YouTube Facebook LinkedIn Google+ Twitter Xinginstagram rss  

By John Dodge

April 7, 2002 | Maybe Compaq, assuming it's still called that when you read this (at press time, HP's Fiorina was claiming victory, but it was not official), should reverse its decision to scrap the Alpha microprocessor line after 2004.

After all, the Alpha chip as the heart of the AlphaServer line has a long list of impressive life science wins. Celera John Dodge Genomics manages a 70-terabyte database with 600 interconnected Alpha processors; the Wellcome Trust Sanger Institute in Cambridge, England, uses 400 AlphaServer systems to operate a 20-terabyte database; and MIT's Whitehead Institute Center for Genome Research has an AlphaServer farm with 100 Alpha processors.

That alone could warrant salvaging the chip's future. The decision to kill it seems downright mad considering that Compaq has entrusted its high-performance computing business to Intel's new, largely unproven Itanium microprocessor line. If Compaq is taking the Intel route, it might as well endorse the proven Pentium 4 as its high-performance chip of choice.

Financial pressures prompted Compaq to announce this fateful decision about Alpha last June. Although cutting Alpha streamlines Compaq's three microprocessor lines into one and removes the company from the hugely expensive microprocessor design business, the measure all but cedes the top end of high-performance computing to those investing in new microprocessor designs. Predictably, Compaq says its high performance computing customers say they are fine with the decision. But a shift of this magnitude, even if one more major Alpha revision is coming between now and the end, keeps customers awake at night.

"We are very fond of Alpha," says Whitehead Genome Center CIO Jill Mesirov, who is resigned to the decision. "Our main issue is upward [software] compatibility. Compaq has [us] assured of it. All I can tell you is the jury is out. But the world changes and we have to change with it. It's not like there's some alternative path. Moving to another architecture would entail a lot of work for us."

While IBM is pumping $100 million into its Blue Gene project, which promises to create a 200-teraflop supercomputer, Compaq is opting for high-volume, off-the-shelf parts, a highly competitive market segment. For Compaq and HP, should they unite, it's just one less advantage to distinguish them from the pack. "Compaq has great wins, but they're abandoning Alpha," says IBM Life Sciences director of strategy Jeff Augen, reveling in a recantation undoubtedly made on IBM sales calls.

Compaq vice president and general manager of High Performance Systems Division Richard Marcella begs to differ, of course, claiming the decision to can Alpha was strategically rather than financially motivated. "The [silicon] building block is becoming a commodity. Differentiation will come from the software and other things higher in the stack," insists Marcella, apparently still bruised from Alpha's inability to win support from hardware builders and software developers a decade ago when Alpha's creator, the late Digital Equipment Corp., yearned to unseat Intel as microprocessor king.

"We never had the ISV [independent software vendor] portfolio that was necessary for Alpha," he says. "By going to Intel, ISV support comes along with it. That's more important than having a 10 to 20 percent performance improvement at the chip level." What's more, Marcella says that after the next Alpha revision, the architecture will have largely exhausted the advances that can be made to it technologically.

Get a Life
Undoubtedly, off-the-shelf computers will become a bigger force in the life sciences given the demand for computing power and storage capacity. For instance, Israeli firm Rosetta Genomics is building a 10-terabyte database using a Dell 8-processor server and Microsoft SQL Server.

On another level, IBM's and Compaq's organizational approaches to addressing the life science market could not be more different. IBM has established a 250-employee Life Sciences Solutions Group, backed by another 300 IT experts within the operating groups with "life sciences" in their titles. Augen's boss, Caroline Kovac, reports into IBM's Public Sector Group, which has estimated revenues of $11 billion and is headed by Curtis Tearte. Tearte is only one level of management removed from new IBM president and CEO Sam Palmisano.

By comparison, Compaq is a poor man's life sciences company. Compaq's High Performance Systems Division oversees the company's assault on what it calls the life discovery sciences, a euphemism for genomics and early-stage drug discovery. But on the organizational chart, the High Performance Systems Division sits 4-5 notches below the top of the company. And Compaq's very future as an independent entity is uncertain, with some of its groups and brands facing the inevitable result of any merger — extinction.

Compaq has neither the research arm nor the legions of life science doctorates that IBM has deployed. Its success in life sciences relies on Alpha and the Compaq Tru64 UNIX operating system, both of which came from Digital. If you harbor any illusions about Intel preserving any Alpha technology when Compaq is done shipping a few hundred Alpha engineers to Intel in exchange for drinking the Itanium Kool-Aid, remember that Digital sued Intel in a nasty patent suit involving Alpha a few years ago.It is difficult to buy Compaq's argument that an edge in microprocessors doesn't count much anymore.

The decision to ditch Alpha and Himalaya, the other Compaq microprocessor line being shown the door, was born out of financial expediency. Of course, this could become HP's problem, and some users worry more about the impact of the merger than Alpha's unlikely salvation.

But at the very least, the Alpha microprocessor and AlphaServer line established themselves as powerful IT brands in the life sciences. It's unlikely even the name will be retained.

Is Alpha worth keeping? Write to me at 

For reprints and/or copyright permission, please contact Angela Parsons, 781.972.5467.