While shrinking transistor sizes to ever decreasing nanometer lengths is a popular topic in the industry today, several companies are working towards removing physical copper connects altogether, replacing them with super high speed laser connections
As computers chip continue to decrease in fabrication size, manufacturers such as AMD and Intel are researching new ways to overcome physical barriers. Die size, performance, operating frequency and heat are all major obstacles in the semiconductor industry. Sun Microsystems announced today that in partnership with Luxtera, Kotura and Stanford University, it is working on an ambitions project to move data transmissions from electrical signals over copper wires to pulses of light using lasers.
With $44 million USD in funds from the Pentagon, Sun will attempt to overcome many bottlenecks that manufacturers are facing today. The goal is to move information faster. According to Sun, processors that are thousands of times faster than ones we have today will be possible if the project is successful.
Today, chips communicate with each other using copper connects, and interconnects for inner-chip communications. Inherent with chip designs, the more bandwidth that's required, the more physical copper lanes are required as well as higher speed processing. With devices shrinking and space consumption becoming a hot direction for the industry, continually adding more lanes is impractical. Replacing a 32 physical lane bus with one that uses one laser lane with higher bandwidth, is the goal.
While the concept appears simple, Sun says that the project is extremely risky. Even with careful designs, Sun expects up to a 50-percent failure rate in producing prototypes. Sun's success rate will depend heavily on manufacturing abilities to accurately align chips. Light channels called wave guides will have to be aligned with high precision, in order for the technology to work - easier said than done.
"We expect a 50 percent chance of failure, but if we win we can have as much as a thousand times increase in performance," said Ron Ho, one of Sun's primary researchers working on the project.
From the press release:
"The historic accuracy of Moore's Law, which predicts a periodic doubling of the number of transistors that can cost-effectively build on a single chip, is partly behind the impressive growth of microprocessor performance over the last 30 years. Today, though, continued improvements are slowing down, as power and size constraints limit the growth of chip clock frequencies."
Sun calls its project macrochip technology, completely redesigning the way in which chips communicate with each other. If this is successful, circuit designs will fundamentally change, both on the silicon level and on the board level.
Sun isn't the only company working on photonics. NEC, Intel, IBM, HP and even MIT are investing vast sums of resources into the same research. Working with photonics firm Luxtera and Kotura, Sun's project was favored by the Pentagon's Defense Advanced Research Projects Agency because of Sun's previous success in aligning silicon dies edge to edge to improve chip-to-chip communications.
Sun claims that its technology is very close to being commercialize. Its partner, Luxtera, already has near-production level prototypes. Sun hopes to produce photonic chips that are able to process millions of trillions instructions per second.