A lifetime tech tinkerer, Gara spent his first career in physics, trying to unlock the fundamental mysteries of matter. Such problems require massive compute power, and in the mid-1980s, supercomputers weren’t yet available off-the-shelf.
Learn more about Gara’s intriguing career from academics to IBM to Intel, his favorite forms of travel, and the kinds of ideas and projects this physicist-turned-supercomputer builder daydreams about.
1. He’s designed and built dozens of world-beating supercomputers. Aurora, which will be at least four times as capable as today’s top supercomputer, stands on top a mountain of Gara-designed machines. Before joining Intel, Gara was an IBM Fellow and chief architect of three generations of IBM’s Blue Gene series. If you look at the list of the world’s top 10 supercomputers over the last 10 years—100 machines in total—you’ll discover Gara was the lead architect for more than a third of them.Why come to Intel after such a run? “One of the things was to really find a place where we could have a broader impact than just the very top end of HPC, which is what IBM was doing,” Gara notes. “The bulk of the HPC market was still being driven by Intel processors. If I wanted to be where the action was going to be, Intel was the place to be.” For the first time since 1997, Intel is the “prime” contractor on Aurora—we signed a $200 million deal to deliver the entire system and its software, built in partnership with Cray, a top supercomputing OEM.
2. His career has gone from computers to physics to computers — to physics. After tinkering with Z80 processors and soldering together “very simple but intelligent things” in high school, Gara double-majored in electrical engineering and physics. “To understand how nature worked was extremely exciting and compelling,” Gara says, which led him all the way to a doctorate in physics at the University of Wisconsin, Madison. But the computers were never far away. Gara studied the field of quantum chromodynamics (QCD) within theoretical physics, which is the calculation of what makes up atoms, how they hold themselves together, and what properties they have (the kind of study advanced at the Intel-boosted Large Hadron Collider in Europe). “The calculations for that were intractable from a paper-and-pencil perspective,” Gara explains, but well-suited for a computer. A really, really fast computer. “That application still uses more supercomputer cycles over the whole world than any other application there is.” Gara’s first big job came at Columbia University, where he focused on both theoretical and experimental physics. “In experimental physics, we built a data-flow processor to do online track reconstruction for a physics experiment that was at Chicago, at Fermilab,” Gara explains. “I got a lot more experience there, and then moved into really utilizing that expertise to do theoretical physics calculations, pulling everything together.” Those early supercomputers were built by very small teams—three or four senior people with experience plus a handful of grad students, Gara says. “It was fun, but it was a lot of work, as you can imagine, too.” In 1999, IBM gave Gara the opportunity to take those systems and build something more general-purpose, and the Blue Gene family was born. Today, Gara’s former academic and research cohorts in QCD are his most eager customers. “That group of people are very adept at using supercomputers, so they soak up every spare cycle that any supercomputer in the world has,” he says. “Typically when a new machine is christened, they are the first ones knocking at the door saying, ‘This whole thing—just give it to us. We’ll use the whole thing.’” “I always loved physics,” he adds. “And QCD has turned out to be a great application to shake out a machine because it pushes many dimensions of the system.”
3. He’d rather be off the beaten path—or pedaling up the Alps. Between the doctorate and the research work, Gara spent six months backpacking all across Asia. “I always had the travel bug,” he says. His favored destinations stray “off-the-beaten-path… places that have rarely seen a footprint,” including “some wild corners of Southeast Asia and India” plus a trip into the Amazon jungle. “It’s harder and harder to find [a quiet place], but it’s still out there,” Gara notes. One beaten path keeps him coming back. “Once a year a group of friends from graduate school try to rendezvous in Europe and ride around in the French Alps,” he says. “We’ve got a little bunch of friends, and we ride some of the same routes that they do in the Tour de France. But we do it, of course, at a different pace than they do.”
4. He’s teaming with his son to write a book on the physics of cycling. Gara’s love of physics, cycling, and tech-tinkering are coming together in one unique little project: Gara and his younger son are writing a book on the physics of bicycling.“There are all these myths and concepts that cyclists have about things, like when you go into a corner, should you lean the bike or should you lean your body? Why should you pulse your brakes rather than squeeze them with constant force? There are about 30 such things we are working through with aspirations of pulling together a short paperback.”He’s built microcontrollers to collect data to support the theory for some of the concepts. “It should be a fun little book,” Gara says. “We still have a ways to go. It’s one of those things that’s slowly in progress.”
5. In his daydreams, he’s trying to invent a vacuum balloon. A scientists’ dream since 1670 (according to Wikipedia), a vacuum balloon is a hypothetical airship where the balloon is not filled with lighter-than-air gas but rather, evacuated to a near-vacuum inside—for maximum lift. The catch? The balloon would have to be extremely strong yet light enough to float (and therefore worth the trouble/challenge).
The idea is “on a very slow back burner in my mind,” Gara concedes, “but I’m approaching it mostly from a theoretical perspective in terms of trying to do all the calculations.” A newer possibility is to take the concept of the carbon nanotube and build a sphere—a carbon nanobubble.
“If you could make these large enough and had a vacuum inside, they would basically form a material that would be lighter than air but built out of carbon bubbles,” Gara explains. “I was doing molecular dynamic simulations to see how stable that structure would be—again, one of these things just playing around with in my spare time—to see if that would actually result in a stable structure. If there were a way to make these stable, it would make for an amazing structural building block.”
“My guess is the answer is ‘No,’ based on conversations I’ve had—but I haven’t completely given up.”
–By Jeremy Schultz, Intel Employee Communications Team