Who wants Itanium processors?

Intel has altered its roadmap for server chips, raising questions about the future of the Itanium chip

Intel's announcement that it has cancelled or delayed several of its forthcoming server chips raised a few eyebrows here at IT Week. In particular, the next version of the 64bit Itanium chip, codenamed Montecito, has now been delayed until next year, and future Itanium updates will occur on an annual cycle rather than on the more frequent schedule that was originally slated.

While most Intel executives are generally loyal to the Itanium chip, there is no shortage of independent observers who think the chip is about to be killed off. There are even more IT directors and chief technology officers who tell anyone who will listen that the Itanium is not an attractive proposition. In recent times there have even been a few Intel executives who seem more focused on x86 compatible chips such as the Xeon than on the Itanium.

As you probably already know, x86 instructions are the ones normally used by Windows and Linux, and are implemented in Intel's Pentium and Xeon processors and AMD's Opteron. They represent the mainstream of business computing systems. In fact, these chips don't actually run x86 code directly - they take x86 code and translate it into some mysterious microcode that the chips can actually run. Intel has several completely different micro architectures that do this, including the NetBurst architecture used by the Pentium 4 and Xeon chips, and the Banias architecture used by the firm's recent mobile processors.

Some might argue that the Itanium architecture is also x86 compatible, but although the Itanium is capable of running x86 software, it makes such a poor job of it that no one would consider running x86 applications on an Itanium system.

At Intel's developer conference last autumn, one of its executives was asked about plans for supporting the x86 instruction set. The executive quoted the creator of Intel's legendary 80386 chip, John Crawford, who in 1985 said, "I don't mind having a rock around my neck as long as it is a gold rock." Software compatibility is a golden rock. Intel still builds stuff into its x86 chips to support software that was written 15 years ago.

But rather than focus on concerns about the Itanium's poor compatibility with their existing software, IT managers usually complain about its poor performance for general business workloads. Intel's Xeons and AMD's Opterons both operate at higher clock frequencies than the Itanium, and this seems to be the key to delivering superior performance for running integer workloads.

I'm not sure the Itanium will go away completely, if only because it's a vital part of HP's plans to retire the PA-Risc chips that power its high-end Unix systems. But the apparent cutting back of Itanium activity suggests that the demand for the processor has been somewhat lower than Intel had expected.

Intel claims the contrary is true; and sales in the first quarter were 170 percent up compared with the same period last year. No doubt these arguments will run on until Intel finally cancels the chip. Of course this is bound to happen one day. The question is when.