Although the chip will continue to appear in notebooks and low-end servers, the Pentium III will essentially be phased out of the desktop market by the end of the year, said Anand Chandrasekher, vice president of microprocessor marketing at Intel.
A new version of the Pentium III, based on the 0.13-micron manufacturing process and code-named Tualatin, is still coming out toward the third quarter, but it won't be actively marketed for desktops.
"The opportunity may be there for some speed bumps, but not for the desktop," he said this week at the Intel Developer Forum in San Jose, Calif. "We don't see a big play for our 0.13-micron Pentium IIIs in desktops."
The Pentium III is Intel's best-selling chip.
Intel can't actually stop PC makers from putting Pentium IIIs into desktops. However, the company can take actions to encourage PC makers to put their energy into the Pentium 4. In the past, Intel has accomplished this by not cutting prices or not increasing speeds on the chips it wants to phase out.
The Pentium III's demise on desktops represents an acceleration of the company's efforts to proliferate the Pentium 4 and creates a number of implications for the marketplace.
For one, if Intel succeeds, it could mean a quicker adoption of the Pentium 4 in corporations. Intel hasn't made a big push in the corporate market yet. But Dell Computer is expected to come out with an OptiPlex corporate desktop with a Pentium 4 next week.
The Pentium 4 could also inherit a fairly sizable share of the consumer market. By the fourth quarter, consumers will be able to buy Pentium 4 computers with a monitor for $1,000, Chandrasekher said. By 2002, Intel's desktop strategy will effectively consist of two chip families: the Pentium 4 and the Celeron, which is aimed at the bargain segment.
Intel released the Pentium 4 in November but, until recently, had soft-pedaled it in the consumer market. "There definitely seems to be an accelerated push on Pentium 4 comparing Jan. 1 to now," said Dean McCarron, principal analyst at Mercury Research.
Then again, Intel's strategy could blow up in its face and open an opportunity for Advanced Micro Devices to finally get into the corporate market.
The potential hitch is related to chip designer Rambus. Currently, Pentium 4 computers need to incorporate Rambus memory. Not only is Rambus memory more expensive than other types of memory, but corporate buyers still recall problems in 1999 when Pentium III computers were recalled or delayed because of problems that somehow were always related to Rambus.
Chipsets that allow computer makers to build Pentium 4 computers sans Rambus are coming toward the midyear. And prices for Rambus memory are going down. Still, memories are long in the corporate world.
"The corporate IT Nazis are absolute fanatics about having a solid, reliable platform," McCarron said. "The experience with the 820 chipset (Intel's Rambus chipset in 1999) put a lot of them off on pushing Rambus in the corporate market."
This "certainly helps" AMD, he added.
How the market will react to the Pentium 4 is impossible to tell because of a number of variables and contingencies.
Chandrasekher, for example, said that the downturn in the economy indirectly helps Intel in the corporate market. Computer makers generally want to work with fewer, rather than more, types of processors because it's less expensive. Corporate buyers face a similar reality: Uniformity in the installed computer base means less overhead for support.
"You have to bet your research and development dollars on the platform that is going to win," Chandrasekher said. "In times of economic slowdown, corporations will do belt tightening."
Slow but steady
Last year, AMD CEO Jerry Sanders predicted that the slowdown in PC buying could delay AMD's plans to land a design win with Hewlett-Packard, Compaq Computer or IBM for a corporate computer by a few months. Nonetheless, the company has made slow, yet steady, progress in landing its Athlon chips into computers for small businesses.
"We're definitely seeing momentum, not just in the U.S. but overseas as well," said AMD spokesman Ward Tisdale. In addition, the company will soon release the first version of Athlon for notebooks and a chipset for making dual-processor servers.
AMD and Intel will continue their fight in the consumer market.
Rambus also remains a wild card. Executives from Samsung, Toshiba and Elpida said at the Intel Developer Forum that manufacturing costs should continue to decline this year as volumes increase. By the end of the year, Rambus will represent 60 percent of Toshiba's output, according to the company, up from 20 percent today.
Rambus could also begin to look better as more computer companies begin to work with double data rate (DDR) DRAM, a competing type of memory that is being matched with the Athlon. Although many of the painful lessons learned with Rambus will make it easier to adopt DDR, computer companies still face a learning curve.
"There are problems with putting DDR into systems," said Dieter Mackowiak, senior vice president of sales and marketing at Samsung.
The lower volumes of DDR also mean that it won't be that much cheaper than Rambus.
"If the price point delta is just 5 to 10 percent, customers will go to Rambus," Mackowiak said.
Still, reducing Rambus manufacturing costs won't be easy, said Peter Glaskowsky, an analyst at MicroDesign Resources. And the efforts to match DDR with the Pentium 4 are unlikely to be attractive to corporate buyers until next year.
If Intel's plan goes awry, Glaskowsky predicts, the company could end up making chips for Pentium III computers "for another year."
McCarron is even more direct.
"They will need a Pentium III platform longer than they are saying they are going to need it," he said.