A team of scientists at IBM Research has achieved a breakthrough in nanophotonics that could hasten the arrival of machines capable of performing one million trillion calculations a second. These so-called exascale computers, which are expected to arrive by 2018, will be about one thousand times faster than today’s fastest supercomputers. The scientists have demonstrated their technology in the lab, and IBM is now transferring it to one of its chip fabrication plants–the first step in commercialization.
The breakthrough is the culmination of eight years of work by a small team at IBM Research, including William Green, Solomon Assefa, Alexander Rylyakov, Clint Schow, Folkert Horst and Yurii Vlasov. They’re presenting their results today at a chip technology conference in Japan.
The new technology, called CMOS Integrated Nanophotonics, integrates traditional electrical components with optical components on a single chip–in fact, on a single layer of silicon on a single chip. In today’s advanced computers, optical elements are placed on separate chips, which impedes the performance of the computers and adds to costs. “This is about the technology of the future, which will enable going to exascale computing,” says Vlasov, the research team manager.
The announcement places IBM at the front edge of chip industry innovation. Several start-ups in Silicon Valley as well as larger companies around the world are pursuing similar goals. For example, Luxtera, a Silicon Valley startup, is already producing chips with integrated optics and electronics, but its approach yields devices that are 10 times bigger than those the IBM technology is designed to produce.
The IBM advance could have a profound impact on the electronics industry. By integrating nano-size optical devices on a single layer of a single chip, the researchers have made it possible to produce large quantities of such devices using the same chip production processes that are already in use. The technology has implications not just for high-performance computing but for such uses as network switching and even consumer handheld devices. It’s useful wherever people need to transmit massive amounts of data at a low price.
Vlasov uses a metaphor to explain the role his team’s technology will have in computing: The human brain includes gray matter, which processes information, and white matter, which connects the gray matter cells. In the brain-like computers of the future, nanophotonics will play the role that white matter does in the human brain.
For more info, go to http://www.research.ibm.com/photonics