Having integrated MMC, AMCC is now the only company that is able to offer a complete set of networking silicon including the physical-layer IC, media-access controller, framer, NPU, search engine, traffic manager and fabric-all interoperable. This is a big step forward for a company that until recently focused mainly on PHYs and other interface chips.

AMCC also gives the former MMC more funding than it had as an independent company. This infusion of cash has helped the company to develop a family of new network processors, traffic managers and switch fabrics.

The company is already shipping the nP7120. This Gigabit Ethernet NPU chip has helped it rack up 100 design wins for its NPUs. AMCC will soon begin sampling the nP7250, designed for OC-48 applications.

The company also plans to sample a 10-Gbit/second NPU by the end of this year. This schedule would make the chip one of the first 10-Gbit network processors, if not the first.

The 10-Gbit/s chip will provide more than just bragging rights. AMCC will be the only company to offer network processors spanning 1 to 10 Gbits, all software compatible.

One flaw is the company's lack of support for industry standards. Although AMCC's chips interoperate, they depend on proprietary interfaces. This makes it difficult for an OEM to combine an AMCC NPU with, for example, Vitesse's switch fabric. These proprietary interfaces can lock an OEM into the AMCC chip set, making it an all-or-nothing decision.

The company plans to support standard interfaces in the future, but even the 10-Gbit/s NPU will use proprietary interfaces.

With so many vendors targeting the NPU market, perhaps the most important success factor is execution. While Intel and Motorola are delayed in upgrading their initial NPU products, AMCC is moving forward. If the company continues to execute to plan, AMCC should be a formidable force in the NPU market.

Founder and Principal Analyst of the Linley Group, Linley Gwennap is the co-author of "A guide to network processors" (www.linleygroup.com/npu).