Power Estimation for Interactive 3D Game Using an Efficient Hierarchical-Based Frame Workload Prediction

In this paper we propose a novel hierarchical-based workload prediction algorithm that integrates the proportion control, integral control, derivative control, frame structure, and analysis of a game or benchmark to make game workload predictions and perform power estimation. By hierarchical analysis of a game or benchmark, one can know the complexity of a future frame of interactive 3D games in advance. Hence, the prediction error is small. Experimental results show that the proposed algorithm requires only 41.2 ms of extra workload, but provides an improvement of more than 15.7% in cycle count estimation as compared to competing algorithms.