Revving Up Moore's Law

Tom Conte likes cars. He gets paid to talk about computers. Problem is, people tend to be more mystified by computers than by cars. So Conte puts the two together and uses cars (and things related to them) to explain computers. He does this a lot.

"For instance, I’ve compared pipelining and parallelism to multi-lane highways, and I’ve compared computer architecture to designing efficient interchanges,” says Conte, professor of computer science. “You use different cars for different purposes, just like computers.

“People,” he concedes, “get sick of my car analogies.”

Tom Conte
 

Or maybe they just recognize passion when they see it—much like the drive and energy toward computers that has propelled Conte intellectually since his teenage days in Delaware, when “computer room” were still novelties in American high schools. Programming came naturally to Conte, who likened their intellectual tidiness to geometry proofs.

“It fired some neurons,” he says.

A few years later Conte was about to finish his bachelor’s in electrical engineering from the University of Delaware, and he was looking around for graduate schools where he could build computers. At the University of Illinois, he met Ed Davidson. Working for Davidson, Tom found that the design for a newsupercomputer had a fundamental flaw; its processor-to-memory interconnect—the multi-lane highway between the memory and the CPUs—would not take six CPU cycles, as they thought. Conte found out it would take more like 11,000 cycles.

In 1992, armed with a brand-new Ph.D. from Illinois, Conte sent out more than 70 job applications, which resulted in a grand total of three interviews and two job offers. Luckily, one of those offers was from the University of South Carolina, which may not have boasted a Top 10 computer science program but did share its locale with one Ms. Catherine Linder, who soon thereafter was set up with Conte on a blind date. The two were married in March of 1994.

Tom Conte

Students may sometimes groan, but Professor Tom Conte of Computer Science gets a lot of mileage out of his car analogies when teaching computer architecture.

 

It was in Conte’s next job, at North Carolina State University, where his career really hit its stride. “N.C. State was a natural choice; I looked for an engineering university in the South that had ties to industry,” says Conte, who joined the school’s electrical and computer engineering faculty in 1995. “They had a new department head, who trusted me as an assistant professor to set the vision for how we should grow computer engineering. We recruited [new Ph.D. graduates] from the top schools, and I’m proud to say [N.C. State’s] computer architecture program is one of the top programs around the country.”

Recruited in 2008 to lead Georgia Tech’s computer architecture group, Conte hopes to accomplish something similar in Atlanta in the field of manycore computing. Conte envisions a future of computation in which single chips carry thousands of processing cores, enabling speeds that vastly surpass current capabilities. The key to building one of these machines, he says, is to tap the ideas and know-how of several disciplines. And that’s why he chose to come further south.

“Georgia Tech has all the ingredients,” says Conte, recently elected to the IEEE Computer Society’s Board of Governors, as well as named editor in chief of the Association for Computing Machinery’s Transactions on Architecture and Code Optimization. “We have a top computer science department, and a top electrical engineering department. I view this as yet another opportunity to do what I love, which is to build computers.”

And talk about cars. And do woodworking and photography. And collect vintage calculators and watches. (“I have too many hobbies,” Conte says.) He is fascinated, he says, by purely mechanical methods of keeping time, by watches and clocks built solely of gears that can stay accurate to within a second a week. Though he carries the title of computer scientist, Conte is at least as much an engineer, something that comes both from personal proclivity and professional training.

“I do an inordinate amount of industry-funded research—you have to, in computer architecture,” he says. “It keeps you practical and grounded in real problems. People ask me about computer architecture: ‘Is it how you build transistors?’ No, I tell them, it’s about how you organize the system. Just like a highway engineer.”