Microsoft needs to wake up NT's alternative processor

Windows NT was supposed to be the architecture-independent OS. I have 1990 Microsoft press literature that says so. Microsoft briefings at the beginning of the decade stressed that the company planned to develop the original version of NT on a little-known RISC processor called the MIPS R4000, a faster-than-the-wind alternative to Intel's 486. Speakers at the time hinted that Microsoft was exploring implementing NT on the 680x0 platform to offer Macintosh users an alternative to the Mac OS. (Mac users, stop guffawing. The next guy on this magazine's routing list won't appreciate spittle on the pages.) One Microsoftie at the time suggested that NT wasn't out of the question as the OS for the IBM 370 hardware platform, although he didn't know how to respond when someone asked him where such a system's mouse would plug in.

As Microsoft developed early versions of NT, it added support for the Alpha and PowerPC chips. When NT 4.0 shipped, NT users had a fairly wide selection of processor architectures. Of course, that was back when NT had a negligible market share and Apple was still something of a market force.

But MIPS fell by the wayside to find a new life in handheld Windows CE machines. And then Microsoft stopped supporting the PowerPC. Might the Alpha be next?

Alpha Has a Faint Pulse
Digital no longer pilots its own course; now it's part of Compaq's fleet. Compaq has announced that it's as committed to selling Alpha systems as Digital was, and I hope that's accurate. The speed of the 533MHz Alpha I've used for the past year is nothing short of fantastic, regardless of whether the machine is acting as a workstation or a server. Compaq would be crazy to stop making Alphas. They're great machines.

However, Alpha applications are scarce. NT and BackOffice come in Alpha flavors, but precious few other programs do. Microsoft offers Word and Excel for the Alpha (at outrageous prices), but there's no Alpha version of PowerPoint. Digital tried valiantly to fill the application gap with FX!32, a cool product that transforms Intel Win32 applications into Alpha NT applications. FX!32 emulates Intel opcodes by searching for Win32 function calls, then rewriting the function calls it finds as Alpha Win32 function calls. Because most NT applications spend much of their runtime calling Win32 functions, an Intel application is fast after you run it through FX!32 because it spends most of its time running native Alpha code rather than emulating x86 code. (For more information about FX!32, see Brian Gallagher, "FX!32," April 1998.) In this respect, FX!32 is terrific—when it runs, it runs fast. However, I've had problems getting particular Intel applications to work under FX!32. Sometimes the software accidentally stuffs Intel files into folders in which NT expects Alpha files, so when I install certain applications on my Alpha, I have to reinstall NT. FX!32 is a good solution to the scarcity of Alpha NT applications, but it's not a great solution.

A Call for Life Support
What does the Alpha need to survive? More support from Microsoft. The Alpha won't become popular until the chip can break out of the vicious no developers will write code for a processor with few users and no users will buy a processor with few applications circle. Microsoft's Windows Logo Program and developer conferences give the company a great deal of influence with developers, but Microsoft hasn't used either forum to further the Alpha. At the October 1998 Professional Developers Conference in Denver, I didn't see one Alpha-oriented session.

If NT is to be a multiple-architecture system and not another Wintel platform, it needs the Alpha for both speed and credibility. Xeons are nice, but Alphas mop the floor with them performancewise. And if Microsoft wants users to embrace NT as an enterprise architecture, it needs to run the OS on a big-horsepower hardware platform. NT needs Alpha, and Alpha needs more Redmond support.