What Happened to Alpha?

In the early 1990s, when David Cutler and his team began to design the next-generation Windows OS, creating an OS that was portable between platforms was a principle design goal—a goal the team borrowed from the UNIX playbook. The team imagined this new technology, which they called Windows NT, as one that could span the entire computer industry. To achieve this flexibility, the team adapted many design concepts from team members who had worked at Digital Equipment on VMS.

To port the system, Cutler's team wrote the microkernel, then compiled the first version of NT for the MIPS processor. The team compiled the second version for the Intel x86 processor family, and the third version ran on Digital Equipment's Alpha processor. The team also worked on an NT version for the PowerPC platform. Microsoft abandoned support for the MIPS and PowerPC processors, but NT has supported Intel and Alpha since the OS's inception—until now.

In late August, Compaq announced that the company will no longer sell Alpha for the NT platform. Rumors circulated that Compaq had laid off 120 engineers who had been working on NT and Alpha at the company's DECwest research facility, but that wasn't the case. A few of the engineers are finishing Service Pack 6 (SP6) for NT 4.0, and Compaq reassigned other engineers to the IA-64 (i.e., Intel's processor architecture) team. Compaq will continue its 64-bit Windows development, but only in terms of developing Windows 2000 (Win2K) for the Intel architecture. So, Compaq will continue to support its current NT-on-Alpha customers, but no future versions of Windows will run on Alpha. Nearly 2 years into Win2K's development, and only a few months before Win2K's scheduled release, Compaq has pulled the plug. This decision caught many people in the NT community by surprise.

Microsoft took about a week to respond to Compaq's announcement. When Microsoft did respond, the company stated that it was discontinuing development of Win2K for the Alpha processor family. The discontinuance was for 32-bit and 64-bit Win2K support. So, any hope of a rescue, or for other companies with licenses to produce Alpha computers (such as Alpha Processor) to carry on production, is gone. From this point forward, Microsoft will design Win2K for only the Intel architecture. The loss of hardware diversity for Win2K isn't good news, and Compaq is taking a lot of heat for this decision. The discontinuance of Win2K support for Alpha sounded like a Microsoft decision, but a senior Compaq strategist said Compaq made the decision: Microsoft realized that even if Win2K for Alpha shipped, Compaq wouldn't sell Alpha systems for Win2K.

In Alpha's long history at Digital Equipment and more recently at Compaq, NT-on-Alpha sales have represented a tiny portion of NT hardware purchases. In current reports, Compaq said NT accounts for about 2 percent of AlphaServer sales. (UNIX accounts for the overwhelming majority of AlphaServer sales.) Terry Shannon, longtime Digital Equipment and Compaq watcher and the editor of Shannon Knows Compaq, said more than 500,000 Alpha systems have been sold, but only between 10 and 15 percent of those systems run NT. When you compare these numbers with the 40 million computers running NT, you begin to appreciate how poorly NT on Alpha was doing.

What happened to Alpha? Alpha systems have always suffered a price-for-performance penalty when compared with Intel systems running NT. This penalty was a major factor in Alpha's poor sales. The other major factor in poor Alpha sales was that a scarcity of software exploited the power of the Alpha. With Compaq's recent introduction of Intel 8-way Profusion servers, Compaq management realized that Alpha sales muddied the water for their sales folk and caused them to compete against one another.

The Alpha processor's considerable performance edge over competitive Intel microprocessors made the processor significant throughout its history. This performance advantage was because of an innovative RISC design, the elements of which illegally ended up in Intel's architecture. When Digital Equipment sold its Alpha chip facility to Intel, the companies reached an internal settlement on the RISC design dispute.

Intel shipped the first prototypes of its 64-bit Itanium (formerly code-named Merced) chip to its partners in September, but Alpha already ran 64-bit software. At the time that Compaq was abandoning further 32-bit Alpha development, the company hoped to continue development for 64-bit Win2K on Alpha. Rumors suggested that Microsoft was contemplating releasing 64-bit Windows on Alpha first. And while Compaq and Microsoft were making the decisions about the future of 64-bit Win2K on Alpha, Cutler's team was booting the first versions of 64-bit Win2K on an AlphaServer (a classic Allan Janus project).

The story emerging is that after Microsoft heard that Compaq would no longer develop Alpha for the 32-bit versions of Win2K and NT, Microsoft lost interest in developing the 64-bit version of Win2K for Alpha. Microsoft felt that Alpha couldn't be commercially successful without continued 32-bit development. And if Alpha wasn't successful in the 32-bit market, how could it be successful on the 64-bit OS?

On Compaq's side, the company probably wondered if other companies would ever commercially deploy 64-bit Win2K on Alpha. Compaq surmised from its discussions with Microsoft that commercial deployment would have depended on when 64-bit Win2K became available, how Intel's architecture was selling, and whether Alpha sales had picked up. So, Compaq faced the possibility that when 64-bit Win2K became available on Alpha, the company wouldn't have enough of a market window to exploit the 64-bit OS on Alpha against the 64-bit OS on Intel's x86 platform. Most likely, Compaq balked at producing 64-bit Win2K on Alpha systems because the company assumed it couldn't sell the Win2K-on-Alpha systems.

The marketplace might have spoken regarding NT on Alpha, but probably not. The vendors greatly contributed to the poor sales of NT on Alpha. Microsoft, preferring to remain a neutral party, didn't advocate earnestly the benefits of Alpha. Digital Equipment and Compaq didn't price and market Alpha correctly.

To fully demonstrate Alpha's advantages, Microsoft needed to deliver a 64-bit OS that could employ the chip's capabilities. Internal memos presented at Microsoft's Department of Justice (DOJ) trial expressed Intel's impatience with Microsoft for the slow pace of its 64-bit OS development. To imagine that Compaq felt the same way is easy.

Despite Compaq's abandonment of Win2K on Alpha, the company continues to express its support for the development of other 64-bit OSs on Alpha. Compaq will makes its Alpha development effort in Tru64 UNIX. And in anticipation of a 64-bit Linux release, Compaq will also target more Alpha sales toward Linux. Compaq is in the middle of porting its language tools over to Linux to aid in this development effort. Although the measurable Linux volume on Alpha is small, Compaq and Alpha Processor are turning Alpha into a Linux play. This opportunity shouldn't have arisen, and Microsoft probably isn't happy about it.

Compaq continues to express support of Alpha to its staff and business partners. Compaq assured Samsung, which is Compaq's Alpha chip partner and backs Alpha Processor, that the company remains committed to selling Alpha systems in volume, which is the task at hand for Alpha Processor.

Enrico Pesatori, Compaq's senior vice president and group general manager of the Enterprise Solutions and Services Group, recently circulated an internal memo to the team. The memo stated that the decision to stop development of Alpha for Win2K in no way diminishes Compaq's strong partnership with Microsoft or the company's commitment to Alpha. Pesatori said, "We continue to invest in Alpha as a core component of our NonStop eBusiness strategy, including next-generation Alpha chip technology and a robust Alpha systems roadmap. We will drive Alpha at the high end of the enterprise market in which our strengths in 64-bit platforms, NonStop technology, and clustering give us a competitive advantage."

None of the factors leading to Compaq's decision are new; the same situation was true of NT on Alpha a year ago. So, I'm left to ponder Alpha's late departure from the Win2K game. I'm drawn to the conclusion that Compaq's decision boiled down to a financial punt. Compaq just ended two losing quarters and needs a couple of quarters to turn the company's financing around. Unless the numbers improve, Compaq will need to cut into the services group to improve its margins. Ultimately, Alpha on Win2K was thrown out of the game because its prospects weren't good.