UT "Stampedes" Ahead With New Supercomputer

The University of Texas is building a supercomputer that is one million times faster than a standard laptop and has 70,000 times the memory.

UT's newest supercomputer, named Stampede, is expected to support more than 1,000 projects in computational and data-driven science and engineering from across the country, putting Texas and the university on the cutting edge of scientific discovery.

“This is really about the United State’s competitiveness in science and engineering,” said Irene Qualters, program manager at the National Science Foundation, which is funding the project. “In order for us to stay in the lead, we need access to the forefront of tools and the services to support these tools.”

The Texas Advanced Computing Center (TACC) at UT announced it is constructing and supporting Stampede in partnership with Dell and Intel through an award from the NSF. The powerful new computer is expected to accelerate the pace of scientific breakthroughs, said John Mullen, Dell vice president and general manager of major public accounts, education, state and local government. 

The system will help scientists across the country address major problems in subjects ranging from designing more effective drugs to climate modeling to energy research.

“This will be one of the most powerful systems in the world,” said Jay Boisseau, director of TACC.

With a peak performance of 10 petaflops — a measure of a computer's processing speed — Stampede is more than 30 times faster than Lonestar 4, the current supercomputer at UT that was deployed in April.

The NSF gave UT researchers $27.5 million for the supercomputer upfront and it's expected to invest a total of $50 million over four years. Stampede will be up and running in January 2013, Boisseau said, and the project could be renewed for another four years in 2017.

Researchers from any open science institution in the country can apply to use Stampede, which will be housed at UT’s J.J. Pickle Research Campus, through the NSF’s Extreme Science and Engineering Discovery Environment (XSEDE) project. A national committee will review the proposals and allocate time on the supercomputer.

Boisseau said 90 percent of the time on the machine would be allocated directly to the national open science community, and the other 10 percent would be used for a mix of industrial partnership projects and new community projects, like urban planning and social sciences. Many of the new community projects will be at UT.

Professor Omar Ghattas, director of the Center for Computational Geosciences in the Institute for Computational Engineering and Sciences at UT, said Stampede would benefit scientists in several ways. First, they will be able to put more resolutions into their models. “The more pieces you can break the model up into, the better you can actually solve the equations,” Ghattas said.

They will also be able to put more physics data into their models to help solve problems. “And because of Stampede’s ground-breaking speed, scientists can also benefit by running many more calculations in a shorter period of time,” Ghattas said.

Historically, supercomputers have been used for conducting simulations of mathematical models. But because of Stampede’s large memory and disk storage, it will support both simulation-based and data-driven science, Boisseau said.

Stampede is not without its challenges, however.

“When we build these systems bigger and bigger, the odds of having a failure are more likely," Bouisseau said. "So we have people develop codes that checkpoint, just in case the systems have to be restarted."

Stampede is also going to need a lot of power, and UT is working to expand its data center to accommodate the energy consumption.