x
Optical/IP

Intel: Comm Chips Commoditizing

DALLAS -- The telecom equipment business will eventually mirror the PC hardware makers by moving toward off-the-shelf chips and components, according to Intel Corp.'s (Nasdaq: INTC) Communications Group VP and chief technology officer Eric Mentzer.

"Building chips is our business," says Mentzer, who spoke at a luncheon organized by the Technology Business Council on Friday. Rather than differentiating from one another by having specially designed chips, Mentzer says equipment vendors "will have to give service providers something they can bill for and make money on."

Mentzer's remarks come at a time when telecom equipment vendors often design their own custom ASICs (application-specific integrated circuits) to consolidate a variety of different data processing functions on a single chip. To him, this harkens back to when PC makers used to differentiate their products on the number of clock cycles, the time that elapses between operations in a computer's main memory. "Eventually, you saw an entire industry learn to stand on the shoulders of the semiconductor business," Mentzer says.

What may intensify that change is Intel's new chip-making process that makes use of 300-millimeter wafers. On these new wafers, Intel can print about two-and-a-half times as many chips as it can on the standard 200mm wafers used today. Intel's new manufacturing process also uses 90-nanometer (0.09 micron) design rules instead of today's 130nm (0.13 micron) rules, Mentzer says.

This new manufacturing technology is important because it means Intel will soon be able to build all flavors of communications chips. The micron process refers to the size of the lines that are etched onto a chip. A fabrication plant with a 0.09-micron process can create chips that are smaller, faster, and consume less than chips made by a plant with a 0.13-micron process. Mentzer also says Intel will be able to make digital and analog chips, which have different voltage requirements, using the same process.

Intel's manufacturing advance theoretically will cut down on the number of plants needed to produce communications chips. This will bring down chip prices even while the functional speed of communication chips continues to increase.

Of course, Intel will have its work cut out for it here. In the past, the networking and telecommunications chip space has been dominated by Applied Micro Circuits Corp. (AMCC) (Nasdaq: AMCC), Cypress Semiconductor Corp. (NYSE: CY), PMC-Sierra Inc. (Nasdaq: PMCS), and others.

Also, Intel's not alone on the road to 0.09-micron designs. Infineon Technologies AG (NYSE/Frankfurt: IFX), for example, expects to be working with 0.09-micron designs in the first half of 2003. "In terms of functionality, there is still a massive amount of things you can do at 0.13-micron," says Jack Basi, VP of marketing for Infineon's Catamaran unit.

The commoditization of telecom equipment components will begin at the network's edge and work its way to the metro core, Mentzer says, and it won't happen all at once, but over many years. "It's time for the Moore's Law revolution to happen in telecom. The telecom business hasn't been known for bringing out new capabilities at lower price points -- or eating their young."

Of course, there are hundreds of variables as to when and where this commoditization might happen. In well understood technology areas, such as the enterprise LAN, the equipment interfaces are standard and the technology could be commoditized more quickly, says Infineon's Basi. That's not the case in other areas, such as carrying Ethernet into the metro network, he says.

What's also up for debate is whether equipment companies such as Cisco Systems Inc. (Nasdaq: CSCO) will embrace Intel's off-the-shelf idea, especially given that two of the last six companies it bought were chip or chip components companies, as are many of the companies it has funded in the past two years.

Mentzer's remarks Friday were couched in a larger discussion about where the communications business is heading in the next decade or so. His vision of the networked home includes Internet access via optical fiber and a plethora of devices that use silicon and optics to process the torrent of bandwidth coming in (see Intel CTO: The Future Is Here!). "We'd better have optics on every single desktop if we're to keep our microprocessors fed."

Unlike most technology speakers these days, Mentzer spared his audience a recap of the telecom boom years or a textbook review of the market's crash: "I can only tell you two things about recessions: They end, and you can't save your way out of them. You need to invest in new technology."

— Phil Harvey, Senior Editor, Light Reading
http://www.lightreading.com Want to know more? The big cheeses of the optical networking industry will be discussing this very topic at Opticon 2002, Light Reading’s annual conference, being held in San Jose, California, August 19-22. Check it out at Opticon 2002.

Register now and save $500 off the registration fee. Just use the VIP Code C2PT1LHT on your registration form, and deduct $500 from the published conference fee. It's that simple!

Page 1 / 3   >   >>
photonsu 12/4/2012 | 10:03:34 PM
re: Intel: Comm Chips Commoditizing "Also, Intel's not alone on the road to 0.09-micron designs. Infineon Technologies AG (NYSE/Frankfurt: IFX - message board), for example, expects to be working with 0.09-micron designs in the first half of 2003. "In terms of functionality, there is still a massive amount of things you can do at 0.13-micron," says Jack Basi, VP of marketing for Infineon's Catamaran unit.

The commoditization of telecom equipment components will begin at the network's edge and work its way to the metro core, Mentzer says, and it won't happen all at once, but over many years. "It's time for the Moore's Law revolution to happen in telecom. The telecom business hasn't been known for bringing out new capabilities at lower price points -- or eating their young."

Is this "MOORE hipe"?
They talk as if they have made significant optics cost reduction, but what does 0.09um electronic features size have to do with optics cost?
Has Intel bought into that which they do not understand? Or have they made a significant breakthrough that allows them to reduce the area of an AWG to 0.09um x 0.09um.

Thought someone should ask!

