Intel-Rambus Relationship Shows Signs of Trouble
The chip maker, which decided several years ago to throw its impressive weight behind the Rambus memory standard (RDRAM), has endured numerous problems over the past year, due to the complexities and high price of the technology (see
Barrett's statement was unexpected, says Avo Kanadjian, vice president of worldwide marketing at Rambus, the company that designed RDRAM technology. Rambus doesn't manufacture RDRAM, but memory vendors pay the company a royalty when they make and sell it.
"We were surprised by the article," Kanadjian says. Nonetheless, the company still anticipates high marks for RDRAM when Intel launches its Pentium 4 processor. "We are very excited about the Pentium 4 launch, and we think the performance gains will speak for themselves," he adds.
Intel executives didn't try to soften Barrett's comments. His statements merely reflect a viewpoint the company has come to acknowledge over the last six months, says Howard High, spokesperson. That said, Intel still believes RDRAM is the best memory technology out there, High adds.
"With RDRAM, we clearly believe the technology is wonderful in the sense of performance," he says. However, it is still too expensive, he says.
In 1996, Intel executives started looking for the next memory technology, High says. "We knew back then that we needed a higher-performance memory for the future," he says. Then, EDO memory was mainstream, and SDRAM was on its way up. Intel projected that once its microprocessors reached the 800-MHz to 1-GHz range, even SDRAM memory would become a bottleneck.
RDRAM seemed the clear choice, he says. Others felt the same way.
"A few years ago, RDRAM seemed the only real option for the P4," says Kevin Krewell, senior analyst with
But Intel eventually opted for RDRAM. The company expected the memory's early high prices to drop, reaching parity with SDRAM pricing while offering superior performance, High says.
Kanadjian disagrees with High on pricing. RDRAM should always cost more than standard SDRAM, although the difference will shrink, he says.
"It is a better technology," he says, and vendors deserve to earn a premium for it. But the premium remains too high today, he acknowledges. He blames the wild fluctuations of the SDRAM market. By the second half of 2001, RDRAM's premium will drop, Kanadjian. He expects it will cost, on an industry average, only about 20 percent more than the same amount of PC-133 SDRAM. A spot-check of retail prices today for a 128MB of memory RDRAM versus the same amount of PC-133 SDRAM shows more than a 100-percent premium. RDRAM should cost about the same as DDR as that memory ramps into the market, he says.
A striking example of the RDRAM problem's influence on Intel's business was the cancellation of the low-cost Timna processor.
Intel planned to use RDRAM with Timna, which was to have integrated graphics and memory controllers. But the company scrapped the budget CPU in September (see
Intel designed Timna to use RDRAM, but when pricing didn't drop, Intel pushed back the launch date to include a memory translator hub to work with cheaper SDRAM. When the company ran into problems with the MTH, Timna fell further behind (see
Intel initially decided to launch Timna six to nine months later than planned, High says. Without the RDRAM price problems, "that product could have been ready for its window of opportunity," High says. Instead, the market changed, and interest waned. Intel killed the project and cut its losses, he says.
Rambus also takes issue with Barrett's recent comments about legal disputes over royalty fees. The company has sued Micron and other chip makers, alleging patent infringement. Hitachi and Toshiba recently reached settlement with Rambus on the issue (see
"We hoped we were partners with a company that would concentrate on technology innovation rather than seeking to collect a toll from other companies," Barrett told the
Rambus has a different business model and view of the technology industry than Intel, High says. "Intel's thrust is we want to move the industry forward. We make money by selling more powerful processors." If people don't want faster chips, than we don't make money, he says.
Intel works to get new technologies accepted in the market, and then it simply works to outperform competitors that also offer that technology, he says. When it gets to a point that one is sacrificing industry growth for immediate profits it's counter to Intel's business model, he says.
"Litigation doesn't move technology forward," High says.
"What we have done in terms of intellectual property is no different from what Intel and others do," Kanadjian says in response. The company has merely asked for compensation for its patents, he says.
Justified or not, continued litigation by Rambus against memory vendors is not good, MicroDesign analyst Krewell says. "When a company sues its customers it is not a healthy relationship," he says. Trying to charge for existing technologies "smacks of greed," he says.