Using In-Situ Macro Models

Early phase space exploration with a focus on power performance tradeoffs has been shown to enable faster design convergence. Tightly coupled physical design considerations and system level models are needed to guarantee time-to- market. This paper describes a methodology for early design space exploration. A Pareto-front analysis using the proposed methodology was applied to an industrial design to perform design space exploration. The obtained low power design solution improved the designers best solution by additionally reducing power by 6% with 2% performance impact. The optimal trade-off solution improved designers best by additionally trading off 1.7% and 2.8% performance for power savings of 8.5% and 7.6% respectively. The absolute amount achieved here is not as important as knowing what the desired design options should be for each functional block in the design very early in the design phase.