Work

Radiation Reliability of System Reboots in Commercial Off-the-Shelf SoC

Year: 2025

Type: article

Abstract: This study investigates the reliability of system reboots in modern low-power system-on-chip (SoC) devices when exposed to terrestrial neutron radiation. The focus is on understanding how single-event functional interrupts (SEFIs) caused by neutron strikes affect system downtime during reboot processes. Three commercial off-the-shelf (COTS) SoCs from NVIDIA, namely, the Jetson Nano with Tegra X1 SoC and two models of Jetson Orin Nano with 4- and 8-GB memory, were subjected to high-flux neutron radiation. The research assesses the number and duration of reboots caused by SEFIs and their consequent impact on overall system downtime. Results reveal that the reboot process has a significantly higher cross section compared to application-level operations such as matrix multiplication and neural network inference, which highlights its vulnerability to radiation. The study further categorizes the reboots based on their triggers and identifies common system messages linked to radiation-induced errors. It also shows that different phases of the boot flow have different radiation reliabilities. The findings underscore the importance of evaluating boot processes in radiation-intensive environments, especially for safety-critical applications where minimizing downtime is crucial.

Source: IEEE Transactions on Nuclear Science

DOI: 10.1109/tns.2025.3539942