exaFLOPclass

exaFLOPclass is a term used in high-performance computing to denote computer systems capable of delivering at least one exaFLOP per second (10^18 floating-point operations per second) of computing power. It characterizes the top tier of supercomputers that approach or attain exascale performance.

In practice, the category is defined by sustained performance on benchmarks rather than theoretical peak. The High-Performance Linpack (HPL) benchmark is the standard reference for exascale classification, though many projects also consider performance on representative workloads such as HPCG or domain-specific simulations. Power efficiency and scalability are critical factors, since reaching exaFLOPS requires thousands of compute nodes and high-performance interconnects.

Frontier, installed at Oak Ridge National Laboratory, is commonly cited as the first exaFLOP-class system, achieving sustained exascale performance on HPL and other workloads. Other announced exaFLOP-class systems include Aurora and El Capitan, with timelines and configurations evolving as hardware and software mature.

Key design challenges include energy consumption, cooling, software portability, fault tolerance, and the need for advanced accelerators and high-bandwidth memory and interconnects to sustain exascale workloads. The term is used in policy, research, and media discussions about the future of computing, as exaFLOP-class machines enable advances in climate modeling, materials science, genomics, and artificial intelligence.