Will Intel and USB make fiber optics mainstream?
The powerful chipmaker hopes its Light Peak project will bring fiber optic speeds to ordinary computer users, and USB is the logical path to get there.
SAN FRANCISCO--You've probably heard about fiber optics for years--some kind of exotic technology used to carry gargantuan quantities of data across continents. But in the not-too-distant future, you might be plugging these tiny glass strands straight into your computer.
That's if Intel gets its way. At itslast week, the chipmaker for connecting many devices to PCs with fiber optic lines. Intel secured major Light Peak endorsement from Sony and now it's has begun trying to make it into an industry standard.
But bringing optical technology to the masses will require more than Intel Chief Technology Officer Justin Rattner taking the stage to connect a thin white Light Peak cable into the back of a prototype PC. According to sources familiar with the situation, the most likely mechanism to carry Light Peak out of the R&D lab to the edge of your laptop will be the venerable Universal Serial Bus, and Intel has begun pounding the pavement to try to make that happen.
"Now all the pieces are in place," Rattner said. "We need to get a standard established to turn on the entire ecosystem to Light Peak."
Even technophobes are familiar with USB. The plug-and-play technology started its journey in PCs and has spread to handsets, consumer electronics devices, digital cameras, and more. And new developments from the group behind the standard, the USB Implementers Forum, could expand adoption more, with a new faster, more power-efficient version and with technology to make it better for charging devices plugged into a computer or power outlet.
The new "SuperSpeed" USB 3.0 has 5 gigabit-per-second data transfer rate, more than 10 times that of the USB 2.0 version that prevails today, and the first USB 3.0 device achieved certification last week. A separate new USB feature increases the amount of power that USB devices can use from 0.5 amps to 0.9 amps while adding another 1.5 amps specifically for charging batteries, making USB for tasks besides just transferring data.
The 5Gbps speed is a big step up; NEC's demonstration of its newly certified USB 3.0 controller showed 500MB of data transferred in 4.4 seconds with USB 3.0 compared to 39 seconds with USB 2.0. But for USB to really break out--to accommodate the data transfer needs of a large 3D TV screen, for example, or to synchronize a terabyte-capacity iPod in moments--there's still more work to be done.
Enter fiber optics.
"At some point the industry is going to have to transition," Jeff Ravencraft, the USB-IF's president and chairman, said in an interview, because copper wires such as those in the current USB 2 and new USB 3 standards have limits on how fast they can transmit signals. "I think the next transition is going to be to optics."
Intel's aspirations and allies
Intel's hope for Light Peak is to create a single connection for video, storage devices, the network, printers, Webcams, and anything else that plugs into a PC. Light Peak uses circuitry that can juggle multiple communication protocols at the same time, and the Light Peak promise is for a universal connector to replace today's incompatible sockets for USB, FireWire, DVI, DisplayPort, and HDMI. It's a hot-plug technology, meaning that devices can be linked when they're up and running.
Intel has pre-production chips and said the technology will be ready to ship in 2010. In its current form, Light Peak can transfer data at 10Gbps each direction along the fiber optic line, but Intel said Light Peak will reach much higher speeds--100Gbps in the next decade, according to Jason Ziller, director of Intel's optical input-output program office.
The Sony endorsement is important, because the company sells PCs, music players, cameras, video cameras, and Blu-ray players. But another company at least as significant had a quieter Light Peak appearance at the Intel show: Apple.
Intel's second demonstration of Light Peak, in which a single cable transported high-definition video and data to a storage system at the same time, used a Mac OS X computer. Apple would be a strong ally: it has influential designs with an emphasis on uncluttered appearance and ease of use, it's willing to take a stand for technology it believes is superior, and its iPhones and iPods that take ever longer to synchronize with a PC as storage capacity expands.
And on Saturday, Engadget reported that Apple isn't merely a Light Peak ally, but that it brought the Light Peak idea to Intel and has plans to bring it to Macs next year. Apple and Intel declined to comment on the matter.
But do we really need to go all the way to optical now? High-speed electrical communications is hard--wires can cause electromagnetic interference, for example, and USB 3 cables can only be 3 meters long compared to 5 meters for USB 2. But technology for transferring data over copper wires, like technology for shrinking computer chips, has defied predictions that it will run out of gas.
