In embedded systems, performance and power are important inter-related issues that cannot be decoupled. Expensive and extensive simulations in a processor design space are usually required to verify whether a design meets both performance and power requirements. In this paper, an analytical co-optimization approach based on an integrated workload, performance and power model for modern processors is described and studied. A design space consisting of more than 15 architectural and workload parameters can be quickly explored for co-optimization. Validation with measured results obtained from simulators as well as physical processors showed that the model has a good degree of accuracy. We shall describe the details of approach and the model, and show how to apply the approach to the problem of co-optimizing the power and performance of processor design. With the completeness, flexibility and efficiency, our approach provides clear insights into the tradeoffs of designs for performance...