All Tier One networking and communications OEMs use NPUs in a range of systems, including edge/services routers, Carrier Ethernet switch/ routers (CESRs), optical-transport platforms, and wireless and broadband infrastructure equipment. Many of these OEMs continue to invest in internal NPUs, which will limit the adoption of merchant products. But even those with internally developed NPUs, such as Alcatel-Lucent, Cisco, and Juniper, are using merchant NPUs in some designs.
After weathering the global economic crisis in 2009, the NPU market in 2010 posted its highest growth rate in four years. For 2011, revenue dipped slightly, but it was still 21% greater than in 2009. After this quick breather, we expect the NPU market to expand in 2012 and continue growing through 2016. Increasing network traffic, which requires expansion of the access network as well as the aggregation and core networks, fuels this growth. In response, NPU vendors have optimized their products for specific network types, such as access versus metro.
Despite being smaller than most competitors, EZchip is the leading supplier of high-end merchant NPUs. The company is shipping multiple product generations including the NP 4, a 100Gbps chip combining a customer-programmable NPU, traffic manager, and Ethernet MACs. EZchip has worked with Marvell to supply semicustom versions of its NPUs to Cisco. Including this Marvell revenue, sales of EZchip’s NPUs grew about 30% in 2011, much better than the overall market. For access equipment, however, the company’s NPUs are less compelling.
In the access market, PMC-Sierra is the leading active vendor. The company entered the NPU market in late 2010 by acquiring Wintegra, which had become a leading vendor of access NPUs because of its flexible architecture and complete data-plane software. Support for legacy protocols continues to differentiate PMC’s WinPath from Ethernet/IP-only NPUs. WinPath has been particularly successful in cost- and power-sensitive wireless-backhaul designs. In base stations, PMC was first to address both transport and RLC/MAC functions for LTE designs.
PMC’s primary competition comes from LSI, which offers its Axxia communications processors for base-station transport and partial baseband processing as well as wireless-backhaul designs. Combining multiple Power CPUs with data-plane technology from the company’s earlier APP line, the Axxia devices represent a hybrid of a multicore processor and an NPU. The company can provide unique solutions by coupling its NPUs with semicustom devices from its ASIC operations. LSI plans to extend Axxia into adjacent markets using future ARM-based processors.
With its new BCM88030, Broadcom sampled the industry’s first full-duplex 100Gbps NPU. The company combines this NPU with its 200Gbps configurable traffic manager, switch fabric, and search coprocessors to offer a packet-processing solution that is more complete than any competitor’s. Customers can combine these components in 400Gbps line cards or use them independently in different designs. The BCM88030 has a design win at NEC, and we expect Broadcom to increasingly challenge EZchip for new opportunities. For access and aggregation applications, Broadcom also offers configurable Carrier Ethernet switch chips in its popular StrataXGS line.
Like Broadcom, Marvell competes against EZchip for high-end NPU designs. In 1Q12, Marvell increased its emphasis on the NPU market with the acquisition of Xelerated. The company’s flagship NPU, the HX, can support 100Gbps line cards as well as standalone pizza-box configurations. The NP-4 and HX family offer similar throughput and integration, but EZchip’s architecture offers greater flexibility, whereas Marvell’s offers greater determinism.
Netronome is now the only privately held company developing NPU silicon. It started to generate NPU revenue in 2011 with shipments of its 40Gbps NFP-3240. Although this device cannot match the throughput of other high-end NPUs, it is unique in its ability to perform advanced services such as IPSec, firewall/NAT, load-balancing, and DPI. In 2012, Netronome disclosed its next-generation NPU—the NFP-6xxx—which should close the performance gap relative to other high-end NPUs. The NFP-6xxx, however, will lag the leaders in sampling by more than a year. The other startup offering NPU technology is Ethernity, which develops its packet processors in FPGAs.
Cisco and other leading OEMs combine search coprocessors with their internal ASIC or NPU designs. Because IPv6 and software-defined networks (SDNs) require more-complex and larger lookups, we expect the search-coprocessor market to grow faster than the carrier-equipment market. Through its NetLogic acquisition, Broadcom becomes the leading supplier of search coprocessors. Its primary competition, Renesas, lags in technology and market share. By offering its new Neuron Search product, Cavium has entered the search-coprocessor market with an algorithmic implementation that promises to reduce power and cost while increasing capacity.