Is USB 3.0 a dead end? Intel thinks it is

The Intel Developer Forum is going on in San Francisco. Prior to the start of the conference, observers thought that Intel would announce support for USB 3.0, the new high-speed version of the interface, in a forthcoming chip set. Intel agreed to support the USB 3.0 specification only this summer. However, at IDF, there's precious little USB 3.0 to be seen.

Instead, Intel demonstrated its Light Peak technology on the floor. Light Peak is a fiber-optic interconnection technology that uses ports and connectors that look very similar to, but are incompatible with, USB 3.0. Light Peak has a theoretical maximum transfer speed of 10Gbps on each port, and it can be daisy-chained from device to device. At IDF, Intel demonstrated Light Peak streaming uncompressed 1080P HD to a modified Samsung HDTV at a sustained rate of almost 770MB/second (approx. 6.16Gbps.)

Intel emphasizes the protocol-neutral nature of Light Peak. HDMI, IP. DVI, SDI and other protocols can be sent over a Light Peak connection. Consider that a single Light Peak connection could carry three 3Gbps HD-SDI signals. With the right adapters, future PCs and other devices will no longer be limited in their I/O capabilities. Intel believes that it will deliver Light Peak connections capable of 100Gbps by the end of this decade.

When Intel executives were asked whether Sandy Bridge, the next-generation Intel Core architecture scheduled for delivery by mid-2011 will support either USB 3.0 or Light Peak, they demurred, but they did go on the record saying that Light Peak will be shipping in volume next year. (Update, 16 September 2010: EETimes reports this morning that while Intel plans to ship Light Peak components in 2011, it doesn't expect to see systems implementing Light Peak until 2012. Corrected--the previous version of this sentence said "Sandy Bridge.")

Does this mean that USB 3.0 is a dead end? Not necessarily, since it has a head start in the market and is backward-compatible to USB 2.0. However, if Intel does succeed in shipping Light Peak components next year in quantity, my bet is that we'll see a major move to Light Peak for high-performance connections. Gaming and media creation systems as well as servers will likely adopt Light Peak, as component costs are likely to be high for at least the first few years, and those systems can support higher end-user costs.