Evaluation of Xilinx Deep Learning Processing Unit under Neutron Irradiation

TL;DR

This study evaluates how neutron irradiation affects the reliability of Xilinx's DPU when executing deep learning models, highlighting the impact of radiation-induced errors on system accuracy.

Contribution

It provides the first analysis of neutron radiation effects on Xilinx DPU performance during deep learning inference.

Findings

Neutron irradiation causes errors in DPU computations.

Single Event Effects impact the accuracy of the resnet50 model.

The paper quantifies the error rates under neutron exposure.

Abstract

This paper studies the dependability of the Xilinx Deep-Learning Processing Unit (DPU) under neutron irradiation. It analyses the impact of Single Event Effects (SEEs) on the accuracy of the DPU running the resnet50 model on a Xilinx Ultrascale+ MPSoC.