FYI.

This story is over 5 years old.

Tech

The US Has the Most Powerful Supercomputer In the World Again

If every person on Earth did one calculation per second, it would take 305 days to do what Summit can do in a second.
Summit. Image: IBM/ORNL

On Friday, Oak Ridge National Laboratory in Tennessee unveiled Summit, a supercomputer capable of 200 petaflops, or 200,000 trillion calculations per second. Summit is over two times more powerful than the previous world-record holder, China’s Sunway TaihuLight, which means the US is home to the most powerful supercomputer in the world again after being dethroned eight years ago. The new supercomputer has been in development at IBM since 2014 and is eight times more powerful than Titan, which until last week was the leading supercomputer in the United States.

Advertisement

To put the power of Summit in perspective, if every person on Earth—that’s all 7.6 billion of us—did one calculation per second, it would take 305 days to do what Summit can do in a second.

Although Oak Ridge National Laboratory has been a leader in developing new supercomputing systems for decades—it was the first in the world to perform teraflop (a trillion operations per second) and petaflop calculations, and is gunning to be the first to reach an exaflop (a quintillion operations per second)—most of the leading supercomputers in the world reside in China.

Yet as supercomputers become more important for cutting-edge scientific research and developing artificial intelligence capabilities, the US is investing more heavily in its supercomputing infrastructure.

Read More: How to Build a Mini-Supercomputer for $100

Today, supercomputers at US national laboratories are used for everything from simulating the conditions around detonated nuclear bombs to modeling the Earth’s climate. Creating these types of models is complex because of the huge number of interacting elements involved. Supercomputers can handle this complexity by leveraging parallel processing architectures, in which thousands of CPUs and GPUs are coordinated to run similar processes at the same time.

Installing Summit at Oak Ridge National Laboratory. Image: ORNL/Flickr

Coordinating all this computing hardware makes the machines themselves physically huge and mind-bogglingly complex. Summit uses 9,216 IBM CPUs and 27,648 Nvidia Tesla GPUs. It requires 185 miles of fiber optic cable, takes up the space of two tennis courts, and pumps 4,000 gallons of water through its cooling system every minute.

Summit is already being used to analyze the data from the Million Veteran Program, which aims to collect medical information from a million veterans to aid in the development of precision medicine. The Oak Ridge National Lab is also accepting proposals for research projects that could benefit from Summit’s unrivaled computing power. Later this month, computer scientists from the Top 500 list who rank the world’s supercomputers will swing by the lab to do some tests at which point Summit will officially be named the fastest supercomputer in the world.