Monday, October 31, 2005

IBM's Power processor versus Intel

In 2004, IBM booked less revenue in total hardware (which includes processors, servers, storage, and PCs) than Intel booked in processors alone (IBM 2004 annual report). What makes IBM think they can make the Power processor a successful competitor to Intel’s Pentium series of processors?

It’s true that IBM is looking to achieve a much broader adoption of Power technology than Intel has achieved. This ranges from embedded processors in storage devices, a Power processor driving every game console as of 2006, Power in small to large servers and even ‘powering’ the world’s fastest supercomputer. In a different light IBM has also taken a completely different approach from Intel. They have made the Power architecture an open specification - one that is, however, guided by a consortium of manufacturers including Freescale, Sony, and Cadence among others. This does not guarantee success.

SUN was one of the first large manufacturers to attempt an open-source style specification through the Java Community Process (JCP). It led to what many consider to be the feature-bloat of the language and some performance issues that Microsoft is not facing given their proprietary process. Microsoft, through greater control over the technology underlying dot-Net (including the OS) has arguably been able to squeeze better performance out of their design. Granted, IBM’s model is slightly different from SUN's JCP but is it different enough?

The jury is out on whether open source models for large systems will work in the long run. In the end, as Linux is showing us, large vendors want more say in how an open source asset is evolved over time. Who do you think is contributing the most code and is exerting the greatest influence over how Linux evolves? Is it the lone hobbyist / researcher or the team of developers working for some of the large corporations with a stake in Linux (Red Hat, IBM, HP, and Novell to name a few)? It’s not just about corporate clout but about the practicality of ensuring Linux’ advancement and maturation. That takes great developers and Linux has to somehow do it without paying any of them. When RedHat offers engineering help will Andrew Morton, the current Linux kernel maintainer, say no? This may reduce some of the positive effect the OSS model experiences in innovation. Competing interests vying for control over a common resource invariably weaken its power. The JCP has not led the Java language to innovate as rapidly as it should have, using Microsoft as a benchmark and the rapidity with which they have successfully developed and deployed dot-Net.

In a market such as the design of mainstream processors that run small to large servers, cash is king. Whoever invests the most in R&D, wins.