Better go sell my Intel stock.
rickkepple 12/4/2012 | 10:03:34 PM
re: Intel: Comm Chips Commoditizing Think about it. Hummm... who is getting the margins in the PC business - is it Dell, Compaq or someone else?? Will any self-respecting systems engineer allow Intel 'inside' as a result?
light-headed 12/4/2012 | 10:03:33 PM
re: Intel: Comm Chips Commoditizing Systems already use cheap off-the-shelf commodity chips for MAC, PHY, Framers, Channelization, etc.

It will be a while before using it for all switch fabric, queuing and packet processing as the speed and functionality requirements are not being met by companies like intel at a cost that competes with ASIC. This intel CTO will be retired before we see a core router with an intel packet processor, scheduler, queue manager and switch fabric deployed!
wilecoyote 12/4/2012 | 10:03:32 PM
re: Intel: Comm Chips Commoditizing I don't know. I want to agree with you lightheaded but from everything I'm seeing, there's not a whole lot of real science being built in any of these categories, from independent chip vendors, that anyone's actually going to buy. So that would point to off the shelf beating NIH which no one can afford anymore in terms of price, development risk or time to market considerations. I don't think Intel has the portfolio yet but with their balance sheet they are in the cat-bird's seat. Guess what? They are one of only two outside vendors CSCO is considering for the network processor side of things, as an example. One of two. Scary, huh?
Joeboo 12/4/2012 | 10:03:32 PM
re: Intel: Comm Chips Commoditizing Phil, are you new here? Cypress??
DoTheMath 12/4/2012 | 10:03:32 PM
re: Intel: Comm Chips Commoditizing The commoditization of telecom equipment components will begin at the network's edge and work its way to the metro core, Mentzer says, and it won't happen all at once, but over many years. "It's time for the Moore's Law revolution to happen in telecom. The telecom business hasn't been known for bringing out new capabilities at lower price points -- or eating their young."

....

"I can only tell you two things about recessions: They end, and you can't save your way out of them. You need to invest in new technology."
-----------------------------------------------

This is an excellent statement of vision for this industry. I don't believe telecom is going to recover if the objective is to get customers to pay a lot more for bandwidth. The vision has got to be to offer a lot more bandwidth for what customers do want to pay, i.e about $20-50 a month (range of your ISP to your cable TV). This is $250-600 a year per household, and is can be very profitable, if only the industry structures itself to get this. The challenge is to offer progressively more every year but keeping the overall bill about the same.

I hope Intel will drive this vision relentlessly forward. Eventually service providers and equipment vendors have to embrace this in order to succeed.

redface 12/4/2012 | 10:03:32 PM
re: Intel: Comm Chips Commoditizing "They talk as if they have made significant optics cost reduction, but what does 0.09um electronic features size have to do with optics cost?
Has Intel bought into that which they do not understand? Or have they made a significant breakthrough that allows them to reduce the area of an AWG to 0.09um x 0.09um."

Clearly this article talks about the cost of high speed electronics chips which drive the laser modulators, etc. With smaller and smaller circuit dimensions, the operating speed of CMOS chips get higher and higher - approaching OC48 or even OC192. Remember that today's PC operate at 2.4 GHz clock speed? As a result, commercial CMOS technology (and Moore's law) can be used to drive down the cost of previously highly specialized and very expensive communication chips.

Of course this cost reduction only works for the high speed electronics components, not the DWDM optical components.

gea 12/4/2012 | 10:03:31 PM
re: Intel: Comm Chips Commoditizing Waitaminute...

"Optical" network systems are rarely only optical. Most, if not all, will hav some need for some form of SONET frame processing at least, if not integrate some SONET functionality. In addition, the term "optical" in the network space is almost meaningless now: take the Tellium switch fo example. It's called an "optical cross connect", but there's nothing optical about. That the perfect place where off-the-shelf chips will probably cause Tellium a lot of headaches (and I'm sure they're already thinking about looking at the off-the-shelf switch fabrics...they're "value added" is clearly in software and in thought-leadership a la UNI 1.0 and GMPLS.

Frankly, off-the-shelf chips in the networking space are almost at the point where it would be silly to develop one's own. That's much less true in the Metro Ethernet area, but in "optical" and SONET transport systems, I can't see any systems vendors keeping up. Remember...one small problem in devloping an ASIC and its back to the drawing board and at least 1 year...no systems startup can afford that, so off-the shelf is almost a must. (Of course there will be exceptions like Cisco, but that's almost off=-the-shelf as far as I'm concerned).
light-headed 12/4/2012 | 10:03:31 PM
re: Intel: Comm Chips Commoditizing Cisco will only use those chips in the enterprise gear... they will not appear in the GSR for a while and not in the packet processor or switch fabric of the GSR. High end routing and switching is not a commodity yet.

Using off the shelf chips makes it too easy for a startup to beat you on cost and provide more and better funtions... sure ASIC design is a pain but I don't know anyone who is willing to bet their company on all 3rd party chipsets today or in the near future.
photonsu 12/4/2012 | 10:03:31 PM
re: Intel: Comm Chips Commoditizing The point I was making is this is an optical web-site, and Moore's law doesn't yet apply to optics, which is where the cost is these days, and will for some time limit where optics fits in the network.
We all know Moore's law applies to electronics integration. Now we just need something to drive the investment in research in realizing the same gains for optics.
Page 1 / 3   >   >>
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE