Listen to the audio version of the article
«Exascale computing» sounds like a science fiction definition, and in many ways it actually is. But there is a “formula” that describes its contours in a simple and concrete way: if the human brain can perform on average one simple mathematical operation per second, an exascale computer is capable of processing something like a trillion, that is a billion billion times more. We are talking about the new frontier of super computers, the one that takes floating point calculation capabilities beyond the exaflop limit.
This type of computer promises to support the highest workloads required by convergent modeling and the most advanced simulation models, artificial intelligence applications and the analysis of huge amounts of data.
At the core of exascale machines is a combination of different hardware technologies (tens of thousands of CPUs and GPUs, multi-socket nodes and other silicon processing devices), brought together in a single integrated application management and development infrastructure. Perhaps the most important peculiarity of these systems is that of moving data quickly and without slowdowns between the processors (the central and graphics units) and the storage systems, guaranteeing performances comparable to those of quantum machines, which instead exploit the simultaneity of the superposition and the interdependence of binary coding to reduce processing times with a significantly lower energy impact.
The value of exascale computers, of these cabinets stuffed with cables aligned next to each other inside large buildings, is however undoubted and finds expression in various fields, from scientific research (the study of the origins of chemical elements or physics of atomic particles, for example) to meteorology, from security to healthcare, from medicine (think of the analyzes on new drugs carried out in the Covid period) to astronomy. Inside these big brains, interactions are managed that power billions and billions of combinations, elaborated by as many compiled mathematical equations: only an exascale class computer can solve problems and manage calculations with higher levels of complexity (and considered unsolvable) and model human knowledge to improve decision-making processes in any area.
It is therefore no coincidence that the challenge to excel in this field is played out on a global level and involves companies and government agencies from all over the planet, obviously including China. The first exascale computer revealed in the world flies the American flag and was born in May 2022, when the Frontier supercomputer at the Oak Ridge National Laboratory of the United States Department of Energy went into operation with a performance of just under 1.2 hexaflops and earning the top of the TOP50 ranking among the fastest computing systems in the world. Other “exa-machines” with even greater computational capacity have long been under progressive development to support Frontier, starting with “El Capitan”, hosted at the Lawrence Livermore National Laboratory in California and designed to reach a computing power of two exaflops . A goal, moreover, which is also aimed at in its complete configuration by another stars-and-stripes super computer, Aurora, in operation at the Argonne National Laboratory in Illinois.