A few thoughts on IBM, INTC, and the network processor mkt -
While neither company's recent announcements are a major threat to any established PMC market (although some of the processors claim to be programmable into a T1 interface, an HDLC engine, and can perform various SAR and Sonet framing actions), of the two, IBM seems more of a potential competitor long-term.
Intel has combined Level One, Softcom, and DECs StrongARM technology into a rather disjointed strategic effort. INTC has a history of failing at anything other that its core mkt (witness its recent termination of its graphics chip biz). In addition, while they have introduced a network processor, they are less clear on a fabric switch offering. It is very important to link these two and many times, the 'tightness' of the processor/fabric architecture is what makes or breaks design wins. LEVL was fading as a T1/E1 vendor and was decent competition for BRCM in the LAN. INTC has proven to be a LAN chip and switch vendor of note, but the WAN has been very elusive. I simply don't see their balance sheet helping them as this is overcome by their lack of expertise and engineering talent in this area.
IBM Microelectronics is a viable force. They have mostly been a leader in ASICs and other custom chips (as well as their foundry biz). As an example, not many people know that for all intents and purposes, IBM ASICs (five to be exact) are the foundation of the Juniper M40. The ASIC biz is of course, much different from the standard products biz. And here, IBM is very behind. They announced a processor to go with their production Prizma switching fabric. Thus IBM should have much greater compatibility between the two stages.
The important to thing to note is: THESE OFFERINGS ARE NOT NEW TO THE MARKET. There are many startups and new companies that have been supplying these products (or readying production) for well over a year. These include MMCN, BRCM, MAKR, C-Port, Agere, SiTera, VTSS (XaQti) and Power-X. Some of these companies offer both the processor and the fabric, some only one.
The difference btw is that the processor stage (encompassing central microprocessors and from 6-17 RISC and other supporting port-level or channel processors) takes the bits from the input ports and the phy layer, and processes them before they are sent to the crossbar switching fabric. What processing? Table lookup, header analysis, checksum analysis, input queue and buffer mgmt etc.).
IBM and Intel are trying to port established computer processors and architectures, and mold them into comm-IC offerings. C-Port, SiTera and others are building comm products from the ground up. In either case, their market is nascent at the present time and absolutely no threat to current PMC revs. Will they be competitive in the future? - possibly. Will PMC be standing still? Not with 20%+ R&D keeping them busy.
I continue to believe that PMC has more growth ahead than its had in the past. Broadband, core switching, xDSL (not just Vortex, but all the Freedm chips and ATM chips needed in various DSLAMs), continued growth in SONET (nice piece today by Morgan Stanley on SONET) and of course ATM (still a long way to go as witnessed by this recent article by systems vendors and service providers: telecommagazine.com, let alone whatever else these guys have up their sleeves, should provide great momentum for many years to come.
Bulldozer |
