A Survey on Energy-Efficient Microarchitecture Design Strategies for Mobile CPUs and GPUs: Techniques and Future Directions

Abstract

The increasing requirement of high-performance mobile computing has exacerbated the desire of energy-efficient microarchitectures, with modern smartphones being dependent on CPUs and GPUs in order to have the compute-intensive workload. Having battery limitations, thermal and heterogeneous task behavior, the energy consumption optimization and the ability to maintain responsiveness have become critical. In this paper, a review of energy efficient microarchitecture approaches to mobile CPUs and GPUs is conducted covering the major methods that include DVFS, clock gating, power gating, cache optimization, and workload- aware scheduling. The survey also examines new technologies including the GPU-specific technologies such as warp scheduling, task migration, and resource-sharing techniques to tackle the parallel and thermal limits of mobile graphics processors. Also, the analysis identifies new architectural trends including AI-based power management, neuromorphic computing, and reconfigurable hardware that strive to enhance scalability of energy in the next-generation devices. On the whole, the paper summarizes existing discoveries, issues, and design compromises that can assist in creating sustainable and high-efficiency mobile architectures.