For years, chipmakers have bumped against the ceiling of Moore's Law. Current fabrication techniques can't keep CPU speeds climbing at the meteoric rates of decades past. Because of this, today's advances focus on multiple cores and power savings, rather than raw speed. But what if there was a new way to build chips -- one that would accelerate processing literally to the speed of light?
Using light to turbocharge data transfer is not new. SANs have benefited from high-speed fiber-optic links for years, and U.S. telecommunications providers have begun using similar technology to offer customers Internet bandwidth comparable to that of a dedicated LAN.
Building entire computers with optical circuitry would certainly provide significant advantages. First, particles of light, aka photons, travel much faster than electrons. Perhaps more importantly, they do not radiate energy the way electrons do, even at high frequencies. Thus, not only would purely optical chips outperform anything previously known, the cooling problems that plague electronic microprocessors would virtually disappear.
It's no surprise, then, that scientists have pursued optical computing for decades. For many years, their efforts yielded only purpose-built devices suited to specialized tasks. Then, in 1993, researchers at the University of Colorado announced the first experimental general-purpose optical computer. Although the prototype proved optical computing was technically feasible, a practical design with real-world applications remains elusive.
For starters, the Colorado device could move only one photon at a time -- unlike traditional supercomputers, which process thousands of operations simultaneously. Worse, it relied on large bursts of photons to control the individual photons that made up its data, making the computer extremely energy-inefficient.
But the quest isn't over. In 2007, the Lukin Group at Harvard announced an optical circuit that needed a single photon to operate. Still, 15 years after the University of Colorado experiment, we're no closer to optical computers for mainstream use, let alone a working prototype that bests an ordinary computer. Photon-powered processing may yet be in your future, but you just might have to travel at the speed of light to get there.
-- Neil McAllister