IBM's optical chip speeds data transfers

Stephen Withers
Tuesday, 27 March 2007 06:46

Your IT - Home IT

IBM's announcement of a high-speed optical transceiver chip has more to do with building fast multiprocessor systems than downloading entertainment, at least in the short term.

The transceiver chip being demonstrated this week by IBM scientists at the 2007 Optical Fiber Conference is capable of transferring 160 Gigabits per second, eight times the speed of currently available components.

Although IBM chose to describe the chip's speed in terms of the amount of data in a high-definition movie, initial applications are apparently likely to be within computing systems rather than in broadband connections to homes.

Data flow between individual CPUs in a supercomputer or blade servers is currently a bottleneck because of the amount of space required and the heat generated. By using IBM's transceiver in conjunction with densely spaced waveguides in place of copper tracks, designers stand to save valuable real estate on circuit boards as well as reducing power consumption, while simultaneously increasing performance.

Other early applications are likely to include the high-speed routers that control the flow of data across the Internet's backbone and in high-end corporate and scientific/engineering networks.

The chip was built with "standard, high-volume, low-cost technology used for most chips today," company officials said, with the addition of more exotic materials where necessary. When the chip reaches the market - which could take as long as three years - IBM hopes to deliver optical performance at copper prices.

IBM's technology could eventually prove to be a cost-effective way of implementing optical networks in homes and offices. At present, network interface cards (NICs) and related components using copper-based media are cheaper than their optical equivalents at a given speed. The electronic/optical interface has been one reason for the extra cost.

It is possible that the breakthrough will ultimately deliver a more affordable way of implementing 10G or 100G Ethernet, but that will be of little value to consumers in countries that do not take the Fibre To The Home (FTTH) route.