Towards Energy Efficient Parallel Computing on Consumer Electronic Devices

Abstract

In the last two decades supercomputers have sustained a remarkable growth in performance that even out-performed the predictions of Moore’s law, primarily due to increased levels of parallelism [19]. As industry and academia try to come up with viable approaches for exascale systems, attention turns to energy efficiency as the primary design consideration. At the same time, energy efficiency has always been the main concern in the mobile computing area. Additionally, mobile and consumer electronic devices are becoming ever more powerful as the use cases (e.g., Web 2.0 applications, video encoding, virtual and augmented reality) become more computationally demanding. It is therefore an interesting question to ask if these devices are the possible building blocks of future HPC systems. It was the workstation and server market in the past that provided the CPUs that power supercomputers and it might be the consumer electronic market that provides the underlying technology in the future.

In this paper we try to analyze the current state of energy efficient parallel and distributed computing on mobile and consumer electronic devices. We provide an overview of performance characteristics of some current and announced future devices for scientific computation and we build a small proof-of-concept cluster from Apple’s second generation “Apple TV” devices and evaluate its performance on standard benchmark applications. We discuss the limiting factors, and analyze the industry trajectory that we believe could make consumer electronic-based design a feasible technology basis for future HPC system designs.