Guy Kewney: Alpha chance for Intel?

Both Hewlett-Packard and Intel combined forces to develop Intel's current Itanium family of 64-bit chips. Now, rival Alpha chip designers have been added to the pudding.

The promise from Intel is that superior Alpha features will be incorporated into future Itanium family processors.

But the fear is that Intel will carelessly throw away the expertise that the Alpha team will bring, the way it threw away the expertise it got from HP.

It is not a secret in the industry that Hewlett-Packard engineers deeply regret their company's decision to give their 64bit processor design to Intel because, from HP's point of view, Intel simply didn't understand what it had been given. Worse, Intel's bright young chip designers didn't realise they didn't understand.

Today, at the very least, Intel chief Craig Barrett knows he needs help. His own comment on the news that Compaq was giving the Alpha technology to Intel was: "This gives us access to a super engineering staff that really understands the high end of computing" -- which is about as clear an admission as you're going to get to the fact that they didn't have those people before.

The unacceptable truth is that Intel still can't make a processor as fast as the Alpha processor which Digital Equipment designed a decade ago, and which Compaq bought when it took over the computer services side of DEC.

But the equally sad fact is that Compaq can't afford to take the design of the Alpha any further. It doesn't have the resources, but even if it did, there isn't the market for it.

For users of Alpha-based systems, it is still possible to carry on whistling in the dark. For example, they say, the VMS operating system is solid, 20 years old, and vastly to be preferred, for serious business applications, to anything from Microsoft.

And, they add, Intel's promise to take Alpha on means that we'll see a new generation of Alpha processors up to the year 2004, and support for Alpha for the next five years, too.

But in the rest of the computer world, the software that people want to run is Microsoft or Unix. You can no longer get NT for Alpha, and you will never be able to get Windows XP for Alpha; so you're stuck with VMS.

Even if you think VMS is as wonderful as its users say, it's still a tiny proportion of the market; and the Alpha is the only chip that will run it. Producing another two versions of Alpha would be throwing huge, astronomical amounts of research money at a problem with the hope of dominating a tiny market.

But if Compaq was stuck in a hole, Intel was no better off.

Its 64bit range won't actually be able to compete on level, performance terms, with Alpha until the second generation McKinley chipset is released -- not for another couple of years, in short. And even worse, the new 64bit range isn't compatible with the old 32bit Pentium family. The PC world is in the doldrums because the Pentium 4 is still no faster than the Pentium III, and the 64bit machines won't run Windows.

From the point of view of the typical PC user, the move of the entire Alpha engineering team to Intel has to be excellent news in one sense: it means that we stand a chance of seeing truly fast processors on desktops, in three years' time. The Alpha team, as Barrett says, truly does understand fast processor design at the top end. And while there are pundits who will say that the top end is going to be a pure server-based market, the fact is that there will always be a bigger market for standalone machines, and the Holy Grail is to have a design which will do both.

But it is, also, the knell of doom.

With Compaq's admission that it simply can't afford to keep playing in the 64bit market, we are down to two processor designers (unless you take IBM seriously with its Power PC range, which I don't). That is, Sun with its UltraSparc family, and Intel, with everything else.

There are some, like the team at Transmeta, for example, who believe that the future remains PC compatible, and that Intel is off down a dead-end road with Itanium. They predict that AMD's Hammer design, or some other processor with true Windows compatibility, will make the real running in the future.

But if they're wrong, then Intel pretty much owns the computer business in the future. Itanium may not be able to run Windows XP yet. But UltraSparc can't, never could, and never will be able to.

It is, of course, wretched news for Microsoft; it means that it may even have to watch Solaris take over as the de facto server operating system. But frankly, anything would be better than a future where Intel owns all of us.

See Chips Central for the latest on processors and the semiconductor industry.

Have your say instantly, and see what others have said. Click on the TalkBack button and go to the Chips Central forum

Let the editors know what you think in the Mailroom. And read other letters.