Nothing but a deafening silence has emerged from Digital Equipment Corp and Intel Corp over still unconfirmed reports in the Wall Street Journal on Monday that Intel would be acquiring DEC’s Alpha chip technology (CI No 3,262). DEC even went so far as to say it had no plans to issue any such statement. According to Michael Slater, founder of The Microprocessor Report, as quoted in Tuesday’s San Jose Mercury: This deal only makes sense if Intel is ‘buying’ Digital as an IA/64 [aka Merced] customer. The Mercury also reported that Intel chairman Andy Grove, as well as DEC chief executive officer Robert Palmer, are both against such a deal.
Meanwhile Maureen O’Gara, in a ClieNT Server News newsflash, believes that the deal on the table is not about acquiring the Alpha chip per se, but is limited to Intel buying the Hudson, Massachusetts facility and taking a cross license to the Alpha technology. The newsletter reports that fingers were pointing to DEC’s Bruce Claflin, an ex IBMer that never cared for the Alpha, as the most probable source of the story, as leaked to the Wall Street Journal. The newsletter said the leak appeared not to have had sanction from Intel, and that its purpose might have been to ensure that Intel would find it harder to back away from the negotiations.
Reports were emerging just before the Intel rumors broke that Samsung Electronics Co was re-thinking its own agreement to support the Alpha chip. Electronics Weekly reported that N B Park, senior vice president of Samsung subsidiary AST Research Inc, was heard to say that Samsung would have difficulty finding customers beyond DEC itself and a few small companies. Samsung has said it would produce 4,000 Alpha units this year, 110,000 units next year, and would be ramping up to 1m units in 1999 (CI No 3,239). One alternative it’s thinking about is a plan to use the Alpha as the core of a low-cost multimedia processor, to replace its MSP-1 multimedia chip project, abandoned back in May. Mitsubishi Electric Corp is DEC’s other second source for the Alpha chip, of course.
Admission of guilt
And, if the whole business goes ahead, it will make Intel Corp’s initial statements of innocence when DEC took out its original lawsuit look somewhat tainted. If Intel buys the Alpha technology, isn’t it tantamount to an admission of guilt on Intel’s part?
Like Sun Microsystems Inc, DEC is eschewing a distributed shared memory architecture of the ccNUMA variety for an extended SMP symmetric multiprocessing model in future Alpha servers. It says it’s got some big iron coming down the pipe, although its high- end AlphaServers with a maximum of 14 processors are in any case some way behind the size of Unix SMP server configurations Sun or Hewlett-Packard Co can now offer. DEC’s newest SMP Unix server, the AlphaServer 8400 5/625 (CI No 3,259), is powered by up to 14 612MHz Alpha 21164 RISC chips the part once known as EV6. DEC says a 10-way system performs 24,537 tpmC at $110 per tmpC.
Analysts’ figures suggest that DEC sold only 275,000 Alpha chips in 1996, giving it a 1% share of the total market. Revenues derived from Alpha were around $4.6bn in the same year. According to RISC systems analyst Andrew Allison, the Alpha had an 8.9% market share of the total, 51.7bn Unix on RISC marketplace in 1996, up 50% from the 6% share it had the year before. Sun Microsystems Inc took 17% with Sparc, the PowerPC camp 27.5% and Hewlett Packard Co 29.6% with its PA-RISC. The market’s slower growth rates were attributed to the rise of Windows NT on Intel.
Merrill Lynch, which has issued a research bulletin entitled Offloading the Alpha notes that support of Alpha by Intel could help near-term business. On the other hand competitors will say that Alpha is a dead architecture long term. That process has already started. Alpha has had a tough time and it’s an opportunity for Digital to rid itself of it, Sun Microsystems Inc told The Wall Street Journal yesterday. It shows we’re right to call it a two horse race. And it’s a distraction for Intel.
NT on Alpha
Not everybody, however, is ready to dismiss the prospects of NT on Alpha just yet. Alpha is a going concern as far as NT is concerned, and, following the withdrawal of the PowerPC camp, is now the only alternative to Intel. The crucial questions are: how long will it take to port NT to Merced, and how much will Intel rely on compatibility with its current processors – the more they do, the more performance is likely to suffer. Microsoft Corp has made it plain that, after Windows 98, it’s NT everywhere. When Intel moves from Pentium to Merced, the theory goes, there could be a hiatus which could let Alpha in with a vengeance. So Intel may feel the need to keep Digital sweet after all, rather than turn it into a cornered rat. And when the Alpha-optimized 64-bit version of NT comes out, DEC could be even further in front.
Remember those secretive discussions between DEC’s Robert Palmer and Microsoft’s Bill Gates a few weeks ago (CI No 3,242)? It was said that Gates was trying to dissuaded Palmer from abandoning plans to produce a NetPC. Well, we saw on Monday that Gates had won the battle, and DEC scrapped it’s proposed Shark Network Computer instead (CI No 3,262). What else did Gates and Palmer talk about at that session, we wonder?
Candidate for acquisition
If ridding itself of the Alpha makes DEC a stronger candidate for acquisition, step forward Compaq Computer Corp. Although Compaq has already digested Tandem Computers Inc, it could still use DEC’s unrivaled experience with high-end NT systems and clustering. And it has the money, courtesy of its recently secured $4bn credit facility.
As we have already noted (CI No 3,255), DEC looked as if was planning to sell off its Hudson, Massachusetts fabrication facility once it was no longer capable of supporting the next generation of Alpha, around the turn of the century. That way it would have become a fabless semiconductor manufacturer, like Sun Microsystems.
In December 1991 (CI No 1,814) DEC said that it expected the Alpha architecture to take it through the next 25 years, and be scalable for all species of machine from palmtop to supercomputer. Well, it’s made the first six years.
This article is from the CBROnline archive: some formatting and images may not be present.
Join Our Newsletter
Want more on technology leadership?
Sign up for Tech Monitor's weekly newsletter, Changelog, for the latest insight and analysis delivered straight to your inbox.