Why IBM could win the processor wars

COMMENTARY--Ordinarily, foretelling the future requires invoking mystical chants. However, there may be just enough physical evidence available--and with a little help from Skip Garvin, VP of business development at Sector7, a premiere application migration service with several fingers on the pulse of the industry--to evolve an intellectual answer as to which CPU will play the most important role in your future.

AMD is MIA on the IT radar. Although the company has played a much-valued role as Intel's jockey for the last few years, its CPUs are currently not in contention for any significant IT penetration. AMD realizes this shortcoming and claims to be working on improving its corporate image, especially in light of the 32/64-bit Hammer processor on the near horizon. For now, however, its products own the coveted title of "Preferred CPU of teenage overclockers." If AMD can shed that honor and place its low-powered CPUs in portables and blades while infiltrating servers with Hammer, things could change rapidly. That's a big "if," and AMD may indeed prefer to wrap itself around the consumer market while Intel branches out.

Intel, on the other hand, has always been very aggressive about IT processors. The Pentium Pro, Xeon, low-power Pentium III, and now the IA-64 are all part of its arsenal. But there have been some misfires. The Pentium Pro had an SMP problem that initially limited it to 2-way systems. The low-power Pentium III CPU is the darling of the up-and-coming blade market but they lack the high-end performance of the Pentium 4. And the IA-64 (or IPF, for Itanium Processor Family if you ascribe to Intel's attempts to de-emphasize the CPU's nearly exclusive 64-bit nature) may turn out to be a bit too radical for most IT departments for the next two or three years. According to Skip Garvin, "[Itanium] will survive long term, but it will be several more years before Itanium chips can offer the functionality that VMS can offer on Alpha chips today." And Intel's market has never really been the high-count SMP crowd.

How can you argue with Sun? If you do, Scott McNealy will probably yell at you. Perhaps the most outspoken corporate head since Lee Iacocca, McNealy has nothing but good things to say about his company. He probably has a reason to be proud. Sun leads the pack in IT sales in the U.S. and its Sparc products are, for now at least, the most flexible offerings you'll find. Additional Sun/Solaris support from companies like Fujitsu help imprint the brand name on the IT mindset. McNealy also has a reason to be edgy. Sun is running behind on its own timeline in processor development and, as every fox will tell you, that gives the hounds a chance to catch up. Right now Sun leads by sheer momentum. It needs to accelerate and that's a costly thing to do when the economy is singing the blues. Sector7's Garvin says the company might even be in the beginnings of a nose dive: "Sun just lost its number one market position in Europe and dropped Solaris and Intel support, betting everything on what we see as a declining architecture."

Conservative, pin-striped IBM may well emerge with the CPU of choice. According to Garvin, "We think IBM is positioned to be the leading supplier of high performance UNIX-based servers with AIX and the intro of the Regatta systems. They are head and shoulders above anything HP and Sun can offer today." Regatta (Power4) is, in fact, the key to IBM's future plans. Its design began in 1996 as a countermeasure to Sun's future plans for Sparc. Along the way to release, according to IBM Fellow, Ravi Arimilli, the team evolved more than a few features that couldn't be implemented within the time-to-market deadline. There are design enhancements waiting in the wings that will substantially increase the Power chip's desirability, and much of the basic development has already been done.

There are also some consequential benefits. Again, according to Arimilli, thanks to its highly integrated and innovative design, the current Power4 can be shaped into an 8-way system with the performance capability of its competitors' 16-way architecture. Your first reaction might be, "Smaller size, less power, that's typical baseline TCO stuff." Think beyond the box. With so many software developers charging per CPU for their licensing fees, cutting the CPU count in half saves on software charges as well.

Quicker to market and with a lower cost--if somehow that doesn't seem like the IBM of old, you're correct, it's not. It's the result of traditional IBM mainframe philosophy being put into the hands of a new generation of young turks. One of the nice things about being young is that it should be a while before you see a pin-striped suit in the group.

What's your take on the processor wars? E-mail me or post your thoughts in the Talkback below.

Bill O'Brien is a freelance writer and frequent contributor to CNET and ZDNet. He writes Tech Update's weekly hardware column.

Stay focused: Sign up for Tech Update Today, the daily e-mail newsletter for those who need to know.