Rambus no longer makes Intel's eyes twinkle
Willamette to embrace DDR, Micron to push DDR
Analysis A senior Intel source told The Register in February that he was very surprised when Rambus became the unexpected star of its Developer Forum show. We can thank senior Intel VP Dr Albert Yu, however, for putting the spotlight on the amazing dancing Rambus. Yu, demonstrating the Willamette processor at IDF, unequivocally stated that Intel's up-and-coming IA-32 architecture was inseparable from Rambus memory. Although he didn't burst into wild eulogies about Rambus, he said that it was the only memory type suitable for, so-to-speak, Intel's prince of IA-32 chips. As the swan is to other birds, as the sturgeon is to fish, as a king is to men, as musk is to perfume, and as the head is to limbs, so is Rambus to double data rate (DDR) memory, he might have said, if he had let poetry engulf his mind. Perhaps thankfully for his audience, he didn't. Yu's unequivocal support for the Rambus-Willamette dream ticket produced rhapsodies on Wall Street that very evening. The Rambus share price (ticker RMBS) twinkled brightly in the New York firmament, and once more became the darling of the brokers. Over the next day or two, them that had ears with which to hear, heard a very different story both from Yu and from other Intel executives. It transpired that while Willamette was yoked to Rambus, its elder brother, Foster - intended for workstations and servers, would instead use the less noble synchronous memory option, DDR memory. At a press conference at IDF, Peter McWilliams, an Intel fellow, ruled out DDR for the desktop, while Peter Glaskowsky, a senior analyst at Microdesign Resources, said that it was inevitable that there would be DDR chipset solutions for Willamette when it arrived. And, at a series of one-to-one briefings beside the swimming pool on a scorching Palm Springs day, Yu repeatedly refused to be drawn on exactly why Willamette was so closely linked to Rambus. Nor would he expand further on why Foster seemed to have a love-affair with DDR memory. According to two stories on Electronic Buyers' News (EBN) at the end of last week, it now appears that Intel is readying a cunning Baldrick-style plan to do the same for Willamette as it is being forced to for the current generation of IA-32 processors, the Coppermine family. EBN reports that Intel, faced with a threat from Athlon Thunderbirds (which uses DDR memory), is now repositioning Willamette to be a mainstream desktop solution, implying that it has realised it has to prevent AMD from continuing to nibble at the x86 hand that originally fed it. The same stories suggest that Intel is readying a memory translator hub (MTH) solution that will allow Willamette to dance the funky chicken with DDR as well as Rambus. The EBN reporters have been busy indeed. They also reported Friday that Micron will now use its Samurai DDR chipset to support the AMD Athlon, as well as Pentium III processors. It is true, beyond doubt, that Rambus memory is still vastly more expensive than synchronous memory. It is also true that demand for Rambus RIMMs has not yet begun to overwhelm PC manufacturers. And it is also a fact that Intel's i820 Rambus chipset has not gone down a storm with third party mobo manufacturers nor with Intel's major customers, the PC manufacturers. But we suspect that this lengthy Rambus game is not yet over, despite this latest evidence of an Intel u-turn. Between February and June, we are given to understand, teams of frantic mengineers and womengineers have been hammering out a new type of Rambus, with cheaper casing, a lower form factor, and more reliable technology. The marchitecture departments at Rambus are likely to go into overdrive too, in an attempt to persuade the world+dog that RIMMing is good stuff, especially the new, improved version they will show us. We are likely to find out when we attend June's Computex show in Taipei how the PC industry at large views these latest maneuverings in the memory war.