One thing Intel was able to demonstrate at the conference the week of March 6 was a tour de force of staying power.
Not only did the company reveal a kick-butt architecture that will put its processors—particularly the server parts—back on par with Advanced Micro Devices, but it also gave its partners a long view of technologies that are at the moment only a glimmer in the eyes of the companys researchers.
Now, AMD has made pretty good inroads with servers at the high end, sold a lot of desktops at the low end, and made even a bit of headway with mobile in retail. But Intel still dominates among enterprise customers.
What drives this continued enterprise loyalty?
The reason isnt just conservatism in the IT ranks—the “Youll never get fired for buying Intel” adage.
Instead, its that Intel can point far into the future with a reassuring message: “Come what may, were going to be there. We will push the frontiers of Moores Law and continue to bring ever-more powerful technologies to market.”
Intel can talk not only about a road map of two cores, leading to four cores, leading to eight and so on, but also the manufacturing processes that will produce line widths of 65 nanometers, dropping to 45 nanometers, dropping to 32 nanometers, dropping to 22 nanometers. And, more to the point, company engineers can tell you what year theyll be able to achieve these milestones.
This detail and its accompanying investment are reassuring to a large corporate customer with an annual IT budget in the hundreds of millions of dollars, who wants to hear in 2006 just exactly what Intel expects to produce in 2011.
At IDF, with the public still reeling from the idea that multithreaded programming can produce incredible performance gains by spreading work across multiple parallel cores all while saving on energy, Intel said, “But wait, theres more,” and laid out a slew of other improvements, both incremental and order of magnitude, that the company expects to implement in future processor generations.
With its new Core microarchitecture not even out of diapers, Intel started talking about Tera-Scale Computing, an architecture that will be far more massively parallel than todays—and even tomorrows—systems.
By exposing potential customers to this level of research and advanced thinking, the company demonstrates that it is in the game for the long haul, just exactly the message that plays well with risk-averse CIOs.
Of course, many issues remain to be resolved. Otherwise, these technologies would be in todays products. For one, the company has to achieve its process milestones in order to get the number of transistors onto a chip required by the futuristic architecture.
In addition, somebody has to figure out how to farm computing tasks out to the great stack of simple, general-purpose processors and the specialized processors behind them that will form the heart of Tera-Scale Computing.
But even there, the company has a track record. It creates tools and compilers to help software developers work with new architecture, essentially easing the way through higher-level programming. Enterprise buyers believe that these pipe dreams will actually materialize because they have before.
Intel can talk, in varying degrees of detail, about five generations of processors, all in motion at the same time. Thats quite a technological juggling act. And one that greatly impresses the intended audience.
Roger L. Kay is founder and president of Endpoint Technologies Associates.