Bill's Index

Processor Evolution Will Continue

Intel's Itanium has killed off some high end processor chips that were doing well before Intel's massive publicity attack that promised the moon. There is an article on what may have been Intel's aim - kill off competing high performance processors - in The Inquirer.

There is much more to this story, in three dimensions. One of those dimensions is what the future of Itanium might be if it is not simply phased out. The second dimension is what will happen as the playing field is narrowed to x86, very high end processors (VHEP), and 'Others'. The third dimension is simply: Who will survive?
Dimension One: Itanium's Future
In addition to the long and expensive gestation of the Itanium, Intel has something else invested in Itanium: Reputation.

Intel has made a point for nearly ten years that Itanium will be the most powerful chip available, period. They have struggled for years to reach that level and are just reaching the edge of that region. Does it make sense for them to discard that investment of time, money and reputation? I think not. In fact, there are three good reasons for Intel to continue to expend effort on Itanium.

The first reason is to not lose the installations and reputation they have already won. Even though past failures like the 432 and the 860 did not harm Intel, dropping the Itanium would. They could have dropped it before full production, even possibly before the second version.

But now, with systems in the field, promised upgrades and a road map extending to 2006, Intel could not afford the hit to their reputation for delivering on promises. It's just too late to stop. Their only alternative is to push Itanium technology to meet their ambitious goals. Given Intel's resources and their need, I expect them to deliver, regardless of cost.

A second reason for Itanium is to compete at the VHEP level with a proprietary chip that will lock users in, against IBM's Power architecture and possible competition from a Sun-Samsung SPARC alliance. Alpha, MIPS and PA-RISC are gone for good unfortunately, but AMD could be a dark horse in the VHEP arena. This battle is nowhere near over.

Specifically, I expect that Itanium will become a Power killer at the very high end SMP and cluster range for business and scientific processing. I think Intel will further specialize Itanium to that area and eliminate most or all x86 support circuitry, depending on better emulation where x86 is needed.

The third reason for Itanium is profit. Keep in mind that the billion dollar investment in Itanium is already spent. The only way they can show a profit is to sell enough Itanium chips and systems to exceed that year's cost of producing them. Whether they ever make enough to offset the total cost is almost irrelevant because that money is already gone. What's left is whether money can be made in the future.
Dimension Two: Future x86 Processors
Another Itanium effect will be that the reduction of mid-range competition. AMD and Intel will go head to head in x86 and the 64 bit version, but I expect further x86 extensions to enhance their performance potential. It is even possible for those extensions to become conflicting despite Microsoft's wishes.

Notice how the x86 processors are becoming a whole environment by themselves. There are already several classes of x86 processors that increase in specialization with each generation. This is because of the huge volume of x86, which can justify the investment, and the broad array of software that supports this architecture.

This x86 architecture is delivered in a bewildering range of overlapping implementations. It delivers chips from the extremely low power ULV and emulated x86 systems, very low cost Cyrix class and embedded x86 chips, LV and normal notebook chips, high performance notebook chips, desktop and high performance desktop chips to the high end server chips for blades and 2X, 4X, 8X and larger SMP machines.

Only Intel can cover that whole spectrum, though AMD will cover all but the low end. Cyrix and others will avoid the mainstream where standards and volume are crucial to profitability, and concentrate on the specialist chips. Ultimately these specialist chips will push out almost all the other low end chips. Economics and software support will ultimately determine who the winners and losers will be.
The Third Dimension: Survival
The microprocessor environment has a huge set of predators. Survival in this environment requires brains and agility, plus extremely focused attention on specific markets and competitors. No company will survive a major mistake, such as late delivery or entry in the wrong market.

There has already been consolidation in the x86 market. The most recent was when AMD bought National's x86 product line. There will be more as design and marketing costs increase while chip prices decline. Current and future design wins and losses will indicate which company acquires, and which are acquired.

The 'Others' category will remain. There will continue to be specialist requirements that are too small or demanding for an x86 processor to make economic sense. Examples will include radiation hard chips, extremely small pico class processors, highly secure processors and special combinations like the programmable Xilinx chips that contain both embedded processors and customizable circuitry.

One likely survivor will be the widely used ARM architecture, now in its tenth revision. This one just keeps going, and going, and going. Xilinx will be another, but beyond that, the future is uncertain. Even the soundest company can flounder if they fall behind or make a major blunder.

Even Intel could stumble. If Intel cannot get the power demand of the Prescott chips under control and AMD keeps upping the ante with Opteron and it's K9 successor, AMD's market share could match Intel's x86 market share in a few years.

The smaller companies have far higher risk levels. These risks exist in three areas: Size will be less important than where companies stand in their niche. Unless they are first or second in a niche, trouble may be ahead. In addition, if their niche grows enough to interest one of the major players, acquisition may be the best or only alternative.

Yet in the face of these problems, small companies will continue to succeed through innovation and specialization. Timing, execution and luck will always be major factors in their eventual success or failure.

The beneficiaries of this war will be those with the luck and foresight to invest money or employment in the winners, and the rest of us who use their products. This year's high end processors are next year's mid range, and a few years later, junk.

Eventually, physics will determine just how much smaller and faster microprocessors can get. Each time obstacles have appeared, research and new approaches have eliminated them. The only show stopper on the current horizon is power delivery and heat dissipation. That too will probably be solved. The only thing I am sure of is that the future will bring more surprises.

[30]

All content on this site is Copyright 2001 - 2004 by Bill Nicholls
All Rights Reserved