A Tall Tale of 10-Gig
This month, six component vendors have made 10-Gbit/s chip announcements, signaling the first step toward widespread use of higher data rates in carrier networks. But one of the products announced doesn't seem to be on the same wavelength as the rest.
The 10-Gbit/s announcers includes Broadcom Corp. http://www.broadcom.com (May 1); Applied Micro Circuits Corp. http://www.amcc.com (May 1); NewPort Communications Inc. http://www.NewPortCom.com (May 8), PMC-Sierra Inc. http://www.pmcsierra.com (May 4), and most recently, Vitesse Semiconductor Corp. http://www.vitesse.com (May 15).
And the invasion of the 10-Gbit/s chipsets isn't over yet. As this went to press, Conexant Systems Inc. http://www.conexant.com and nSerial Corp. http://www.nserial.com said they planned near-term rollouts.
Of the products announced, all but Broadcom's are Sonet chips, designed to perform a range of functions in routers, switches, and DWDM muxes at rates to OC-192 (roughly 10 Gbit/s). And that's highly significant: Carriers are demanding that equipment vendors take the next step up the Sonet hierarchy. As a result, the OC-192 market is expected to grow 250 percent this year, according to research by the Dell'Oro Group Inc. Sonet's Shooting Star: OC-192 ADMs. These are the components that will make it happen.
Where does this leave Broadcom? Sadly, in a curious limbo of its own making.
Initially, Light Reading fell prey to the furor generated by Broadcom's claim to have developed "the world's first single-chip 10-gigabit per second Ethernet transceiver." We were in good company. Broadcom's shares rose 7.75 points in one day, to 182 the day after the announcement. (Since then, the share price has drifted steadily earthward. It closed at 166 5/8 on May 16.)
After the announcement, however, some disappointing facts surfaced about Broadcom's chip. First off, it isn't really a 10-Gbit/s chip. Instead, it offers four channels of 3.125 Gbit/s. To obtain 10-Gbit/s rates in real-world devices, an equipment vendor needs to aggregate the output of the channels and send that to a single laser driver. And while Broadcom says it has worked with a partner DWDM vendor to successfully achieve the combination of channels for delivery over fiber, it refuses to identify the vendor.
Broadcom also isn't the first to offer a chipset that supports 10-Gbit/s Ethernet. In fact, all of the recent announcers of 10-Gbit/s chips claim to be able to support Ethernet at up to 10 Gbit/s rates and beyond -- albeit over Sonet.
But none except Broadcom claims to support standard 10-Gbit/s Ethernet. The reason is simple: There is no standard to support.
"An optimistic date for completing the standard is March 2002," says Richard Taborek, chief technology officer at nSerial Corp. Taborek served as an editor for the IEEE 802.3z Gigabit Ethernet specs and is presently active on the IEEE 802.3ae committee (although he says it's too soon yet for any specific duties to have been assigned). "We are still in the process of defining the issues we need to address. Broadcom's announcement is compatible with the direction of our presentations," he says. "But for any vendor to claim compliance with a standard we expect to see in 2002… well, that's stretching things."
Others weren't so kind. "Broadcom's way past the acceptable levels of fact distortion," said one reader, who claimed to be a Broadcom customer but asked not to be identified. "By now, I am pretty used to seeing press releases filled with distorted facts from .com companies and startups, but a profitable company like Broadcom should not have to use these kinds of tactics."
-- by Mary Jander, senior editor, Light Reading http://www.lightreading.com