Intel's Itanium: A leap too far?

Intel says Itanium has not done as well as it expected. We hate to say we told you so, but...
Written by Leader , Contributor
The move to 64-bit computing is inevitable: when it's no longer unusual for desktops to ship with up to 1GB of RAM, the 4GB limit imposed by 32-bit computing will seem uncomfortably close.

Speed, of course, is the other factor; 64-bit processors are just faster than 32-bit processors. You only have to take a look at ZDNet UK's processor shoot-out to see Athlon 64s clocked at 2.4GHz equalling and in many cases beating 32-bit Pentiums running a whole gigahertz faster. That test is not meant to be an empirical comparison between 32-bit and 64-bit chips with common baseline set-ups, but you get the point. Sixty-four-bits are great. So why then are Intel's flagship 64-bit Itanium chips failing to sell -- as the company has finally admitted?

The admission won't surprise anyone familiar with the chip industry; Itanium was always an ambitious project:

  • It took nearly 10 years to move from HP's early EPIC architecture to the full-blown Intel Itanium.
  • The first Itanium release was pitched as a 'trial' version.
  • The company finally admitted as much at its Developer Forum in San Francisco.

Of course this doesn't spell disaster for Intel, or for Itanium. People are buying the things, just not in the volumes that the company had hoped. AMD meanwhile snatched the crown with its Opteron launch a year and half ago, which offered a much more comfortable upgrade path to 64-bit computing.

Intel's problem with Itanium is several-fold: it discarded native 32-bit support in favour of a solution that requires an emulation layer to run your existing software; and it has to be very careful that Itanium processors don't cannibalise its Xeon range, which are widely used in one and two-way servers. The result for buyers is an expensive processor that is slow at 32-bit apps, and has to be pushed into the exotic heights of eight, 16 and 32-way servers.

What this means in practice is that for many, a move to Itanium is almost as big a leap as it would be to switch to Sun's Sparc architecture. This is partly why AMD's strategy with its 64-bit Opteron processors seems like such a stroke of genius, even if it was in reality little more than common sense: it allows you to buy a new system but retain your current software investment until you're sure you want to make the move. It's more gradual, and all the more comfortable for it.

Intel's followed AMD's successful lead last month with the launch of a Xeon processor (codenamed Nocona) with 64-bit extensions. This is the 64-bit Intel architecture that many buyers will have been waiting for. While Intel remains bullish about Itanium's long-term prospects, buyers should pay more attention to Xeon.

Editorial standards