With quantum leaps in performance and legions of competitors, life is tough for those developing networking chips. But not developing them can be even tougher.
Consider Clearwater Networks. The startup burned through three years, three CEOs and $35 million to develop a simultaneous multithreaded network processor. But the company recently closed its doors after failing to complete the design.
I knew that Clearwater was in trouble when the company couldn't tell me, even under nondisclosure, how far the schedule had slipped. But this isn't the only network processor (NPU) company with execution problems.
EZchip announced plans to sample its first NPU in April 2001; the chip sampled in April 2002. Yet, the startup's NP-1 still became the first full-duplex 10-Gbit/second NPU to sample.
Agere preannounced its 10-Gbit/s NPU last summer, promising samples in early 2002, but now says that product will sample in the fourth quarter. Also announcing a nine-month slip in its 10-Gbit/s NPU was Vitesse, which blamed the delay on the closure of its Xaqti design center.
The problem extends beyond network processors. Startup Corrent had planned to sample a 10-Gbit/s security processor in the first quarter but has now redefined that product and pushed it back a full year, promising a 5-Gbit/s part in the third quarter to bridge the gap.
Some vendors are not even reaching their performance targets after delays. Azanda had planned to deliver an OC-192 traffic manager by the end of 2001, but the startup now plans instead to deploy an OC-48 version by mid-2002.
Several of these vendors cite weak demand for 10-Gbit/s products. Even if we delivered on time, they cry, no one would buy it. While the 10-Gbit/s market has not grown as forecast, there will be dozens of design wins this year and many more next year. The companies that deliver 10-Gbit/s silicon sooner will be in a good position to lock up design wins.
Despite this inventory of ineptitude, several companies (such as AMCC and Cavium) are delivering products reasonably close to schedule and promised performance. Others (such as Bay and Terago) are at least disguising delays by not preannouncing schedules and specs.
The moral: Before you buy into some PowerPoint presentation, check out the track record of the design team.
Linley Gwennap is Founder and Principal Analyst of the Linley Group (www.linleygroup.com).