The current buzz is that Microsoft's Windows 2008 Server R2 will be the last version of the operating system that will support Intel's Itanium. My colleague Mary Jo Foley posted Microsoft pulls the plug on future Itanium support on April 5th on this very topic. It's not surprise that many industry media outlets are acting astonished and are castigating Microsoft for leaving people behind. For long-time industry watchers, however, the view is so what else is new? This has happened many times before.
A little historyFor those who remember the history of Microsoft's Windows NT operating system, the predecessor to the current family of Windows 2008 products, this is nothing new.
How many of you remember the ACE initiative? For those of you who are scratching your head and wondering what Dan is mumbling about this time, the Advanced Computing Environment was part of an effort by an industry consortium that was designed to attack the growing Windows/Intel market dominance and also take a swipe at Sun Microsystems' workstation market position as well.
It included a long list of computer and technology companies that all promised to develop a compatible series of desktop systems that be based upon a selection of one or more microprocessors from Digital Equipment Corporation (the Alpha), Intel (the i486 and later Pentium), IBM (the PowerPC), and MIPS (the R4000). These systems would support two operating systems - SCO UNIX and what would become Windows NT. The announcement of this initiative was in April 1991. Remember now?
Companies such as Acer, Digital Equipment Corporation, Compaq, Control Data Corporation, Kubota, Microsoft, MIPS Computer, NEC Corporation, NKK, Olivetti, Prime Computer Pyramid Technology, Siemens, Silocon Graphics, Sony, Sumitomo, Tandem, Wang Laboratories and Zenith Data Systems were behind this effort.
SCO UNIX never appeared on any other platform but X86.
It is clear that with few exceptions this effort failed.
Reality raises its interesting headVery technical issues, such as how floating point data was stored in a machine word became sticking points in efforts to create compatible systems. Some of the microprocessors stored exponents at one end of a machine word while Intel stored this data at the other end of the machine word. DEC's Alpha could support either approach if it knew ahead of time how the data had been stored. ACE Initiative members couldn't come to a final agreement on which approach was best. A splinter group, the Apache group, was formed by some of the members in the hopes of taking this group in a "big endian" direction.
As Digital's Intel/UNIX business manager, I found myself caught up in the effort to support this initiative. It was very challenging to get the Windows-oriented folks to think about the world the same way as workstation people thought. The audience for PCs and the audience for workstations were different. they wanted different things, different applications, had hugely different levels of technical expertise and expected a different approach to marketing and sales.
Other issues also came to the forefront including the fact that some members built their business on high value/low volume products while the others focused on low value/high volume products. Getting everyone to agree on a single approach to the market became a big problem.
In spite of the confusion, Microsoft kept its focus and made sure that Windows NT was one of the survivors.
A focus on WindowsHow many of you remember that Windows NT (which stood for "New Technology" by the way) initially was made available for systems based upon Digital Alpha, IBM PowerPC, Intel i486 and Pentium, and MIPS R4000 microprocessors? Although a version one product, Windows NT was launched as "version 3.1". Microsoft's hope was, of course, was to allow this new operating system to play on the broad acceptance of Windows 3.1.
As we all know, Microsoft's mantra is volume, volume and more volume. At that time, their sole focus was low value/high volume products. This means that the company started dropping support of low volume microprocessors very early in the life of Windows NT. By the time Windows NT V3.5 was launched, only Intel i486, MIPS R4000 and Alpha chips were supported. Later when Widows NT V3.51 was launched only Intel and Alpha chips survived. Windows NT V4 killed off Alpha support.
If we fast forward to the end of 2003, Microsoft was persuaded by its buddy Intel to support its new Itanium architecture with special versions of Windows XP and Windows Server 2003, both of which were successors to Windows NT.
Snapshot analysisIt is clear that the volume of Itanium-based systems never met Microsoft's goals. Furthermore other processors from both Intel and AMD have surpassed the performance that was offered by Itanium. These processors were more compatible with many organizations' applications and so, the market adopted them rather than changing everything to work well on Itanium.
Microsoft, being strongly focused on its own bottom line, decided that Itanium would never take over the computing world and so, decided to drop it. I'm sure that they'll figure out a way to make their customers move applications to newer Intel or AMD-based systems and like it. Overall performance and lower cost are likely to play in that story.
As an aside, this is at least the third time that Intel has jumped the shark when attempting to offer a new microprocessor architecture. Similar issues happened during the 8-bit to 16-bit transition, the 16-bit to 32-bit transition and the 32-bit to 64-bit transition. Each time, Intel believed that people would be willing to re-architect everything, including things that were working happily on earlier microprocessors, to an incompatible architecture (the level of incompatibility differed each time). Each time, the golden rules of IT were invoked by IT decision makers and Intel's efforts found their way into the pages of IT history.
When we consider the history of both companies, we'd have to come to the conclusion that we've seen this film before and go on with our lives.