Longhorn no window of opportunity for Itanium

It's rare that Microsoft is troubled by a lack of self-confidence — even when richly merited — but today's announcement of Itanium support for Longhorn shows just how nervous the Redmond giant has been in high-end systems. The company's less-than-stellar reputation for reliability and security has created a cynical, critical audience for its pitches at the posh end of the market, and that's not a community whose objections are easy to ignore.

However, Microsoft is now gaining confidence — even if it's not doing too well on shipping products. It will have a clustered edition of Windows, when the schedule stops slipping. It has bought into 64-bit, across the board.. And now, if you train your telescopes to the red-shifted end of the Longhorn Server release schedule in 2007, it will have Itanium support to boot.

That's good news for Intel, but no news would have been very bad. Microsoft has nowhere else to go: all of the other chips in that world are owned by its competitors. The day Windows will be ported to Power or any other RISC architecture will be when Bill Gates releases it under the GPL, the day after Zen Buddhism becomes the state religion of Saudi Arabia.

The other signs for Itanium's success are mixed at best. Sales are up, but no target has ever been hit. Lots of chips are going into a few very large systems, where its undoubted superiority at floating point maths make it a good choice, but that's no market for a young architecture with ambitions. How many have been sold at book price? You may ask. Why won't Microsoft include Itanium support in its first version of clustered Windows? Another good question.

Meanwhile, system designers report that in other matters Itanium remains a gristly chew. It's hard to write efficient systems software for the chip, it won't virtualise easily and there's no groundswell of new and interesting products, hard or soft, on the horizon. Apart from maths, the chip's big advantage has been its ability to cope with huge amounts of performance-boosting memory: here, 64-bit variants of the x86 architecture are massing to invade the low end of high end before Itanium's even found them on the map.

Montecito will fix some of the technical problems through architectural tweaks: dual core and massive cache will hide some of the rest, for a while. Unless that's enough to boost the architecture into serious sales well before 2007, the long wait for Longhorn will have been in vain.