Quantifying Power in Silicon Photonic Neural Networks
Alexander N. Tait
Abstract
Due to challenging efficiency limits facing conventional and unconventional electronic architectures, information processors based on photonics have attracted renewed interest. Research communities have yet to settle on definitive techniques to describe the performance of this class of information processors. Photonic systems are different from electronic ones, and the existing concepts of computer performance measurement cannot necessarily apply. In this paper, we quantify the power use of photonic neural networks with state-of-the-art and future hardware. We derive scaling laws, physical limits, and platform platform performance metrics. We find that overall performance takes on different dominant scaling laws depending on scale, bandwidth, and resolution, which means that energy efficiency characteristics of a photonic processor can be completely described by no less than seven performance metrics over the range of relevant operating domains. The introduction of these analytical strategies provides a much needed foundation and reference for quantitative roadmapping and commercial value assignment for silicon photonic neural networks.