Saturday, November 18, 2023

The Computer That Will Change Everything: "New Breed of Supercomputer Aims for the Two Quintillion Mark"

Back in March we looked at this machine in "The Computer That Will Change Everything".

Here's more from the Wall Street Journal, November 14:

That’s two billion billion operations a second, enough to explore the brain and discover drugs

Inside a vast data center on the outskirts of Chicago, the most powerful supercomputer in the world is coming to life. The machine will be able to analyze connections inside the brain and help design batteries that charge faster and last longer.

Called Aurora, the supercomputer’s high-performance capabilities will be matched with the latest advances in artificial intelligence. Together they will be used by scientists researching cancer, nuclear fusion, vaccines, climate change, encryption, cosmology and other complex sciences and technologies. 

Housed at the Energy Department’s Argonne National Laboratory, Aurora is among a new breed of machines known as “exascale” supercomputers. In a single second, an exascale computer can perform one quintillion operations—a billion billion, or a one followed by 18 zeros. 

Aurora is the size of two tennis courts, weighs 600 tons and is expected to be the world’s first supercomputer capable of two quintillion operations a second at peak performance, scientists at Argonne said.

Aurora, built by

and , is slowly being turned on, rack by rack. Unlike regular computers, these high-powered machines take months to bring online as technicians look for flaws like mechanics testing a Formula One car before a race. Aurora is expected to become fully operational in 2024, Argonne said. 

On Monday, the Top500, a ranking of supercomputers, said tests showed Aurora, only partly functional, was already the second-most-powerful supercomputer in the world. Oak Ridge National Laboratory’s Frontier, which came online last year and was the first operational exascale computer, retained its title as the world’s No. 1 computer. Aurora will likely “exceed Frontier…when finished,” the Top500 said. 

There are more of these powerful machines coming soon. Lawrence Livermore National Laboratory in California is building a $600 million exascale supercomputer called El Capitan, after the famed rock formation in Yosemite National Park. It is expected to be deployed next year and could eventually exceed Aurora’s computational firepower, a spokesman for the lab said. 

is spending more than $1 billion to build an exascale supercomputer called Dojo, its chief executive, Elon Musk, has said. China might have exascale machines, but it doesn’t provide them to outsiders for testing. Computer scientists in the U.K. and elsewhere are trying to produce their own exascale computers.

Behind Aurora’s computing muscle are more than 60,000 graphics processing units, or GPUs, technology developed for advanced videogaming systems that has become the powerhouse of supercomputers. That compares with the nearly 40,000 GPUs in Frontier. 

The added capacity helps provide Aurora with roughly 70% more memory than Frontier, said Olivier Franza, Intel’s chief architect of the project. The extra memory will allow the computer to “tackle problems that no other machine can approach,” he said. Researchers recently used Aurora to screen 22 billion drug molecules an hour, accelerating potential drug discovery. 

The extra memory will also give Aurora the ability to handle the biggest large language model—a predictive AI system similar to ChatGPT—ever deployed, Franza said. Another potential task is mapping connections in the brain, a task so complicated it could take Aurora a full day to process a tiny sliver of the brain. 

The supercomputer will be able to sharply increase the reliability of climate-change forecasts, a notoriously difficult computational challenge. By running thousands of climate scenarios using billions of parameters, forecasters will have higher accuracy in their predictions of impacts on ever-smaller parts of the globe, such as a power plant or a military installation, said Rao Kotamarthi, a senior environmental scientist at Argonne. The more-accurate estimates will allow planners to better prepare for the potential impacts of floods, wildfires or storms on a facility or neighborhood....

....MUCH MORE