Startup Claims Chip Breakthrough
Under normal circumstances, that would hardly merit much of a mention in Light Reading, but there’s a lot more to West Bay than meets the eye. The startup claims "a revolution in chip architecture" that will enable it to make semiconductors that can perform complicated tasks with fewer transistors than those of other vendors. The upshot for telecom equipment vendors is more functionality, less real estate, and lots of cash saved -- a big deal, in other words.
Here’s the score: The traditional way of developing more powerful chips is to cram more and more transistors onto them. In fact, the number of transistors that can be squeezed onto a chip has doubled every 18 months for so long now that the phenomenon has become known as Moore's Law.
The big problem for telecom equipment vendors is that their performance requirements are growing faster than Moore's Law, as traffic volumes on the Internet continue to soar. As a result, chip makers are finding it difficult to keep up.
West Bay’s founder, president, and CEO, Tino Varelas, reckons he can solve the problem -- and leap-frog a couple of Moore's Law cycles -- by using brains rather than brawn. Specifically, West Bay is arranging the transistors inside chips in a special (secret) way so that together they deliver a much bigger bang for the buck.
West Bay's chips will have the functionality of competitor's chips four to eight times the size, according to Varelas. West Bay's first product will be a 16xOC3 framer, samples of which are expected to ship in the first quarter of next year. This chip sits behind the optics and processes Sonet frame information. First, it processes the header to identify where the information is sitting in the payload; it then presents that information to the backplane.
Concentrating on Sonet is one of the things that makes West Bay stand out in a crowd. "Communications chip makers are concentrating on 'fat pipes' for data at the expense of good old-fashioned voice circuits. But we're not in a totally data world yet," says Varelas. In the real world, most services come in at low channel speeds like STS-1 (51.84 Mbit/s), he adds.
Varelas claims that West Bay's chip can handle both data and voice (Sonet) information equally well. The key here is to design chips that take low-speed channels, such as OC3 (155 Mbit/s), and send them on separate paths through the chip.
This brings big benefits to equipment manufacturers. One 16xOC3 framer chip will do the work of 16 individual OC3 cards sitting in the equipment rack, according to Varelas. What's more, the OC3 channels can carry a mix of protocols on different channels. In fact, West Bay’s chips can subdivide each 155-Mbit/s data stream so that it can carry three independent payloads, he claims.
It's worth pointing out that others, like Applied Micro Circuits Corp. (Nasdaq: AMCC) and PMC-Sierra Inc. (Nasdaq: PMCS) have announced so-called channelized products in the past year (see AMCC to Supply Chips for Alcatel Router and AMCC Intros RHINE IC).
However, these other companies don't take the concept far enough, says Varelas. He claims that West Bay's framer chip will have almost three times the functionality of AMCC's product. Though both are nominally 16xOC3, West Bay's design processes 48 path streams, while AMCC's Rhine chip only handles 16. AMCC hadn't responded to requests for comment at press time.
Of course, since West Bay's chip is still in the lab, it's difficult to prove these claims. Are they believable?
Here are three reasons why Light Reading thinks they could be:
One: The management team of West Bay has high-powered credentials. It was started by a group out of PMC-Sierra: Oon-Sim Ang, Tze-Wo Leung, Barry Tsuji, and Tino Varelas. Varelas himself was one of the founders of PMC (before it linked up with Sierra Semiconductor) back in 1988. The startup has lured away a bunch of other people from PMC-Sierra as well as from companies like Agilent Technologies Inc. (NYSE: A) (see Chong Ong Moves On).
Two: West Bay has secured sufficient funding from private investors to last for three whole years without any further finance. Varelas declines to put a figure on this funding, but it must be sizeable, considering that the company already has a staff of 24.
Last, but not least: According to Varelas, the startup already has the commitment of a tier-1 equipment manufacturer -- a Cisco, Lucent, or Nortel -- to use its product. Varelas couldn't be coaxed into revealing exactly which.
-- Pauline Rigby, senior editor, Light Reading http://www.lightreading.com