Has the Itanium got a future? It's question that many people have asked in recent years, as the development and sales pace of Intel's most powerful -- but also slowest-selling -- processor continue to dribble along in a most unIntel-like manner. We asked the question we asked back in September but things have moved on since then -- not least HP's acquisition of a new leader.
On the downside, in January this year Microsoft pulled out from its Itanium-based XP development programme, in favour of the more popular Xeon. The move marked the end for Itanium 2 in Windows-based workstations and followed the abandonment by all major hardware vendors of the 64-bit chip for use in workstations.
Even Intel agreed the latter was a smart move, as spokeswoman Erica Fields said, "The workstation market really has never been a main focus for the Itanium. Xeon with 64-bit capabilities really provides the best overall price performance for the workstation market."
Intel has also acknowledged that the chip has not met its expectations. We were promised the next all-singing, all-dancing processor back in the early 1990s, when Intel first trumpeted a new architecture that it, along with HP, would develop for the bright new world of the twenty-first century. It took much longer than expected but, finally, it arrived. And after the initial brouhaha, few people took much notice. Since then, it has chugged along towards the bottom end of the high-end Unix market with, according to Intel, some five per cent share.
On the upside, it has a major champion in HP, the importance of which is that HP is the high-end server market leader, according to market analyst IDC's figures.
And in December 2004, with the aim of boosting Itanium's flagging sales performance, HP announced it would spend over $3 billion (£1.54bn) over the next three years on Itanium-related product development, money to be spent on R&D, software and hardware design, and marketing. Given the scale of that investment, the recent departure of its CEO notwithstanding, HP is unlikely to abandon Itanium any time soon.
HP watchers, such as Denys Beauchemin, a past chairman of HP user group Interex and migration consultant at software developer Sector7, noted HP's plans to discontinue Alpha and PA-RISC-based systems, and said that HP was not in the chip building and designing business anymore. They put all their eggs in the Itanium basket.
Analyst Rich Partridge from DH Brown Associates agrees that users are likely to see a speeding up of transitions away from Alpha and PA-RISC and a reduction in the number of products the company has to support. I think HP has no choice but to move forward with Itanium.
And let us not forget, the Itanium remains a powerhouse of a chip, which is a compelling argument for many compute-starved applications. With 592 million transistors and a 432 sq. mm die size, it's the largest server die in full production today. The multi-processor version runs at 1.6GHz and houses up to 9MB of L3 cache -- if your hardware design can cope with the maximum power dissipation of 130W. In its market space, few have argued that it cannot cut the mustard, MIPS-wise.
There's one more, and potentially life-saving crutch for Itanium -- and that's Linux. Even Dell offers Red Hat Linux on its high-end servers, while on its (even higher) high-end systems HP offers only Linux or its own OpenVMS -- which despite the protestations of some of its ardent supporters, looks to be going nowhere today.
While the jury is still out as to whether Windows or Linux which is better -- and let's face it, it depends on your situation, applications and skillsets as much as anything else -- the applications are there for Linux, while Microsoft remains focused on the sweet spot of the pure x86 architecture.
HP is not alone either. In March, market researcher IDC reported that even IBM makes more out of Linux on Itanium than it does Linux on Power. It accounted for 4.8 per cent of IBMs worldwide Linux factory revenue for 2004, while Linux on Power accounted for 3.1 per cent. IDC also reported that nearly two-thirds of all the x86 servers shipped by Sun were shipped with Linux -- not Solaris.
The Itanium most likely does have a future, provided that HP and the Linux community -- either combined or by some happy but unconnected coincidence -- continue to support it. But if I were running Intel's Itanium business -- and two of those have recently quit -- I'd consider having a fall-back plan in my pocket.