========================================================
TEST CENTER REPORT InfoWorld.com
========================================================
Monday, July 21, 2003
Advertising Sponsor - - - - - - - - - - - - - - - - - -
TECHWORLD: THE NEW WEB SITE FOR UK IT PROFESSIONALS
Techworld is IDG's new Web site for the IT professional.
It has been set up with one aim in mind; that of making the network and IT manager's job easier. It is written by people with experience of running networks, people who are aware of the technical problems that are thrown up in the course of the working day. People like you in fact. Let us know what you think at techworld.com
- - - - - - - - - - - - - - - - - - - - - - - - - - - -
INTEL TURNS 35: NOW WHAT?
By David L. Margulius
Posted July 18, 2003 3:00 PM Pacific Time
For many in the computer industry, the title of Intel Chairman Andy Grove's 1996 book Only the Paranoid Survive came to symbolize the relentlessly competitive attitude that helped make Intel a dynastic success. Over three decades, Intel's fanatical focus on design and manufacturing excellence, and its big bets on blockbuster products have enabled it to dominate the microprocessor business and become a Silicon Valley legend.
But today, on the eve of Intel's 35th anniversary, a more apt title might be Only the Diversified Survive.
Faced with softening demand for faster microprocessors, the company founded by Robert Noyce and Gordon Moore in July 1968 is revising its playbook. Instead of betting heavily on a single next-generation product as it often had in the past, Intel is spreading its bets among several emerging markets and technologies, including 32- and 64-bit microprocessors, mobile device chip sets, and a raft of futuristic innovations.
"The good stuff is incredibly good," says Martin Reynolds, a Gartner vice president and research fellow, lauding the company's legendary 32-bit Pentium microprocessor products. "The challenge is they struggle to have anything other than processors be successful. It's a company that's trying to find that next big thing."
Of course, Intel is still one of the industry's strongest companies. Though revenues have been flat at $26 billion to $33 billion over the past five years, the company has $12 billion in cash in the bank and 13 state-of-the-art fabs (semiconductor fabrication facilities) worldwide, and it will spend roughly $4 billion on R&D this year, according to Intel's CTO Pat Gelsinger. The company is also poised to benefit from major enterprise trends such as the migration away from Unix to Linux on Intel.
Expanding Moore's Law
In a way, the enormity of Intel's success has created its current dilemma. From its scrappy start by a handful of Fairchild Semiconductor veterans and through numerous market transitions, the company established a pattern of quickly scaling up to massive production capacity for each new semiconductor product. It optimized yields, speeds, and quality, giving it a major cost advantage and making it the only vendor that could deliver huge quantities of the latest chips to high-volume OEMs.
Along the way, Intel established a culture of thinking big and betting big. "There's no opportunity to sit back and rest on your laurels," said Moore, the company's chairman, in a 1993 Intel brochure. On the contrary, the company continually invested in new factories and manufacturing processes that were the envy of the high-tech manufacturing world. "Intel has done a great job in process technology, and that's surely been fundamental to their leading the industry," says Michael Dell, CEO of Dell Computer.
But the success model has depended on an ever-increasing
demand for microprocessor power, and today that demand seems
to be flagging.
"They've hit some kind of limit in terms of how much performance they can sell," explains Gartner's Reynolds. "Processor volumes are continuing to grow, but the average selling price is dropping at the same time."
Reynolds partly blames a lack of CPU-hungry software. "We haven't seen major advances in software performance demands for several years now ... to some extent Intel is still waiting for a new generation of software to come along," he says.
In response to the demand slowdown, "Intel is deliberately putting on the brakes of tech innovation" by slowing the frequency of new technology releases, says Peter Kastner, chief research officer at the Aberdeen Group. "The number of speed bumps per year has decreased."
That's not how Intel's Gelsinger sees it. "Intel has made a huge bet that we can keep the pace of Moore's Law going," he explains, referring to the famous 1965 prediction by Moore, Intel's co-founder, that the number of transistors per integrated circuit and, therefore, processing power would expand over time. Over 30 years, this has roughly equated to a doubling of the number of transistors every couple of years.
Intel's strategy, according to Gelsinger, is two-pronged: It both extends and expands Moore's Law. The effort to extend Moore's law (or keep the speed increases coming) involves both new materials (what do we need to add to silicon to allow it to keep getting smaller?) and new ways of building transistors, including Tri-Gates, carbon nanotubes, and new packaging types.
Intel engineers are already working on 30-nanometer CMOS transistors, much smaller than the 90-nanometer transistors the company plans to start producing in volume later this year. And Gelsinger believes that adding threading capabilities, or the ability to process more instructions in parallel, will result in additional performance gains throughout Intel's product lines while lowering power consumption requirements.
But even more important is the effort to expand Moore's law -- to address new markets by integrating additional capabilities on silicon chips. "We're going to be at billions of transistors per chip in the next year or two," Gelsinger says. "What else can we do with them?" (Intel has a set of research programs to explore potential semiconductor products for wireless, optical communications, and even precision biology applications.)
Potential applications include using the silicon chips to analyze protein and DNA at the molecular level, for example. Or using the mechanical properties of silicon for cooling or display applications. Or building sensor networks consisting of miniature computers and radio transmitters combined on a single chip that can monitor everything from health care environments to highways and homes. Gelsinger notes that although the bulk of Intel's R&D takes place inside the company, Intel also has 400 active grants in research institutions around the world and investments in over 500 startup companies through its venture arm.
