2000 character limit reached
Temperature-Aware Monolithic 3D DNN Accelerators for Biomedical Applications (2203.15874v1)
Published 29 Mar 2022 in cs.ET and cs.AR
Abstract: In this paper, we focus on temperature-aware Monolithic 3D (Mono3D) deep neural network (DNN) inference accelerators for biomedical applications. We develop an optimizer that tunes aspect ratios and footprint of the accelerator under user-defined performance and thermal constraints, and generates near-optimal configurations. Using the proposed Mono3D optimizer, we demonstrate up to 61% improvement in energy efficiency for biomedical applications over a performance-optimized accelerator.