The short answer is there is a need. Video screens are getting larger, expanding beyond HD TV's 1920x1080 pixels, and 3D video requires a doubled data transfer rate. Richard Doherty of the Envisioneering Group expects even the newer DisplayPort video standard has only about 24 to 30 months before new technology needs more capacity than it can supply.
"Optical may be the only way to do it," he said, saying the need for 60Gbps transfer rates is on the horizon.
USB group to standardize Light Peak?
Ravencraft wouldn't comment on whether the USB group is working with Intel on adopting Light Peak for the coming transition to optical communications, but there are indications that could happen.
For one thing, the USB 3.0 specification explicitly accommodates optical lines in the cable's connector, a move to try to "future-proof" the standard. For another, when Intel demonstrated Light Peak, it used USB connectors on its prototypes. Ziller said in an interview that nothing should be read into that choice, but it was conspicuous nonetheless.
Light Peak discussions are under way at the USB group, said Steve Roux, senior director of business development at NEC and a member of the USB Implementers Forum board.
"Through the USB-IF we're looking at it. It's clearly something we'll have to pay attention to," Roux said, adding, "We don't see it as a USB 3.0 killer."
The politics of standardization are another reason the USB-IF makes sense for Light Peak. Along with leading developers such as Intel, Hewlett-Packard, NEC, Texas Instruments, ST-Ericsson, and Microsoft, there are more than 200 companies involved in USB development. The USB-IF devotes two whole pages crammed with corporate logos in its presentation to illustrate who widespread USB buy-in is. And to meet Intel's ambition, Light Peak will need to win over the video community as well as those who presently use USB.
There are other groups where standards are set--the Institute of Electrical and Electronics Engineers, or IEEE, for example, which oversees the USB competitor FireWire as well as 802.11 for wireless networks and 802.16 for Ethernet networks.
Doherty believes both are possible. The USB group could get a connector defined rapidly for consumer use, and the IEEE could work on a variation for higher-end systems such as servers, with optical lines linking processors together and linking computers to storage systems.
"If this starts out as a 100-gigabit USB 3F (F for fiber) connector, there's nothing precluding it from going to IEEE and becoming a 10-terabit link with the same connector," Doherty said.
The money question
There's another obstacle besides politics that Intel and any allies must reckon with: cost.
Optical networking, in which lasers send information as photons down transparent fibers, doesn't come cheap.
Doherty believes high-volume production could lower its costs, though. And USB is nothing if not high-volume: about 3 billion USB devices ship a year right now, according to In-Stat.
"There's every indication that if they get down to USB economics, they can get the cost of the connector down to tens of cents instead of the tens of dollars of most high-performance fiber optic connectors now," Doherty said.
One way to cut costs would be to use plastic fibers rather than higher-quality glass, Doherty said. That limits data-transfer capacity compared to glass, but plastic is cheaper and also is more flexible, Doherty said. Here's a sign Intel agrees: Ziller said of Light Peak, "You can tie a knot in it and it'll still work."
Intel has conducted plenty of research into silicon photonics, in which lasers are built into processors themselves, but Light Peak uses more conventional technology for the optical modules that convert ones and zeros into light at and vice versa. Ziller said Intel is using optical modules from mainstream manufacturers such as Avago Technologies, SAE Magnetics, and Foxconn.
Linking two wires is well understood, but how exactly does that work with two fiber optic lines? In Rattner's demonstration featured a hot-plugged Light Peak cable, so evidently Intel has an idea how to make it work economically.
High-end fiber connections are made by fusing the optical lines, but Doherty believes a gel-like adhesive, perhaps protected by a sheath that snaps back when the connector is plugged in, could be used. "It may not be for things you take on and off a hundred times a day," he said, but such a connector could be used dozens or hundreds of times.
Plenty of Intel ideas have flopped, but the company does have more experience than most introducing complicated technology. And it's not putting on the hard sell for Light Peak.
"We're talking hundreds of millions of ports over next the few years, which really will help drive the costs down and make it an attractive technology," Rattner said. "Fundamentally, we believe the time has come for the optical technologies to go high volume."
Correction, 6:03 a.m. PDT: This story was updated to correct a statement about the time needed to transfer 500MB over USB 2.0 in NEC's demonstration. The time was 39 seconds.