Exascale computing refers to computing systems capable of performing at least one exaFLOPS, which is equivalent to a quintillion (10^18) floating-point operations per second. This immense computational power allows for simulations of unprecedented complexity and fidelity, as well as the training of massive AI models. The standard benchmark for measuring this performance is the High Performance LINPACK (HPL) benchmark, using 64-bit precision. Exascale systems represent the pinnacle of supercomputing and are essential for tackling some of the world's most challenging scientific and engineering problems.
The pursuit of exascale computing, often referred to as the 'race to exascale,' began in earnest in the late 2000s and early 2010s, following the achievement of the petascale (10^15 FLOPS) milestone. Governments and research institutions around the world, particularly in the United States, China, Japan, and Europe, invested billions of dollars in research and development to overcome the significant technological hurdles. The culmination of these efforts was the announcement of the first public exascale supercomputer, Frontier, at Oak Ridge National Laboratory in the United States in May 2022.
Exascale computing is having a transformative impact on a wide range of fields. It is being used to create highly detailed climate models to better predict the effects of climate change, to simulate the aging of nuclear stockpiles to ensure their safety and reliability, and to advance astrophysics research by modeling phenomena like black holes and supernovae. In medicine, exascale systems are enabling personalized medicine by simulating how drugs interact with individual patients' cells and by accelerating the discovery of new treatments. The development of exascale computing is also driving innovation in computer architecture, interconnects, and software, with a growing emphasis on energy efficiency to manage the immense power consumption of these systems.