Carl, you are FOS. Full of Spin, that is. A more credible report of the meeting mentions nothing about Micron and the other Dramurai walking out of the Intel conference a day early. While I may not agree with the conclusions of the Insight 64 analyst, this report appears to be balanced - in direct contradistinction to your incessant FUD.
Intel gathers memory makers to boost Rambus plans
By Ian Fried and Stephen Shankland
Staff Writers, CNET News.com
May 18, 2000, 5:30 p.m. PT
URL: news.cnet.com
Intel held a closed-door meeting with memory makers in Arizona yesterday to discuss the status of controversial Rambus-based memory.
The chip giant confirmed a meeting took place but would not comment on whether any new ground was broken. Intel, long an ardent proponent of Rambus' proprietary standard for next-generation memory, is counting on an increase in Rambus-based memory output to support the launch of its Willamette processor later this year.
"It was a meeting with Intel customers," said Intel spokesman Dan Francisco. "It was invitation-only."
A Rambus representative had no comment on the gathering. Rambus doesn't make memory itself but rather licenses its designs to chipmakers that make high-speed memory known as DRAM.
Intel has said its companion chip set for Willamette-based desktop computers will only support Rambus memory and has been trying to spur memory makers to produce more of the fast but costly chips.
Intel's need for Rambus-based memory intensified as Intel disclosed a glitch with a translator hub that was supposed to allow its 820 chipset to use standard memory. That means that the chipset is currently good only for Rambus-based systems.
Intel also has offered to replace standard memory with Rambus chips in hundred of thousands of existing 820-based motherboards found in Pentium III systems. Analysts have questioned whether Intel will be able to scrounge up enough costly Rambus chips if vast numbers of customers take advantage of the recall offer.
As for future systems, Intel spokesman Howard High said he recognizes that it's tough to convince people to move to Rambus memory.
"It's a chicken-and-egg problem," High said. Until PC makers buy Rambus in quantity, the memory will be expensive to manufacture. But until PC manufacturers demand lots of Rambus chips, memory makers won't make them in high volumes at low cost.
And Intel is in a somewhat awkward position, since it controls neither the memory suppliers nor the PC makers. All it can do is talk to both sides--and give money to memory makers in the form of hundreds of millions of dollars in equity investments.
Today, many of the memory makers were in San Jose at a forum for Double Data Rate (DDR) memory, a more evolutionary technology seen as alternative to Rambus.
Insight 64 analyst Nathan Brookwood, who attended the DDR event, said the memory makers' decisions will only become tougher as memory chip supplies tighten later this year.
Memory makers will be faced with the choice of starting wafers for traditional memory or DDR--chips they know they can sell and make money on--or producing Rambus-based chips that could sell for a higher price but where demand is more uncertain.
"That's what they are basically trying to come to grips with," Brookwood said.
Meanwhile, time is getting tight for Intel. Without a faster memory system to replace existing synchronous (SDRAM) memory, speed gains in new Intel chips will be wasted, High said.
Intel wants to ensure system balance, in which faster components of a computer aren't held up by slower parts. Intel maintains today's SDRAM is too slow for an 800-MHz CPU. "This is the crossover point," High said. "The disparity only gets greater, not less" as CPUs get faster. |
