Top NSF petascale supercomputer and expert staff accelerate discoveries for nation's scientists

Published on May 27, 2014 by Aaron Dubrow

Innovation: Integrating New Technologies

When the National Science Foundation released their solicitation for proposals for a new supercomputer to be deployed in 2013, they were looking for a system that could support the day-to-day needs of a growing community of computational scientists, but also one that would push the field forward by incorporating new, emerging technologies.

"The model that TACC used, incorporating an experimental component embedded in a state-of-the-art usable system, is a very innovative choice and just right for the NSF community of researchers who are focused on both today's and tomorrow's scientific discoveries," said Irene Qualters, division director for Advanced Cyberinfrastructure at the National Science Foundation.

"The results that researchers have achieved in Stampede's first year are a testimony to the system design and its appropriateness for the community."

The Intel Xeon Phi [aka the ‘many integrated core (MIC) coprocessor] squeezes 60 or more processors onto a single card. In that respect, it is similar to GPUs (graphics processing units), which have been used for several years to aid parallel processing in high-performance computing systems, as well as to speed up graphics and gaming capabilities in home computers. The advantage of the Xeon Phi is its ability to perform calculations quickly while consuming less energy.

"We wanted to put an innovative twist on our system and look at the next generation of capabilities," said TACC's Dan Stanzione. "What we came up with is a hybrid system that includes traditional Intel Xeon E5 processors and also has an Intel Xeon Phi card on every node on the system, and a few of them with two.

"The Xeon Phi is Intel's approach to changing these power and performance curves by giving us simpler cores with a simpler architecture but a lot more of them in the same size package," Stanzione said.

As advanced computing systems grow more powerful, they also consume more energy — a situation that can be addressed by simpler, multicore chips. The Xeon Phi and other comparable technologies are believed to be critical to the effort to advance the field and develop future large-scale supercomputers.

"The exciting part is that MIC and GPU foreshadow what will be on the CPU in the future," Stanzione said. "The work that scientists are putting in now to optimize codes for these processors will pay off. It's not whether you should adopt them; it's whether you want to get a jump on the future."

Though Xeon Phi adoption on Stampede started slowly, it now represents 10-20 percent of the usage of the system. Among the projects that have taken advantage of the Xeon Phi co-processor are efforts to develop new flu vaccines, simulations of the nucleus of the atom relevant to particle physics, and a growing amount of weather forecasting.

Back to overview >>