Nikos Hardavellas Receives Prestigious Honor for Young Faculty
High-risk project to address enormous energy waste of modern computing systems
- Only a fraction of computing capability on processors currently can be used
- Computer engineer designing new computer chips to address this problem
- Hardavellas works at intersection of computer architecture and nanophotonics
EVANSTON, Ill. --- Nikos Hardavellas, assistant professor of electrical engineering and computer science at Northwestern University’s McCormick School of Engineering and Applied Science, has received a prestigious Faculty Early Career Development (CAREER) Award from the National Science Foundation (NSF).
The CAREER award, the NSF’s most prestigious honor for junior faculty members, supports early career development of individuals who exemplify the role of teacher-scholar through outstanding research, excellent education and the integration of education and research. The minimum CAREER award size is $400,000 for a five-year period.
“Computing consumes inordinate amounts of power, but the vast majority of this power is wasted,” Hardavellas said. “The NSF CAREER award will enable my group to commit substantial resources to carry out a high-risk, high-reward research project using nanophotonics and other novel devices to design new computer chips to address this major problem.”
Hardavellas’ expertise is in parallel computer architecture, especially at the intersection of computer architecture and nanophotonics, and design for dark silicon, memory systems, approximate computing and data-oriented software architectures.
The lab he founded and directs, PARAG@N (Parallel Architecture Group @ Northwestern), works on all these issues and addresses problems in the general area of power and energy-efficiency of computing systems, often by incorporating emerging technologies and unconventional approaches.
His CAREER project is “Energy-Efficient and Energy-Proportional Silicon-Photonic Manycore Architectures.” The power consumption of modern computing systems is already a limiter of growth and big science, Hardavellas says. This constraint is so severe that even today only a fraction of the computing capability put on a processor can be used. Large systems quickly become prohibitive: A single next-generation supercomputer could consume as much power as the households of an entire county with 20,000 residents, for example.
Hardavellas’ project aims to utilize nanophotonics to design novel energy-efficient processor architectures that break free of the area, power, bandwidth and yield constraints of traditional chip designs. This could lead to processors that reach scales of thousands of cores, at a performance and power envelope impossible to realize with conventional technology.