Network Processor Revival

Network processors are so hot, even the dead products are doing well!

Okay, that's an exaggeration -- and whether a $120 million year is "hot" is debatable, considering some half-dead companies make more than that in a quarter. Still, customer wins with Tier 1 players were enough to increase market share for Vitesse Semiconductor Corp. (Nasdaq: VTSS) -- to 11 percent from 8 percent in 2003, according to figures released this week by The Linley Group -- even though the company has cut future network processor development.

"They've got a couple of Tier 1 design wins," says Bob Wheeler, a Linley analyst. "You figure those have a shelf life of maybe a couple of years, so Vitesse could keep bumping along at this level."

Other big communications chip businesses appear to have stabilized. IBM Corp.'s (NYSE: IBM) PowerNP line earned flat revenues -- albeit while losing market share -- even though IBM halted product development early in 2003. Hifn Inc. (Nasdaq: HIFN), which acquired the PowerNP, reports the parts did $2.1 million in the past quarter.

Wheeler says Hifn is even reviving the NP4GX, a next-generation chip IBM had canned. "It won't be out until 2005, but at least now that product line has a roadmap," he says.

Even a few OEM projects are coming back from the dead. "We're getting calls from people we had design wins with a year or two ago that we had written off," the Intel spokesman says.

Overall, network processors are blossoming after two years of winter frost. The market should grow to $120 million in 2004 from $85 million last year, Wheeler says.

In terms of revenues, Applied Micro Circuits Corp. (AMCC) (Nasdaq: AMCC) remains the king of network processors, with Intel Corp. (Nasdaq: INTC) in second place. Here's how they stacked up:

Table 1: Net Processor Market Share by Revenues
2003 share 2002 share
AMCC 39% 38%
Intel 16 11
IBM 13 19
Motorola 12 15
Vitesse 11 8
Agere 6 9
EZchip 2 < 1
Source: The Linley Group

AMCC led in last year's market study from Linley as well, but the Intel spokesman says his company isn't sweating over the stats. Intel placed No. 1 in a similar survey by In-Stat/MDR, the difference being that AMCC's nP5700 traffic management chips are included in Linley's survey.

Among the disappointments, Motorola Inc. (NYSE: MOT) managed to ship a long awaited traffic manager, but its market share dropped. "Relative to AMCC and Intel, they just haven't delivered significant new products, and that's starting to catch up with them," Wheeler says (see Motorola Stuck on C-Port).

— Craig Matsumoto, Senior Editor, Light Reading

xbar 12/5/2012 | 1:57:04 AM
re: Network Processor Revival The market is tanking; that is why both IBM and Intel are exiting. AMCC still cannot understand that the light in the tunnel is the train approaching:)...
dash_riprock 12/5/2012 | 1:57:01 AM
re: Network Processor Revival The market is tanking; that is why both IBM and Intel are exiting
IBM is out. But Intel ? where did you get this from ?

sigint 12/5/2012 | 1:56:46 AM
re: Network Processor Revival xbar: (paraphrased)
Intel is exiting

that's BS. the Network processor group at Intel has in fact broken even. took a lot of cost-cutting, but it finally did happen.
edgecore 12/5/2012 | 1:56:38 AM
re: Network Processor Revival My 0.02,

The NPU market, it is a bit of a running joke (for the last 3 years at least). Many questions seem to not be clear to me:

How big is it really? How much money can be made?

The market may be starting to get interested (the design engineers this time, not just the marketing folks)...but again, how much has been invested and how potential ROI is really out there?

Think of the huge cost/resource burdens...given the fact that AMCC and Intel not only need to design and bring to market complex NPU's but you also need about 12 million sw engineers on staff to bring to market a useable and productive Linux and VxWorks based development tool (or workbench) to write the actual microcode that gets to run on the forwarding engines inside the NPU. Hardware companies have show time and again that it is extremely challenging to be a good hw and sw company at the same time.

The HW and sw costs seem astronomical given the small potential market (in relation to the sunk costs so far by major NPU players).

What about the actual ecosystem (good marketing word) around the development of micro apps for the NPU's, how will that work? Will we see miilions of Teja's spring up and provide value to Tier 1's, will AMCC and Intel convinve the world that this is a service better provided by them (for a huge fee). Where do the Cisco, Lucent, Nortel L2/L3 experts get to play in this MPU equation/ Do they need to learn brand new tools?

If AMCC has a strong lead, then they may be fine, but does that make Intel, the only other that can afford to stay in the game long enough to see the day where they make profits?

...OK, now I am just rambling, but I feel this has been very good theropy, thanks for listening!


hoopmeister 12/5/2012 | 1:56:30 AM
re: Network Processor Revival With the maturity of router design and potential flexibility of the NP in processing high layer functions, there maybe a bigger role for the NP. What that means is that folks working high layer applications can use NP and application specific H/W co-processor to boost the processing performance. In the early days of NP, this was not ready.
Anyhow, does anybody have insights with Xelerated (www.xelerated.com), a swedish company that claims the lead in NP with so called synchronous data flow architecture. It says its X11 has exceptional processing power (40Gbps) with only 11 watts power consumption. Will this company survive?
thud 12/5/2012 | 1:56:24 AM
re: Network Processor Revival
I am surprised that Bay Micro didn't make the list. They have had a product in the market for a while, right?

niblick 12/5/2012 | 1:56:17 AM
re: Network Processor Revival I think the future for NPUs is bright. With Carriers moving to more packet based (IP/MPLS) networks you need flexibility to track standards and changing requirements. Flexibility is important as these packet technologies are not mature yet to be baked in an ASIC. Therefore any flexible engine such as NPUs will be crucial to build systems in future.

While I agree that past NPUs were immature in performance and in SW tools they are getting better by the day.

I also think for NPUs to become mainstream the NPU vendors should just focus on getting the silicon right and let other vendors develop SW tools and applications. Similar to the PC world.

npufx 12/5/2012 | 1:56:10 AM
re: Network Processor Revival I'm a consultant that has been writing applications for AMCC NPUs for almost 4 years now. From "the trenches", I'd agree with much of what edgecore has to say regarding the software development issue. IMHO the tools (simulators, debuggers/etc) have improved over the years - that allows me to develop solutions faster/cheaper. The real question for chip companies (like AMCC) is whether or not they want to enter the professional services business and provide truely customized solutions for their clients. Sample code is just that - and clients are typically more fustrated trying to figure out the vendors microcode and try to make changes to make it fit into their system. It's far easier (and more cost effective) to bring in an expert to help them develop a custom solution and realize the full potential of their application on the NPU.

I believe that chip companies should focus their resources on the tools, developing app notes, and an ecosystem that facilitates 3rd party software vendors/consultants.

Just my 0.02, a bit bias I'm sure since this is how I make my money :-)

-- npufx

Sign In