Temperature-Aware Monolithic 3D DNN Accelerators for Biomedical Applications

Prachi Shukla; Vasilis F. Pavlidis; Emre Salman; Ayse K. Coskun

Temperature-Aware Monolithic 3D DNN Accelerators for Biomedical Applications

Emerging Technologies 2022-03-31 v1 Hardware Architecture

Authors: Prachi Shukla , Vasilis F. Pavlidis , Emre Salman , Ayse K. Coskun

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.

Keywords

deep neural network acceleration

Cite

@article{arxiv.2203.15874,
  title  = {Temperature-Aware Monolithic 3D DNN Accelerators for Biomedical Applications},
  author = {Prachi Shukla and Vasilis F. Pavlidis and Emre Salman and Ayse K. Coskun},
  journal= {arXiv preprint arXiv:2203.15874},
  year   = {2022}
}

Comments

This paper was accepted to be presented at the Design, Automation and Test in Europe Conference (DATE) 2022 workshop on "3D Integration: Heterogeneous 3D Architectures and Sensors"

Temperature-Aware Monolithic 3D DNN Accelerators for Biomedical Applications

Abstract

Keywords

Cite

Comments

Related papers