UT is Now Home to the Fastest Supercomputer at Any U.S. University

By Anna Daugherty in 40 Acres, Nov | Dec 2017 on October 27, 2017 at 12:44 pm | No Comments

The term “medical research” might bring to mind a sterile room with white lab coats, goggles, and vials. But for cutting-edge researchers, that picture is much more high-tech: it’s a room filled with row after row of metal racks housing 300,000 computer processors, each blinking green, wires connecting each processor, and the deafening sound of a powerful machine at work. It’s a room like the one housing the 4,000-square-foot supercomputer Stampede2 at The University of Texas’ J.J. Pickle Research Campus.

At peak performance, Stampede2, the flagship supercomputer at UT Austin’s Texas Advanced Computing Center (TACC), will be capable of performing 18 quadrillion operations per second (18 petaflops, in supercomputer lingo). That’s more powerful than 100,000 desktops. As the fastest supercomputer at any university in the U.S., it’s a level of computing that the average citizen can’t comprehend.  Most people do their computing on phones the size of their hands—but then again, most aren’t mining cancer data, predicting earthquakes, or analyzing black holes.

Funded by a $30 million grant from the National Science Foundation, Stampede2 replaces the original Stampede system, which went live in 2013. Designed to be twice as powerful while using half the energy of the older system, Stampede2 is already being used by researchers around the country. In June 2017, Stampede2 went public with 12 petaflops and was ranked as the 12th most powerful computer in the world. Phase two added six petaflops in September and phase three will complete the system in 2018 by adding a new type of memory capacity to the computer.

For researchers like Rommie Amaro, professor of chemistry at the University of California, San Diego, a tool like Stampede2 is essential. As the director of the National Biomedical Computation Resource, Amaro says nearly all of their drug research is done on supercomputers.

Most of her work with the original Stampede system focused on a protein called p53, which prevents tumor growth; the protein is mutated in approximately half of all cancer patients. Due to the nature of p53, it’s difficult to track with standard imaging tools, so Amaro’s team took available images of the protein to supercomputers and turned them into a simulation showing how the 1.6 million atoms in p53 move. Using Stampede, they were able to find weaknesses in p53 and simulate interactions with more than a million compounds; several hundred seemed capable of restoring p53. More than 30 proved successful in labs and are now being tested by a pharmaceutical company.

“The first Stampede gave us really outstanding, breakthrough research for cancer,” Amaro says. “And we already have some really interesting preliminary data on what Stampede2 is going to give us.”

And it’s not just the medical field that benefits. Stampede has created weather phenomena models that have shown new ways to measure tornado strength, and produced seismic hazard maps that predict the likelihood of earthquakes in California. It has also helped increase the accuracy of hurricane predictions by 20–25 percent. During Hurricane Harvey in August, researchers used TACC supercomputers to forecast how high water would rise near the coast and to predict flooding in rivers and creeks in its aftermath.

Aaron Dubrow, strategic communications specialist at TACC, says supercomputer users either use publicly available programs or create an application from the mathematics of the problem they are researching. “You take an idea like how cells divide and turn that into a computer algorithm and it becomes a program of sorts,” he says. Researchers can log into the supercomputer remotely or send their program to TACC staff. Stampede2 also has web portals for smaller problems in topics like drug discovery or natural disasters.

For Dan Stanzione, executive director at the TACC, some of the most important research isn’t immediately applied. “Basic science has dramatic impacts on the world, but you might not see that until decades from now.” He points to Einstein’s 100-year-old theory of gravitational waves, which was recently confirmed with the help of supercomputers across the nation, including Stampede. “You might wonder why we care about gravitational waves. But now we have satellite, TV, and instant communications around the world because of Einstein’s theories about gravitational waves 100 years ago.”

According to Stanzione, there were nearly 40,000 users of the first Stampede and an approximate 3,500 projects completed. Similar to Stampede, the new Stampede2 is expected to have a four-year lifespan. “Your smartphone starts to feel old and slow after four or five years, and supercomputers are the same,” he says. “They may still be fast, but it’s made out of four-year-old processors. The new ones are faster and more power efficient to run.” The old processors don’t go to waste though—most will be donated to state institutions across Texas.

In order to use a supercomputer, researchers must submit proposals to an NSF board, which then delegates hours of usage. Stanzione says there are requests for nearly a billion processor hours every quarter, which is several times higher than what is available nationwide. While Stanzione says nearly every university has some sort of supercomputer now, the U.S. still lags behind China in computing power. The world’s top two computers are both Chinese, and the first is nearly five times more powerful than the largest in the states.

Regardless, Stampede2 will still manage to serve researchers from more than 400 universities. Other users include private businesses, such as Firefly Space Company in nearby Cedar Park, and some government users like the Department of Energy and the U.S. Department of Agriculture. Stanzione says all work done on Stampede2 must be public and published research.

“Being the leader in large-scale computational sciences and engineering means we can attract the top researchers who need these resources,” he says. “It helps attract those top scholars to UT.  And then hopefully once they’re here, it helps them reach these innovations a little faster.”