Carl:
Your take on:
john
electronicnews.com
Rambus Looks to Non-PC Markets
Its 1066MHz RDRAMS could reduce its need for support from Intel
By Paul Kallender
The Rambus Inc. roller coaster took another twist last week when the company said it intended to break out of the PC memory market and into consumer applications with its new 1066MHz direct Rambus DRAM (RDRAM).
The announcement comes just days after Intel Corp. officially abandoned its requirement that RDRAM would be the sole memory for its Pentium 4. However, the 1066MHz part has done little to spur confidence that Rambus can engineer RDRAM volume in any form at all next year.
Aimed at consumer, graphics and communications applications—and based on a simple bin split of the PC 800MHz RDRAM—Rambus sold the speedy 1066MHz part as a win-win situation for makers to find new markets while ramping 800MHz RDRAM. In this case, "bin split" means testing the same parts to see which ones can run successfully at higher clock rates.
"Nobody should question that we want to grow our customer base. We are doing the right thing," said Avo Kanadjian, Rambus' vice president of worldwide marketing. Big guns Samsung Semiconductor Inc. and Hyundai Electronics as well as test equipment house Advantest Corp. came out in support of the new chip. So Kanadjian believes the 1066 will have an infrastructure in place and in 2001 a market to raise the RDRAM.
Rambus is asking for 1066 MHz production by the DRAM makers to show that RDRAM is scalable and as a transition to lowering its price.
"The implication is that the 800 (MHz version) is coming on line," said Bob Merritt, analyst for Semico Research, Phoenix.
"Exactly," said Kanadjian about DRAM makers' ability—and willingness—to ramp next year. To support his optimism, Kanadjian cited figures from Samsung saying the world's largest memory maker thinks it can raise June's 60 percent yield to 80 percent on the 800MHz parts by December 2000. Along with Samsung, other RDRAM suppliers seemed likely to achieve the same yield range. Micron said it would be coming on-line later this year, together with a mystery company next year.
Avo Kanadjian, vice president of worldwide marketing, Rambus Inc.
"Production volume can be made to meet required demand as soon as that demand is ready," Kanadjian added.
Further, Intel said that it believed producing the 1066MHz chip will not crimp 800MHz production.
"To the best of our knowledge, it will have no impact," said Intel fellow Pete MacWilliams.
But demand, or lack of it, will remain a crucial issue that the 1066 will at best only marginally assuage, said observers.
Turning to the bigger question of the 800MHz part's general availability, MacWilliams would only say Intel believes that makers are through the technical difficulties that have made the 800MHz difficult to produce. But he would only predict "some volume" next year.
"That's why we are adding the PC-133. It would be foolhardy to put all our eggs in one basket," said an Intel spokesman. "We do strongly believe that RDRAM is the primary memory solution for performance desktops."
The best way for RDRAM makers to reduce costs would be to invest more money in production, said Bert McComas, founder and principal analyst at InQuest. Which leaves the question of what sort of applications market the 1066 MHz device will generate. Kanadjian said he is confident OEMs will want to design-in the chip because of DRAM makers' strong show of support for it.
"They've got to be bullish about next year; this year was the year they were supposed to go into volume production," said Jim Feldhan, president at Semico Research. "If they can garner demand … then their likelihood to penetrate the PC market is higher."
But garnering that demand will be problematical, he said. In the short term, OEMs will take six months to come on-stream.
So, then who will adopt the technology? "There's not a lot of activity," Feldhan said of games designers' interest in the 1066MHz chip. "Part of engineering a market for a whole new architecture is that you have to create demand."
Turning to overall demand, Semico is just now considering downgrading its 73 million unit sales estimate for RDRAM in light of RDRAM's June performance, Feldhan said. As for 2001? His estimate is 81 million RDRAM parts. "It's pretty small," said Feldhan.
"Ultimately, all strategies end up leading back to the PC market, it's just another road to get them there," he said. |