Intel is killing Itanium one comment at a time
It's for mainframes only now
Intel's erstwhile high-volume enterprise chip - Itanium - is now being positioned for the lowest-volume market of them all - the mainframe. In one of his first public interviews since being tapped as Intel's next CEO, Paul Otellini shafted the multi-billion dollar baby by declaring that Itanic's future lies with the dinosaurs. It's a revealing comment from Otellini, who only a year ago was touting the "year of Itanium." Now, as CEO, he may well be the one that finally pulls the plug on the chip.
"Long term, the architecture Itanium needs to aim at is [IBM's] Power line," Otellini told BusinessWeek. "We have nothing in our existing 32-bit line capability that can compete with Power. It's a very high performance line requiring liquid-cooling capabilities. The mainframe isn't dead. That's where I'd like to push Itanium over time."
Has it really come to this - the mainframe isn't dead? To his credit, Otellini may be doing all he can just to keep the thought of Itanium alive by being realistic.
In September, HP killed off its line of Itanium workstations - the products key for software development on the EPIC architecture. Now we learn that Itanium is not competitive for low-end servers either.
"For a while, we had ambitions to drive it down to two-way servers and workstations," Otellini told BusinessWeek. "It just doesn't work in terms of the economics of the low end of the industry."
Why not? Intel has never managed to make Itanium affordable. An HP source explains -
"Intel never dropped the price for Itanium," said a high-ranking HP insider. "This was a really rotten deal from the HP side of things. You just couldn't get it to volume or be competitive with that strategy."
But have the OEMs received the memo on this new positioning? If the economics don't support it, why in the world is Intel making not only standard versions of Itanium for two-way servers but also low-voltage versions of the chip for rack mount boxes and blades? Whatever fluff Intel was once handing out about the 64-bit versions of Xeon not eating into Itanium sales can now safely be ignored. Intel has firmly relegated the chip to the high-end of the server market.
Customers and investors are to understand that the "twenty-year" architecture that Intel and HP spent billions on has been condemned to the role of mainframe contender. Intel and its OEMs will take on IBM in a market where it enjoys something pretty close to a monopoly. Intel and its OEMs will do this as software makers - little companies like Microsoft - pull away from the processor.
At this point, you have to wonder if Itanium will even work as a simple PA-RISC replacement. The messaging around Itanium has, just in this year, gone from "all is well" to "its future lies on the mainframe." And Intel only has itself to blame for Itanic's fate.
It tossed a mediocre first-generation product on the market, and big name vendors like IBM and Dell quickly pulled away from the chip. It made up some ground with the third-generation Madison, but now customers will have to wait until 2006 for a dual-core, fourth-generation chip. Meanwhile, IBM and Sun Microsystems are pumping out dual-core products that double the processor count of their servers with ease. Think about what this means as software makers inevitably start adjusting their licensing models to count dual-core chips as one processor.
Beyond these failings, OEMs - namely HP - have every right to despise what Intel has done with Itanium. Intel and HP did try to woo customers with steep, as in free, discounts on Itanium systems, but this strategy only gained some big name customer wins. You can't give away product to everyone. Intel needed to lower Itanium's price and take some financial hits in order to generate volume for the famed "Itanic ecosystem."
Intel has now made matters even worse by upping its focus on the 64-bit Xeon product, pulling engineers off of other efforts (Itanium included?). This can't be welcome news at HP where the management decided to bet the company's server future on Itanium. HP will surely move more Xeon-powered servers than any other vendor, but it's looking less and less competitive at the high-end where the real money is.
Canceling Itanium would be little more than a pride swallowing exercise for Intel. It would be a total disaster for HP.
You can't help but think Otellini's willingness to turn so publicly on the chip only adds to the notion that Intel is and has been Itanium's worst enemy. ®