Does anyone have a direct line to the Pope? We've witnessed another Itanic miracle and think it's time Intel's server processor chief Mike Fister begin his walk toward sainthood.
Joe Clabby, of analyst firm Clabby Analytics, is the third man to undergo the miraculous transition from Itanium basher to EPIC worshipper. In a recent note, Clabby professed to having "seen the light" after Intel executives spoke to him during a server processor presentation. Intel cleansed Clabby of all his Itanic wickedness and, in so doing, managed to elevate the Itanium processor to a religious status that far exceeds that of mortal chips.
"Only a year ago Clabby Analytics would have classified EPIC/Itanium architecture as a failure," Clabby writes.
The analyst points out that it took Intel close to ten years to bring Itanium to market, the chip cost billions in research and development, it had poor yields and poorer adoption and failed to capture the hearts of developers.
And then on the 3,700th day of Itanium's creation, Fister placed his hand on Clabby's head and all became clear - in an instant.
"But what a difference a year makes! In 2003 Intel got its Itanium act together, shipping 100,000 units while expanding its ISV offerings to over 1,000. And, in 2003, Intel managed to capture most of the leading performance benchmarks with EPIC/Itanium architecture. After almost a decade of research and development; after a re-spin of Itanium to produce Itanium 2; and after a major campaign to capture ISVs (including “seeding” the ISV community with funds to port applications to Itanium), Itanium finally established a solid foothold in the 64-bit computing market-place."
Praise the Lord!
The first man touched by "The Fist" was Merrill Lynch's server analyst Steve Milunovich - known at El Reg as "The Loon." Miloonovich once favored RISC processors from Sun, IBM and HP, but as the analysts around him began hyping Itanic, Merrill's top dog jumped to Intel's side, calling Itanium an "industry standard" when it held less than one percent of the server processor market - far less than one percent. The man who once favored differentiation succumbed to Intel's grace.
The second healing took place in Berkeley. For it was there that UCB graduate student Nick Weaver went from Itanium hater to Itanium lover overnight. Of all the Itanic miracles, this one was the most spectacular.
To fully understand Clabby's transition, we must travel back to July of last year. At that time, the analyst issued a scathing critique of HP's planned Itanium migration. HP would make life very difficult on its customers by shifting them to an immature processor that lacked the necessary software for serious business use, Clabby said at the time.
Clabby maintains his concerns about HP. The company's plan "to bet the farm" on Itanium is risky. How could shifting thousands of customers onto a new architecture be easy?
But for Intel the story is much brighter.
Clabby rightly points out that Intel has destroyed the competition on most benchmarks with IBM's Power processor the only rival even close to keeping pace. In addition, Intel has delivered a rich set of porting and migration tools to help customers move from RISC to EPIC. And last, Microsoft finally delivered code for Itanium in 2003.
Most importantly, customers have started picking up the third generation Madison chip. It's this processor that pushed Intel to 100,000 unit shipments and prompted the company to declare 2003 "The Year of Itanium." The 100,000 figure, however, ignored the fact that Itanium server shipments have yet to break the 5,000 unit per quarter barrier, relegating the chip to a rather insignificant place in the overall high-end server market.
Clabby is also bullish on where Intel plans to take the Itanium architecture. By the time Tukwila rolls out in 2007, Intel will have multiple cores per die and multithreading on the processor. This should give the chip about 7 times the performance of Madison and outpace performance gains that trend alongside Moore's Law.
But while Intel bills itself as being on the cutting edge of multicore design, the reality is that it is woefully behind IBM and even trails Sun. IBM paved the way for dual core processors with Power4 and is ready to deliver Power5, and Sun plans to have chips with tens of cores by 2007.
Clabby, however, notes that Intel is paying special attention to using EPIC to its advantage
"Intel's competitors are also building multi-core, multi-threading processors, but do not preprocess instructions before handing those instructions to a CPU for processing," he writes.
Er, we'll see.
Clabby is right that Intel finally started to show some momentum with Itanium in the latter half of last year. This helped breath life into a processor many people in the industry had given up on.
But other analysts have noted that this success may be short lived. Intel is already facing competition from AMD's x86-64-bit Opteron processor and may well start competing with itself should the fabled Yamhill come to market. Based on the emergence of x86-64-bit chips, IDC has lowered its 2007 forecast for Itanium sales down to $7.5 billion from $8.7 billion.
To add a bit of perspective here, IDC once forecasted that 2004 Itanium sales would reach $28 billion.
It seems a little premature to celebrate the second-coming of the Itanic just yet. ®