Squeezing the technology orange
Another issue is a lack of a clear technology transition. Initially, vertical-cavity surface-emitting lasers (VCSELs) will be used to deliver optical interconnects in computers "but it is clear VCSELs will not scale," said Beausoleil, noting the industry will need to transition to dense wavelength division multiplexing (DWDM) to deliver multiple links per channel efficiently on chip.
"How do we migrate from VCSELs to DWDM," asked Kash.
"There has to be a migration path, and certainly we don't see it today," he said. At the same time, "there's got to be some smaller markets that we can address on the way to the big vision" of silicon photonics in PC processors, he added.
Another key challenge is reducing costs of silicon photonics, as much as 70 percent of which are now in packaging and testing of optical components. "Testing has to take ten seconds and that's orders of magnitude changes from what we have today," Paniccia said.
The economic downturn might push out the transition to silicon photonics as companies cut back on long term research projects. In his keynote, Beausoleil said the capability needs to be in place in time for chips in 2017.
The economic reality will push out some of the technology," said Paniccia. If you can squeeze the sponge on the current generation technology--that's the default."
"It's like squeezing an orange, and in two or three years it will be like squeezing the peel," said Beausoleil, extending the metaphor. "The physical limits of wires are real and performance is already rolling off, so within a few years a brick wall will appear on the horizon," he added.
"I tell my team to focus on first downs and stay in the game," said Paniccia. "We are just at the beginning of this next generation of technology, but once we get there we will have built out an enormous amount of infrastructure, so if we are successful optics becomes a technology enabler like wireless—it goes everywhere," he said.