Analysis You'll read a lot this week about the wonders of Intel. That's because Intel is hosting a developer conference in San Francisco, and the press can't gush enough over the processor maker's wizardry at these things. So, before the hype machine starts, it seems like an appropriate time to remind everyone how fallible Intel has proved itself to be over the past two years.
Of all the true tech powerhouses, Intel tends to be the most professional geezer on the scene. Without question, Intel owns the smartest, classiest PR team. It knows how to market a product. It knows how to sell a product. It knows how to throw an event. Its management is calm, cool and talented. It makes money and then makes some more. Intel does most of the big things right and just about all of the little things right. It's a disturbingly efficient machine - the kind of organism only an engineer could truly love.
This nauseating list of positive attributes underscores why it's so touching when Intel flounders.
Over the past two to three years, we've watched as Intel scoffed at the idea of x86-64-bit chips, preferring instead to insist that Itanium wasn't a huge failure. Intel also brushed off the multicore hype from the likes of IBM and Sun Microsystems. All the while, Intel plodded along, upping the GHz of its products and wrapping them with more and more cache. Then, one day, Intel woke up with a fleet of tubby, super-heated antiques on its hands and had to read about AMD signing up IBM, HP and Sun as server customers.
The big dog slipped up and had to change course - quick. Problem is the processor game requires long turnaround times.
One of the marks of Intel's professionalism appears when the company tries to act humble. CEO Paul Otellini, for example, has admitted that Intel "had a fairly weak product line in 2004." Twelve months is a pretty long stretch on the Moore's Law timeline Intel cites so often. Then we find an engineer last week confessing that Intel cobbled together a dual-core chip just to save face and try and compete against AMD.
Given how tight-lipped Intel usually is, these are huge declarations.
Less savvy executives would come right out and say, "We relied too heavily on GHz. We thought this cache cramming thing would work for a couple more years. With all the resources on the planet and the best engineers, we failed to see the immediate, obvious trend in computing toward lower-powered multicore chips. We're the world's biggest and most successful chipmaker, and we were the last to adapt our technology to the realities of current processor manufacturing."
You can't very well expect an Intel executive to cough out any of those lines, but they're true.
Intel suffered from classic big company thinking. It embraced a working formula and couldn't change course while the world changed around it.
Rivals saw that processor performance was outstripping memory performance by greater and greater margins. They saw that single core chips were getting too hot to make their way into servers and even desktops. They saw that lower-power, multicore processors would be the wave of the future. IBM, Sun and AMD all marched toward these goals.
This year, for example, IBM's Blue Gene supercomputers took up half of the top 10 spots of the most powerful supercomputers. All of these systems run on lower-powered chips and are well suited to handling technical computing types of workloads. IBM's success came as a real shock to Intel, according to our insiders. Intel put quite a bit of money and effort behind trying to secure the #1 spot for the Columbia Itanium-based system at NASA but to no avail. IBM's Blue Gene boxes sit at #1 and #2.
Then, while Intel struggles to introduce Itanic and Xeon to the dual-core era, IBM, Sun and HP have pumped their RISC server lines full of dual-core products. The same vendors also have dual-core Opteron parts from AMD. This interest in Opteron helped AMD capture 10 per cent of the x86 server processor market for the first time, according to a recent study from Mercury Research. Meanwhile, Sun is expected to roll out its multicore Niagara processor late this year - months ahead of schedule, heaping more competition against Xeon.
As the wealthiest chip company in the world, Intel has the luxury of scoffing at these relatively minor milestones from competitors. And, over the past couple of years, Intel did plenty of scoffing.
At the Intel Developer Forum this week, however, the vendor will talk a lot about how it has followed rivals. There will be chatter about Intel's move to bring forward the release of a dual-core Xeon processor, shipping it late this year instead of 2006. The lack of such a dual-core part had become a major embarrassment for Intel and its lone Intel-only Tier I customer Dell.
Intel will also hype up its "new architecture" that makes use of mobile processor cores in multicore chips for servers and desktops. The idea with the new designs being that Intel can strap together multiple lower-power cores. Sound familiar?
Intel's competitors often have to cheer their own, future efforts with a giant megaphone to draw attention away from the chip king. In many instances, this leads to the likes of Sun or AMD overselling products they can't quite deliver on and then watching as Intel matches them. Over the past two years, however, Intel's rivals have largely shipped all they promised to customers and, in some cases, have produced better than expected processors.
This push from rivals resulted in a bad scene for Intel which didn't adjust quickly enough to changes in the processor industry and which is still married to detrimental bus architecture. Intel has been forced to talk about "pushing forward" dual-core Xeon releases and hyping a "new architecture" that won't arrive for many months in the hopes that it can appear to be on par with rivals. In actual fact though, Intel remains well behind the competition in a number of areas. Lucky for Intel, the best processor rarely wins.
But during Intel's extended lull, you'll notice that it actually made some progress. Yes, Intel has been forced to share IBM, HP and Sun with AMD. That's a big loss. Still, Intel has picked up Apple as as a customer. It has maintained Dell as an Intel-only shop. In addition, some of our most connected sources and traditional Intel doubters say the company's upcoming product line - 2007 and on - will send rivals home crying. It seems that 2007 will stand as the year Intel finally catches up to the current trends in the processor market by delivering sophisticated multicore designs and by reworking its memory subsystems. The question for competitors is how much more damage they can do to Intel in the meantime.
As the stories roll out this week about Intel's glory, remember that big companies can make big mistakes. Intel knows this lesson well. It's just damned good at ignoring its own follies and so is the public. ®