Editor: Linley
Gwennap In
This Issue
Time is running out to reserve your place! On June 14 The Linley Group will host a one-day seminar on High-Speed Interconnect Design at the DoubleTree Hotel in San Jose. Hear industry experts present the latest solutions in board and system-level interconnects. The seminar wraps up with a panel discussing emerging trends in interconnects. Qualified attendees earn FREE admission courtesy of our sponsors, AMCC, Freescale, StarGen, Tundra, PLX, IDT, Dune, Pericom, and The Linley Group. For a complete program listing and registration information, visit our web site. Last week, PMC-Sierra introduced the MSP8520, the first product in a new family of integrated processors. The new chip combines a 1GHz CPU, based on PMC's own E9000 CPU design, with a security engine licensed from SafeNet. The highly integrated processor also includes a DDR2 memory controller, dual GbE MACs, and other I/O. A version without the security engine is available as the MSP8510. The company has been quietly sampling the new devices since late last year and expects them to reach production this month. These devices are the first to implement PMC's new strategy for high-end processors. Last year, the company decided to cancel development of the 1.8GHz RM11000 CPU and lay off its CPU development team. Due to limited interest from Cisco and other key customers, PMC no longer recommends its integrated RM9000-family processors for new designs, although it is supporting existing customers. As the MSP8520 shows, however, PMC continues to offer high-performance integrated processors. The MSP8510 is, in fact, identical to the old RM9150, and the MSP8520 adds the encryption engine to better position it for security applications. The company plans to develop additional versions of the MSP8000 family for specific vertical applications, such as storage. These products extend the MSP line, which also includes processors for VoIP and DSL/PON gateways. Having lost its role as Cisco's primary supplier of control-plane processors, PMC's new strategy leverages its existing processor portfolio for new applications while maintaining low development costs. So far, the company has used a shrewd combination of acquisitions and IP licensing to enhance its processors with new capabilities. If PMC continues to execute well, it will remain a key supplier of high-speed processors. —Linley Additional
coverage of PMC-Sierra's MSP processors will appear in our
upcoming report A Guide to SOHO Gateway Processors.
In May, Bay Microsystems disclosed a new chip, called Chesapeake, designed to be the first integrated NPU/TM device to operate at 40Gbps. Based on Bay's existing Montego architecture, the leading 10Gbps NPU/TM device, Chesapeake cranks up the performance to 122 million packets per second, enough for two full-duplex OC-192 ports or 24 GbE ports at minimum packet sizes. The chip includes a wire-speed SAR, allowing it to perform interworking among Ethernet, POS, ATM, and other protocols at these speeds. The chip also contains a sophisticated traffic manager that is equally adept at handling unicast and multicast (e.g. IP TV) traffic while providing QoS for voice and video and enforcing service-level agreements (SLA). The chip's unique internal architecture is powerful yet programmable. Bay does all of the programming of the internal engines, providing the OEM with an API to simplify creating the data-plane code. Most remarkably, the chip is expected to require only 16W at 40Gbps. The company expects Chesapeake to sample this summer. Although Xelerated's X10 was the first NPU to hit 40Gbps, that product lacked TM or SAR functions. With its interworking capability and high level of integration, Chesapeake is ideal for multiservice switches (MSSW) and other Metro equipment. A single chip can support a complete pizza-box design or a powerful line card. Assuming Bay can deliver Chesapeake as promised, the company should have smooth sailing in the metro NPU market. —Linley Additional coverage of Bay's products appears in our report A Guide to Network Processors.
At Globalcomm this week, AMCC unveiled a chip set targeted at cellular infrastructure. The set includes a new network processor, called the nP3665, as well as AMCC's Amur framer and the company's newest high-performance CPU, the PowerPC 440GRx. Rounding out the offering, the company provides production-quality NPU firmware for AAL2, AAL5, IPv4, and IPv6. The nP3665, the new NPU, derives from the nP3700 family and most closely resembles the nP3705. Like that device, the 3665 integrates a single 700MHz packet engine, based on AMCC's well-established nP architecture, and a traffic manager. The 3665 has fewer interface options than the 3705. It sports a GbE and a SPI-3 port along with either a second GbE port or a Utopia-2 interface, omitting the 3705's options for supporting SPI-4.2 and a dozen Fast Ethernet interfaces. AMCC rates both chips at 2xOC-12. The 3665 is sampling now with production expected by the end of the year. The Amur framer has already been introduced and is also positioned as a companion to the 3705 for multiservice access applications. Amur connects to the 3665 using SPI-3. On the line side, it connects to either one OC-12 or four OC-3 interfaces and can terminate up to 1,024 DS0 channels. The 440GRx is a PowerPC CPU operating at up to 667MHz. It adds DDR2 support, a pair of GbE interfaces, and security acceleration to the proven 440 core. AMCC's microprocessor product line is already commonly used in mobile infrastructure, providing AMCC a starting point for launching the new three-chip set. Competition in the cellular-infrastructure market is heating up. Rival Agere used Globalcomm to announce another mobile-infrastructure design win at NEC, and Wintegra is also targeting this market. AMCC distinguishes itself from competitors by offering a complete chip set, including the popular PowerPC CPU—a key technology that only Freescale can match. As cellular operators provide the triple play of voice, video, and high-speed data services and incorporate IP into their networks, NPUs become attractive because of their flexibility. —Joe Complete coverage of AMCC's nP3705 and Amur framer appears in our report A Guide to Access Processors.
Residential and SOHO gateways are evolving rapidly as home networking technology changes and service providers roll out faster broadband links with new services such as voice and video. Powering these gateways are inexpensive yet flexible processors packed with features to support these changes. Leading this market are companies such as Broadcom, Conexant, Texas Instruments, and Atheros that have combined their wireless, broadband, or VoIP technologies with basic RISC CPUs. Competitors such as Ikanos, Infineon, and PMC-Sierra are challenging the incumbents with new designs that have integrated voice processing and dedicated packet processing to handle faster networks. The result is a rich array of options from several new and established vendors. A Guide to SOHO Gateway Processors provides an in-depth look at the products and vendors for this market. Building on our "Guide to Communications Processors" published last year, this report provides extensive coverage of processors with integrated DSL, cable, 802.11, and/or voice functions. It also provides background on this market and helps you sort through the plethora of CPE and networking technologies. This detailed report, packed with valuable information, brings you up-to-date on the latest developments in broadband and home networking. It can provide the analysis and product comparisons you need to help choose a supplier or partner in this field. Order by June 30 to take advantage of a special prepublication discount. For more information on this report, visit our web site. Did you know that The Linley Group now publishes a newsletter focused on semiconductors for consumer electronics? Here are headlines from the latest edition of Linley on CE:
To receive The Linley Wire via e-mail, click here
|
