New 200-petaflop supercomputer to succeed Titan at ORNL

 

A new 200-petaflop supercomputer will succeed Titan at Oak Ridge National Laboratory, and it could be available to scientists and researchers in 2018, a spokesperson said this week.

The new IBM supercomputer, named Summit, could about double the computing power of what is now the world’s fastest machine, a Chinese system named Sunway TaihuLight, according to a seminannual list of the world’s top supercomputers released in June.

Sunway TaihuLight is capable of 93 petaflops, according to the list, the TOP500 list. A petaflop is one quadrillion calculations per second. That’s 1,000 trillion calculations per second.

Summit, which is expected to start operating at ORNL early in 2018, is one of three supercomputers that the U.S. Department of Energy expects to exceed 100 petaflops at three U.S. Department of Energy laboratories in 2018. The three planned systems are:

• the 200-petaflop Summit at ORNL, which is expected to be available to users in early 2018;
• a 150-petaflop machine known as Sierra at Lawrence Livermore National Laboratory near San Francisco in mid-2018; and
• a 180-petaflop supercomputer called Aurora at Argonne National Laboratory in Chicago in late 2018.

“High performance computing remains an integral priority for the Department of Energy,” DOE Under Secretary Lynn Orr said. “Since 1993, our national supercomputing capabilities have grown exponentially by a factor of 300,000 to produce today’s machines like Titan at Oak Ridge National Lab. DOE has continually supported many of the world’s fastest, most powerful super-computers, and shared its facilities with universities and businesses ranging from auto manufacturers to pharmaceutical companies, enabling unimaginable economic benefits and leaps in science and technology, including the development of new materials for batteries and near zero-friction lubricants.”

The supercomputers have also allowed the United States to maintain a safe, secure, and effective nuclear weapon stockpile, said Orr, DOE under secretary for science and energy.

“DOE continues to lead in software and real world applications important to both science and industry,” he said. “Investments such as these continue to play a crucial role in U.S. economic competitiveness, scientific discovery, and national security.”

At 200 petaflops, Summit would have at least five times as much power as ORNL’s 27-petaflop Titan. That system was the world’s fastest in November 2012 and recently achieved 17.59 petaflops on a test used by the TOP500 list that was released in June.

Titan is used for research in areas such as materials research, nuclear energy, combustion, and climate science.

“For several years, Titan has been the most scientifically productive in the world, allowing academic, government, and industry partners to do remarkable research in a variety of scientific fields,” ORNL spokesperson Morgan McCorkle said.

Summit will be installed in a building close to Titan. Titan will continue operating while Summit is built and begins operating, McCorkle said.

“That will ensure that scientific users have access to computing resources during the transition,” she said.

Titan will then be decommissioned, McCorkle said.

She said the total contract value for the new Summit supercomputer with all options and maintenance is $280 million. The U.S. Department of Energy is funding the project.

McCorkle said the Oak Ridge Leadership Computing Facility at ORNL has been working with IBM, Nvidia, and Mellanox since 2014 to develop Summit.

Like Titan, a Cray system, Summit will be part of the Oak Ridge Leadership Computing Facility, or OLCF. Researchers from around the world will be able to submit proposals to use the computer for a wide range of scientific applications, McCorkle said.

She said the delivery of Summit will start at ORNL next year. Summit will be a hybrid computing system that uses traditional central processing units, or CPUs, and graphic processing units, or GPUs, which were first created for computer games.

“We’re already scaling applications that will allow Summit to deliver an order of magnitude more science with at least 200 petaflops of compute power,” McCorkle said. “Early in 2018, users from around the world will have access to this resource.”

Summit will have more than five times the computational power of Titan’s 18,688 nodes, using only about 3,400 nodes. Each Summit node will have IBM POWER9 CPUs and NVIDIA Volta GPUs connected with NVIDIA’s high-speed NVLinks and a huge amount of memory, according to the OLCF.

Titan is also a hybrid system that combines CPUs with GPUs. That combination allowed the more powerful Titan to fit into the same space as Jaguar, an earlier supercomputer at ORNL, while using only slightly more electricity. That’s important because supercomputers can consume megawatts of power.

China now has the top two supercomputers. Sunway TaihuLight was capable of 93 petaflops, and Tianhe-2, an Intel-based system ranked number two in the world, achieved 33.86 petaflops, according to the June version of the TOP500 list.

But as planned, all three of the new DOE supercomputers would be more powerful than the top two Chinese systems.

However, DOE officials said it’s not just about the hardware.

“The strength of the U.S. program lies not just in hardware capability, but also in the ability to develop software that harnesses high-performance computing for real-world scientific and industrial applications,” DOE said. “American scientists have used DOE supercomputing capability to improve the performance of solar cells, to design new materials for batteries, to model the melting of ice sheets, to help optimize land use for biofuel crops, to model supernova explosions, to develop a near zero-fiction lubricant, and to improve laser radiation treatments for cancer, among countless other applications.

Extensive work is already under way to prepare software and “real-world applications” to ensure that the new machines bring an immediate benefit to American science and industry, DOE said.

“Investments such as these continue to play a crucial role in U.S. economic competitiveness, scientific discovery, and national security,” the department said.

DOE said its supercomputers have more than 8,000 active users each year from universities, national laboratories, and industry.

Among the supercomputer uses that DOE cited:

• Pratt and Whitney used the Argonne Leadership Computing Facility to improve the fuel efficiency of its Pure Power turbine engines.
• Boeing used the Oak Ridge Leadership Computing Facility to study the flow of debris to improve the safety of a thrust reverser for its new 787 Dreamliner.
• General Motors used the Oak Ridge Leadership Computing Facility to accelerate research on thermoelectric materials to help increase vehicle fuel efficiency.
• Proctor and Gamble used the Argonne Leadership Computing Facility to learn more about the molecular mechanisms of bubbles—important to the design of a wide range of consumer products.
• General Electric used the Oak Ridge Leadership Computing Facility to improve the efficiency of its world-leading turbines for electricity-generation.
• Navistar, NASA, the U.S. Air Force, and other industry leaders collaborated with scientists from Lawrence Livermore National Lab to develop technologies that increase semi-truck fuel efficiency by 17 percent.

Though it was once the top supercomputer, Titan was bumped to number two behind Tianhe-2 in June 2013. It dropped to number three this June.

As big as a basketball court, Titan is 10 times faster than Jaguar, the computer system it replaced. Jaguar, which was capable of about 2.5 petaflops, had ranked as the world’s fastest computer in November 2009 and June 2010.

The new top supercomputer, Sunway TaihuLight, was developed by the National Research Center of Parallel Computer Engineering and Technology, or NRCPC, and installed at the National Supercomputing Center in Wuxi, China.

Tianhe-2 was developed by China’s National University of Defense Technology.

In the United States, DOE said its Office of Science and National Nuclear Security Administration are collaborating with other U.S. agencies, industry, and academia to pursue the goals of what is known as the National Strategic Computing Initiative:

• accelerating the delivery of “exascale” computing;
• increasing the coherence between the technology base used for modeling and simulation and that for data analytic computing;
• charting a path forward to a post-Moore’s Law era; and
• building the overall capacity and capability of an enduring national high-performance computing ecosystem.

See previous stories on Titan here.

See previous supercomputer stories here.

Read more about Summit here: https://www.olcf.ornl.gov/summit/.

See a Summit fact sheet here: https://www.olcf.ornl.gov/wp-content/uploads/2014/11/Summit_FactSheet.pdf.

More information will be added as it becomes available.

Copyright 2016 Oak Ridge Today. All rights reserved. This material may not be published, broadcast, rewritten, or redistributed.