Skip to main content

Itanium: The dead-end Intel drove right into

There's been a lot of Itanium talk lately, most of which is about how Intel seems to be losing interest in a chip that has pretty much become a custom part for HP's Integrity servers. HP, of course, was part of the development effort and helped design the chip.

For those of you who were not tuned in during this hysterical moment in history, the Itanium was a product line introduced with outrageous fanfare in the late 1990s. I witnessed this hoopla and took an immediate dim view of the whole thing since it was bucking every known trend.

The chip actually began life as a joint development effort with HP and Intel to exploit an obscure architecture called "explicitly parallel instruction computing" (EPIC). They began in earnest with what they called the IA-64 architecture in 1994, and this is when the ballyhoo started. Everyone signed on to the hype machine, including most analysts, pundits, and research companies, and predicted it would kill all competition.

Apparently, nobody considered three things: First, the previous failed attempts by Intel to create a better architecture, including the hopeless iAPX432; second, the absolute juggernaut the x86 had become; and third, the idea that HP and Intel could actually work well together.

It got funnier and funnier until 2001, when the chip was finally rolled out, after having been delayed, of course.

Immediately, the chip was condemned as an underperformer. All the supporters bailed out and the chip was soon nicknamed the "Itanic." Intel made efforts to fix the problems and it did improve the chip, but that same bubble universe that extolled the chip now turned on it. Worse, AMD decided to milk the x86 and rolled out a 64-bit version ahead of Intel, catching Intel flat-footed.

Although AMD has offices in Silicon Valley, its milieu is out of the local bubble and in the Texas milieu, a different mentality altogether with problems of its own.

So this week, we hear that Intel is not even going to use its newer manufacturing process for the next – and possibly the last – iteration of the Itanium, which now almost exclusively sells to HP.

All the while, other pundits were telling us how mobile devices would be the way we access the Internet and how important they would eventually become. If Intel had kept that in mind during all its development efforts, instead of the Itanium, one can only imagine what sort of scene we'd have today. Instead, ARM rules that low-power roost by lucky coincidence.

The last laugh is that the x86 is indeed still alive while the Itanium turned out to be the dead-end architecture that Intel feared all along. The company was so afraid of the dead-end that it actually created one.

I'm reminded of one of the rules of motorcycling, especially in the dirt: When you are zooming around and see a boulder up ahead that you may hit, do not look at the boulder because you will hit it. Look at where you want to go. Intel kept looking at the dead-end boulder and indeed hit it. It's as if the company manufactured the dead-end boulder so it could hit it. It's almost metaphysical. Or maybe wish fulfilment.