Wireless Push
By far, the biggest market Intel hopes to expand into is wireless, and the company is pushing hard to create new chip set products for mobile phones, PDAs, and laptops. "We've dubbed our program 'Radio Free Intel,'" Gelsinger says. "We want to put radios into every chip." The company sees devices based on the convergence of computing and communications as an opportunity to build another huge franchise by selling the core hardware building blocks at high volume to device manufacturers.
In addition to a big recent marketing push for its Centrino 802.11 chip set for laptops, Intel is heavily touting its Personal Internet Client Architecture, a framework for integrating a microprocessor, GSM/GPRS baseband radio, and flash memory into a single chip, which will help lower the cost and power consumption of next-generation multimedia cell phones and PDAs. Currently most PDAs and cell phones have separate chips for these functions.
Intel, which has traditionally sold flash memory for small mobile devices, currently holds about a 50 percent worldwide market share for flash memory, according to Tony Sica, a vice president of Intel's Wireless Communications and Computing Group. In 1998 it acquired Digital Equipment's StrongARM portable wireless processor line, which was used in most PocketPC devices before being replaced recently by Intel's own next-generation XScale processor product line.
But the big push has been to provide an "all on one die," integrated single-chip solution, explains Sica. Code-named Manitoba, the PXA800F cellular processor product was introduced in February and is currently shipping in sample quantities.
Sica notes that Intel has "probably invested more than $2 billion in acquisitions" to become a player in this emerging wireless market where creating demand is still the key challenge. "The first thing you think about," Sica says, "is how do I get enough customers to buy all this stuff?"
Processor Product Lines
Will Intel be able to execute on these new market
opportunities to keep its fabs humming at capacity? The
company has had a mixed record with new businesses in the
past. It largely abandoned such markets as Web hosting, consumer devices, and DRAMs but also picked some winners such as chip sets, motherboards, flash memory, and networking components.
Intel may actually find its biggest diversification challenge in its core processor product line, where the market is segmenting and the company is trying to make inroads into the high-end server market against RISC architecture chips. While it has continued to win low-end enterprise server business by squeezing dramatic performance gains from its 32-bit processor (Pentium and Xeon) product lines, it has failed so far to significantly dent the high-end server business with its 64-bit Itanium I and II chips launched in 2001 and 2002.
The company has executed well on its 32-bit products, proving that the Pentium 4 architecture is scalable, and adding more than a gigahertz of speed improvement to the 32-bit line over a one-year period. It has integrated an on-chip memory cache into the XeonMP, which is important in larger-scale four-way servers that handle database and large application workloads. "Pentium 4 and Zeon products through the '90s gained good market share in the volume server market," says John Crawford, an Intel fellow in the company's Enterprise Platforms Group.
Yet the 64-bit Itanium, which went late to market and was updated with the Madison release on June 30, has been considered a relative disappointment due to its underwhelming performance. "It has taken a fair amount of shots," says Gartner's Reynolds, who believes Itanium is slowly ramping up in the market and might get a boost if Sun decides to use it to replace its own proprietary Sparc chips. "One has to view Itanium as a long-term play," he says.
But the competition claims customers have snubbed Itanium because Intel chose to develop a new instruction set that was not backward compatible with software written for its 32-bit architecture products. "Intel has made some major miscalculations where they really didn't listen to what the end-users want," says Kevin Knox, director of enterprise business development at competitor AMD. "Customers are saying we don't need more performance, we need more stability."
Intel's Crawford believes "a prolonged transition" to 64-bit chips will be led by the high-end server market, which requires the ability to address large amounts of memory. He thinks improvements in 32-bit products "will probably satisfy the volume server market for some time." And with the exception of some high-end computational workstations, Crawford doesn't think desktop machines with Itanium chips will catch on. "It's really hard to see why anyone would need 64 bits in the forseeable future," he says.
Can Intel generate enough success with its 64-bit microprocessors and other emerging products, such as wireless and new silicon applications, to make a strategy of "only the diversified survive" work? Will the company's legacy of great chip design and large scale manufacturing excellence remain relevant?
"In this industry, fortunes can change rapidly," Gordon Moore said in 1993. The question for Intel at 35 is whether it can innovate rapidly enough to sustain its fortune as the world's dominant semiconductor company.
David L. Margulius is a contributing editor at InfoWorld.
- - - - - - - - - - - - - - - - - - - - - - - - - - - -
To join, or start, a discussion on this or any
IT-related topic, please visit our InfoWorld forums at idiscuss.infoworld.com. Here you can interact and exchange ideas with InfoWorld staff and other readers.
- - - - - - - - - - - - - - - - - - - - - - - - - - - -
QUOTE OF THE DAY:
"The differences between spam and conventional e-mail are pretty subtle. Sure, you can screen out all messages with multiple exclamation marks, but then you might miss your brother-in-law's frantic note headed, 'Egad -- we're having triplets!!!'"
--InfoWorld CEO and Editorial Director Kevin McKean
infoworld.com
- - - - - - - - - - - - - - - - - - - - - - - - - - - -
SUBSCRIBE/UNSUBSCRIBE/CHANGE E-MAIL
To subscribe, unsubscribe or change your e-mail address
for any of InfoWorld's e-mail newsletters,
go to:http://www.iwsubscribe.com/newsletters/
To subscribe to InfoWorld.com, or InfoWorld Print,
or both, or to renew or correct a problem with any
InfoWorld subscription, go to iwsubscribe.com |
