Abstract:The AMD UltraScale+ XCZU9EG device is a Multi-Processor System-on-Chip (MPSoC) with embedded Programmable Logic (PL) that excels in many Edge (e.g., automotive or avionics) and Cloud (e.g., data centres) terrestrial applications. However, it incorporates a large amount of SRAM cells, making the device vulnerable to Neutron-induced Single Event Upsets (NSEUs) or otherwise soft errors. Semiconductor vendors incorporate soft error mitigation mechanisms to recover memory upsets (i.e., faults) before they propagate to the application output and become an error. But how effective are the MPSoC's mitigation schemes? Can they effectively recover upsets in high altitude or large scale applications under different workloads? This article answers the above research questions through a solid study that entails accelerated neutron radiation testing and dependability analysis. We test the device on a broad range of workloads, like multi-threaded software used for pose estimation and weather prediction or a software/hardware (SW/HW) co-design image classification application running on the AMD Deep Learning Processing Unit (DPU). Assuming a one-node MPSoC system in New York City (NYC) at 40k feet, all tested software applications achieve a Mean Time To Failure (MTTF) greater than 148 months, which shows that upsets are effectively recovered in the processing system of the MPSoC. However, the SW/HW co-design (i.e., DPU) in the same one-node system at 40k feet has an MTTF = 4 months due to the high failure rate of its PL accelerator, which emphasises that some MPSoC workloads may require additional NSEU mitigation schemes. Nevertheless, we show that the MTTF of the DPU can increase to 87 months without any overhead if one disregards the failure rate of tolerable errors since they do not affect the correctness of the classification output.

Neutron-induced single event effect in Xilinx 16nm MPSoC configuration RAM (CRAM) using white neutron and 2.72 similar to 81.8meV neutron in CSNS-BL20

Neutron Irradiation Testing and Monte Carlo Simulation of a Xilinx Zynq-7000 System on Chip

Analysis of multiple cell upset sensitivity in bulk CMOS SRAM after neutron irradiation

SEU Effects Induced by Reactor Spallation and Accelerator Neutron Sources in Nano-SRAMs

Single-Event Upsets in SRAMs with Scaling Technology Nodes Induced by Terrestrial, Nuclear Reactor, and Monoenergetic Neutrons

Single Event Upset on Static Random Access Memory Devices Due to Spallation, Reactor, and Monoenergetic Neutrons

Neutron induced single event upset in CMOS simulated with Geant4

Microbeam Heavy-Ion Single-Event Effect on Xilinx 28-nm System on Chip

Experimental Study of the Impact of Temperature on Atmospheric Neutron-Induced Single Event Upsets in 28 nm Embedded SRAM of SiP

Electron-Induced Single-Event Effect in 28 nm SRAM-Based FPGA

SEE Sensitivity Evaluation for Commercial 16 nm SRAM-FPGA

Identification of Pulsed-Neutron-Induced Upset Bursts in Static Random Access Memories using Monte-Carlo Simulations

Comparative Analysis of Upset-Multiplicity Occurrences Due to Flash X-rays and Pulsed Neutrons in Commercial SRAMs

Neutron-induced effects on a self-refresh DRAM

Single Event Effects Assessment of UltraScale+ MPSoC Systems under Atmospheric Radiation

SEU in SRAMs Due to 2.5 MeV, 14 MeV and Reactor Neutrons

Single Event Upsets characterization of 65 nm CMOS 6T and 8T SRAM cells for ground level environment

Simulation and Experiment in Neutron Induced Single Event Effects in SRAM

Single event effects evaluation on Convolution Neural Network in Xilinx 28nm System on Chip

X-Ray and Proton Radiation Effects on 40 Nm CMOS Physically Unclonable Function Devices

J un 2 01 9 Ultraslow Dynamic Annealing of Neutron-induced Defects in n-type Silicon