In-Depth
Microsoft Throws in the Towel on IA-64
Microsoft's decision to drop support for Intel's Itanium chip was anticlimactic. Nevertheless, it's the kind of non-event Itanium fans couldn't have envisioned a decade ago.
In the end, Microsoft Corp.'s decision to drop support for Intel Corp.'s Itanium microprocessor was anticlimactic.
Dan Reger, the senior technical product manager for Microsoft's Windows Server unit, announced the decision in a blog post. Reger's posting had the flavor of a just-before-the-weekend-document-dump, in part because it was published at 2:59 PM ET on Friday, April 2.
"Having been involved in Windows Server support for Alpha processors in the past, and seeing many of the early chapters of 64-bit Windows Server written on Intel's Itanium architecture, it's with a sense of nostalgia that I pass on some important news," Reder wrote.
Microsoft isn't hanging Itanium users out to dry. The company plans to continue to support Itanium on all currently supported products. In the case of Windows Server 2008 R2, Reder indicated, this means customers can expect a minimum of eight more years of support (encompassing both mainstream and extended support), which should last through July 10, 2018.
Reder added that the "natural evolution of the x86 64-bit … architecture has led to the creation of processors and servers which deliver the scalability and reliability needed for today's 'mission-critical' workloads."
Industry watchers say there's something to be said both for and against Microsoft's move. "On the plus side," writes industry veteran Charles King, a principal with consultancy Pund-IT, the future of x64 chips -- from both Intel and competitor Advanced Micro Designs (AMD) Inc. -- looks bright. "The new Intel Xeon and AMD Opteron processors are quite simply the best and most able x64 server offerings either company has ever produced," King comments. "More importantly, Intel's and AMD's respective multi-core technologies look extensible for a long time to come."
On the other hand, HP has a huge stake in Itanium by virtue of a long-standing (and Itanium-centric) partnership with Intel. HP planned to port its flagship HP-UX operating environment off of its proprietary PA-RISC architecture and onto commodity silicon. Intel wanted to develop a 64-bit architecture that it could position as a high-end successor to its CISC x86 line. Both HP and Intel collaborated to develop Itanium, with its Explicitly Parallel Instruction Computing (EPIC) architecture. Itanium, or IA-64, debuted very late and was dogged by early performance issues. Meanwhile, Intel competitor AMD had other plans on the 64-bit computing front, introducing its own line of x86-based 64-bit chips in 2003. AMD's move forced Intel to develop its own 64-bit x86 chips.
Two years after AMD introduced Opteron, Microsoft delivered a version of its Windows Server 2003 operating system for x64. Thereafter, Microsoft would also deliver an x64-compatible version of its Windows XP Professional desktop OS.
The fate of Windows-on-Itanium -- like those of Windows-on-MIPS, Windows-on-POWER, or Windows-on-Alpha -- might then have been sealed.
In every case, the value proposition of a Windows operating system running on commodity hardware trumped the putative performance or reliability advantages of a Windows OS running on not-quite-commodity platforms.
In the case of Itanium and Intel's latest Nehalem-EX processors, the performance and reliability gaps have been narrowed even further.
"The fact is that Microsoft's Itanium efforts -- beginning in the mid-1990s -- have never really resonated commercially. According to recent survey data, less than 5 percent of Itanium systems sold in 2009 supported Windows Server, while … [Unix-based] Itanium sales continued to grow," King writes, noting that the bulk of Unix-on-Itanium configurations consist of HP-UX running on HP's Integrity Itanium servers.
"Microsoft's move is a significant reality check for boosters who continue to posit Itanium as the foundation of Intel-based mission-critical enterprise computing. If any further evidence were needed, Microsoft's decision cements x64's role at the center of those scenarios."
For this reason, King sees Microsoft's move almost as a non-event. "HP
owns [more than] 90 percent … of the Itanium market and leverages the platform across a variety of common … and legacy operating environments," he points out, citing HP-UX, Linux, and Windows as examples of the former (or common) and Open VMS and NonStop as examples of the latter.
Microsoft's decision to pull the plug on an IA-64 version of Windows might put a kink in HP's mixed workload pitch -- e.g., the ability to host HP-UX, Windows, and Linux on the same SuperDome Integrity servers -- but King doesn't foresee HP pulling the plug on Itanium. The irony, after all, is that Palo Alto is facing a kind of PA-RISC redux: Itanium is now slightly less of a niche product (with support from Silicon Graphics Inc., Groupe Bull, and NEC) than PA-RISC (which was also OEM-ed by NEC, along with Hitachi, Oki, and -- for a time -- Mitsubishi).
"The size of the company's relative commitments to Itanium makes any radical changes in HP strategy doubtful. In addition, given the company's close relationship with both Microsoft and Intel, one expects HP will easily step up to support its partners' shifting strategies," he concludes.
"The same is even more likely for smaller players, like Bull and SGI, which leverage Itanium in Linux-based systems. In fact, they and the vendors who do offer Windows on Itanium systems, including NEC and Fujitsu, are also enthusiastic supporters [of] x64-based technologies. One expects that by 2018, many if not most of those vendors' Windows on Itanium customers will have long since migrated to x64."