Richard,
The Register agrees with you:
theregister.co.uk
AMD, Via invited to join Intel's Six Dramurai gig
When we reported last week that Intel was behind a plan to push a new memory alliance last week and that Rambus Ink was not there to fire its poppers or toot its horn along with the rest, our ears pricked up.
Sho'enuff, when we called Chipsetzilla the next day to seek confirmation of these reports, representatives said that Rambus was not part of the project but that Intel welcomed "any interested parties" to join that wanted to. At the time we wondered if we should throw our hat into the ring -- we're a very interested party, after all.
The story did its usual transmutation through the week, somewhat drowned out by Transmeta's announcements, but leading up to the remarkable news last Friday that INTC would be happy not only if Rambus become an "interested party" but that AMD and Via, too, were welcome to join in.
There was, however, nary a word from Rambus Ink as to what its executives felt about the new initiative, which must surely affect its future. The plan is a broad-brush scheme aimed at producing a future generation of memory from the year 2003 onwards.
What's behind all of this? It's a complicated question full of intriguing variables and drawing in Intel's plans for its IA-64 platform too.
The first rumblings that all was not well with the Rambus platform came at one of these "fests" in the middle of last year, when PC vendors developing future platforms using Rambus memory reported that they were having technical problems making prototype boards run sufficiently fast. Major Intel customers, including Hewlett Packard, IBM and Compaq, all voiced their complaints, which coincided with a push by another group for Intel to join the PC-133 consortium set up some months before.
Via, now a prime target of Intel's ire, was a major ideas holder in the PC-133 initiative, and at the important Taiwanese trade fare at Computex last June, it was clear that there was no great enthusiasm for Rambus from either the third party mobo makers, nor from the memory businesses on the island. Mosel-Vitelic, in particular, was outspoken about the unsuitability of Rambus as a PC platform for the future, and said it would not be making chips for Rambus RIMMs.
Three months later, things had progressed (or degraded) sufficiently for Intel, at its autumn Developer Forum, to cobble together its now famous Seven Dramurai consortium -- where some rather embarrassed looking executives of major memory companies publicly took the line that Rambus was, in the words of 1066 and All That, a "good thing".
However, even by the time of the Forum, it was evident that Intel's customers had had their wicked way and put pressure on Chipsetzilla. INTC was showing the first system prototypes of its Merced (Itanium) server and workstation and once we'd persuaded a friendly chappie on the stand to open the box, we could see no trace of anything resembling a Rambus RIMM. Later, in a private meeting, one insider informed us that HP had got really stroppy and told Intel that there was just no-way that Rambus could be used in Itanium machines...
Meanwhile, Intel spinolas were puffing and blowing as they had to handle the difficult issue of the up-and-coming i820 chipset, which, they said, would not be able to support Rambus memory to the extent the designers had first intended. So much so, that senior executives from the Corporation were forced to admit that they would now, after all, be creating a PC-133 chipset.
Only six months earlier, when problems with the i820 chipset first emerged, staffers at The Register had phone calls from Intel more or less suggesting that Satan Clara Would Freeze Over before the Corporation supported PC-133 and double data rate (DDR) memory. Three weeks after the Caminogate i820 debacle, and new motherboards were withdrawn because of a technical problem, it was obvious that if the Rambus game wasn't over, there certainly was a new game in town.
The memory manufacturers, as usual, had to be polite to all parties. Already furious because Intel had pushed them into using Rambus, which demanded a royalty payment for using its Intellectual property, they also began to see companies such as AMD and Via beginning to make waves in the market. As AMD now claims a 20 per cent market share in the desktop area, and that is likely to grow, neither it, nor Via, can be ignored as future revenue streams.
If the reports are true, and AMD and Via join the Six Dramurai to make it the Eight Dramurai, it gets INTC off any number of hooks. First of all, it is publicly displaying that it now has realised the error of its ways, and sees that an open memory standard, for all, is a "good thing". Secondly, the memory manufacturers themselves will be happy that they're no longer being seen to be bullied into developing memories against their own will. And, thirdly, INTC's PC customers will be happy too that they are being listened to.
During the week, the intro of Transmeta microprocessors hogged the headlines and the release of Rambus Ink's Q1 financial results was pushed into the background. Its take on the quarter can be viewed here, and it's pretty evident that it prefers to talk about the Sony PlayStation and Nintendo rather than beat its chest about the success of the platform for the PC market.
Given that the first RIMMs have only just started appearing and that even Kingston Technology, a leading Rambus partner, has claimed that companies making the chips only get a 50 per cent yield, a real question mark must now hang over the future of the platform. © |
