We thought they were dead, but supercomputers are back in action. They are now benefiting from a growing global market, thanks to the exploitation of 21st century black gold.e century: data at the service of artificial intelligence (AI). They have even become an issue of sovereignty, to the point that the United States, in its trade war against China, is going so far as to deprive its Asian rival of microprocessors intended for these machines. In Europe, the survivors are coming out of the woods. Starting with the French State, which is following the negotiations around the dismantling of the Atos group. In its subsidiary Eviden there is indeed a nugget to be pampered, among the few that survive in Europe: the supercomputers resulting from the takeover of Bull in 2013. A new factory should even be built in Angers by 2027.
These behemoths are called to the rescue to develop and vaccines more quickly, against Covid-19 in particular, refine weather forecasts in the face of climate change, improve the aerodynamics of planes and other vehicles to consume less energy, fight against increasingly formidable cyberattacks or simulating a nuclear explosion in the name of deterrence. Machine learning (machine learning) and quantum computing (simultaneous and mive calculations on the atomic scale) need them.
Long confined to university research or industrial or nuclear simulation (defense), the line of supercomputers is regaining strength with artificial intelligence. “Large-scale AI models are growing very rapidly, and new buyers are starting to use high-capacity machines, with sales prices ranging from tens of millions to hundreds of millions of dollars each. The fight against cyberattacks will also require major computing power”, predicts Earl Joseph, general manager of Hyperion Research, an American research company specializing in the global HPC market (high performance computing, or “high performance computing”). Because the more powerful the computers, the more expensive they are.
Billions of billions of operations
Heavyweights in computing, these new supercomputers have relegated their ancestors from the 1960s to the featherweight category. At the time, the pioneering manufacturers were called IBM, Univac, Control Data Corporation, Cray Research or Silicon Graphics Inc. Moore’s law – predicting the doubling of the power of electronic chips every two years – harmed them in the 1980s-1990s, but there was no extinction of these dinosaurs. A few hundred thousand instructions per second (or KIPS, for kilo instructions per second) for the old ones, the current generation reaches millions of billions of operations per second, expressed in petaflops/s (from 1015) for the so-called models “petascale”, or even billions of billions of operations per second, in exaflops/s (from 1018) for the strongest in calculation, called “exascale”. “If every person on Earth performed one calculation per second, it would take more than four years to do what an exascale computer can do in just one second”indicates the Dane Anders Dam Jensen, director of EuroHPC, the European alliance determined to make the Old Continent a paradise for supercomputers.
You have 72.98% of this article left to read. The rest is reserved for subscribers.