The irony of Compaq's return to OEM roots

Ambrose McNevin looks at Compaq's decision to put down the Alpha.

Texans, as illustrated by George W Bush, are not renowned for their sense of irony. But they are clearly a rum lot over at Compaq.

The Houston company's decision to abandon the Alpha platform in favour of Intel's Itanium is a huge fillip for companies such as Hewlett Packard (HP), which threw its lot in with Intel years ago.

Ironically, this has helped to develop the very Itanium chip on which Compaq's high-end servers will now run. It also raises a question over the future of Sun's Sparc chipset, but that's another matter.

When HP decided to port its HP-UX operating system to run on Intel, it effectively put a date on the death of its own PA Risc chipset. At the time many people reacted with horror. This was a bold move because it wasn't clear how the chip market was going to evolve. It was a gamble that users would trust a jumped-up desktop chipset on which to run their mission-critical systems.

But Intel has done a fine job of buying mind share and the people who were buying the servers turned out to be those who had bought Intel for their desktops.

Compaq inherited Alpha, the first true 64-bit chipset, when it bought Digital Equipment Corporation (DEC), after the firm had spent billions developing it. But neither firm could find enough original equipment manufacturers (OEMs) to opt for the platform.

It is widely believed that DEC's attempts to develop chips are what eventually brought the company down, because it tried too early to prolong a world where every disparate sector of the IT industry was proprietary.

At the time of the Alpha chip launch, DEC thought it had Microsoft on its side and that it would offer the proprietary chip set of choice for Windows NT, which was promised to be the de facto standard server operating system.

There was a great deal of excitement that Wolf Pack, or Windows NT clustering, running on Alpha was going to be the ultimate in client/server architectures.

Notwithstanding the dubious notion of naming an IT system after a fleet of Second World War submarines, Windows NT clustering never took off. The Alpha never achieved breakthrough, but both the resellers that shipped it and the customers that bought it couldn't find praise high enough. And don't the ironies keep coming: clustering is set to make a comeback.

Even after Compaq took over DEC it continued for a while with Alpha-based mid-range servers until it became obvious that it couldn't compete against IBM, Sun Microsystems and Intel/HP in the chipset market.

At the high end of the scale, users of Compaq's Himalaya range of fault-tolerant super servers (basically two mainframe computers strapped back to back), which have run on Mips chips since before Compaq bought Tandem Computers, will also have to get used to running their systems on Intel chips.

After a few years of attempting to make itself a 'we own this technology and can tell you where it is going' hardware player, Compaq has more or less returned to its roots: being an OEM for Intel and Microsoft.

All of this is part of a plan for Compaq to concentrate on software and services. So it is probably just as well, if not a little ironic.