Single-Event Effect (SEE) Characterization of the Xilinx UltraScale+ MPSoC and Reliability Analysis of FPGA-accelerated Large-Scale Systems

Future exascale systems will be more powerful but will fail much more frequently. Projections from current large-scale systems indicate that the Mean Time Between Failures (MTBF) is expected to be a few minutes. Thus, preserving MTBF at acceptable levels is a major concern for future systems. Radiation-induced effects are the major factor for the growth of transient errors on HPC systems. High-energy neutrons interact with the silicon die, creating a secondary cascade of charged particles. These can create current pulses that invert the values stored either in memories or produced by combinatorial logic. The proposed test strategy will attempt to study the vulnerability of the hardware technologies integrated on exascale systems. In particular, it will examine an FPGA device that is part of the H2020 EuroEXA’s project node and analyze the reliability of different HPC applications.