Last week, we feared the worst. We feared that IDC - in a bid to tout Itanium for customers HP and Intel - had completely ignored the major issues facing the chip in the coming years. We feared that IDC's latest forecast for the server marketplace had once again missed the mark, continuing an ignominous seven-year tradition.
As it turns out, we were wrong. The critical part of IDC's Itanium assessment does exist, only it was written by the much smaller analyst firm Clabby Analytics.
Regular readers will remember Clabby Analytics and its chief prognosticator Joe Clabby. It was he who was touched by St. Fister in 2004 and cured of all Itanic wickedness. Itanium's apparent 2004 sales ramp and ISV embrace had turned Clabby into an EPIC believer.
Well, Clabby believes no more.
Where IDC championed the recent $10bn fund raising effort by the Itanium Solutions Alliance (ISA), Clabby, in a fresh report, has questioned how effective such an investment will be. He also urged readers to consider what Dell and IBM's abandonment of Itanium means to the chip's future, how ever improving x86 64-bit chips will challenge sales and how the RISC advancements made by IBM and Sun Microsystems will affect Intel's 64-bit dynamo.
Detractors will say that Clabby received funds from IBM for his initial anti-Itanium report and then got more funds from HP for his 2004 pro-Itanium report and is now back in IBM's pocket again with the new diatribe against Itanic. Frankly, we don't care.
IDC counts Intel, HP and IBM as its customers as well. The difference, however, between Clabby and IDC is that one of the firms can admit when it's wrong, and one of the firms can consider all sides of an issue. That firm happens to be Clabby Analytics.
IDC, by contrast, seems to have trouble coming to terms with the reality of the server market place. For example, it once predicted that Itanium server sales would reach $33bn in 2002. Then, it lowered that forecast slightly a couple of years later to say that Itanium servers sales would hit $28bn by2004. Well, 2004, came and went with Itanium server sales worth $1.4bn.
Most analyst firms would consider that a disappointing result. Not IDC.
"Purchase intent is fairly strong," the analyst firm said in its most recent report. "The current market perceptions of Itanium-based servers are quite positive." "Future Itanium server purchase intent is also quite optimistic." "Satisfaction among current Itanium customers is high." "IDC believes that the formation of ISA may ultimately prove to be one of the most significant elements in driving future sales of Itanium servers." And, of course, "Consistency in messaging will be key to future Itanium adoption."
Nowhere does IDC address its huge, past mistakes predicting the shape of the server market. Nowhere does it address how Dell and IBM's decision to distance themselves from IA-64 might affect the "consistency in messaging" for Itanium. Nowhere does IDC even discuss how improving x86 servers will compete in the 64-bit market. Instead, you're told that SGI, Bull and NEC will save the processor, and you're provided with gems of wisdom such as, "the vendors most likely to succeed in the Itanium ecosystem are those that will be able to articulate a cohesive message around delivering solutions to solve business problems, rather than delivering the best-performing Itanium server." (And here we thought the companies most likely to succeed with Itanium would be those that could raise more money from the French government or find suitors for their graphics technology.)
Clabby, meanwhile, does address all of the crucial points.
"While Itanium floundered, Intel has introduced another 64-bit architecture (the hybrid 32-/64-bit Xeon) that is meeting with solid market acceptance in the 64-bit computing space," Clabby writes. "This chip now presents another way for Intel customers to get into the 64-bit world without having to significantly alter their existing applications using the EPIC pre-compile method – and, the new chip has, accordingly, undermined Itanium acceptance."
"To compound these problems, Intel has continued to miss its Itanium roadmap targets - pushing back its first dual-core Itanium model (code-named Montecito) from 2005 to mid- 2006 (and delaying related functionality such as power management too)."
IDC does not even acknowledge the roadmap slips.
On the competitive front, Clabby points out that both IBM and Sun have dual-core RISC chips already. In fact, IBM has been shipping a dual-core product since 2001.
The analyst notes that IBM's Power and Intel's Itanium have stood as the clear performance leaders, putting Sun to shame on most occasions. Clabby then goes on to point out that Sun has endured product slips and that IBM has been forced to push back the release of its Power6 chip. This raises obvious questions about whether or not Intel's Itanium delays will really hurts its position in the competitive landscape.
Clabby addresses these questions by looking at how much actual investment Intel puts into Itanium. Like other members of the ISA, Intel pledged to funnel millions into boosting interest around the chip. And yet it has let Itanium products slip way behind competitors, while rushing new Xeon products out ahead of schedule in order to try and stay competitive with AMD. This strategy makes sense, given that Intel relies far more on Xeon than it likely ever will on Itanic. Unlike years past, Intel has fewer resources to spare on pushing Itanium forward given that it's still trailing AMD on the x86 technology front. This likely means that Intel won't make Itanium a top priority, and that's bad news for Itanic server vendors, according to Clabby.
"For vendors that sell high-end Itanium-based SMP solutions, constant roadmap schedule slippage and functionality slippage puts these vendors in a competitively disadvantageous position (because Sun and IBM own their own microprocessor development and can assign resources as they deem necessary to meet competitive challenges)," Clabby writes.
"Failure to own microprocessor resources gives leading HPC vendors HP and SGI little control on when they get their next respective spins of Itanium. Meanwhile, IBM and Sun can exploit this to their respective advantages - assigning even more resources if they so desire to pull way-out-in-front of Itanium."
Even on the ISA front, Clabby remains a skeptic and remarks that members of the club "would be loathe to open their books to show us what they're spending now - and what they plan to add to that spending." The idea there being that not a whole lot of "new" money is coming out of the ISA. (Contrast that with IDC, which took the $10bn pledge at face value.)
To Clabby's point, the financial backing of Itanium has long been murky. HP has tossed out an unknown quantity of gear for free to ISVs willing to port their code over to EPIC or to big name customers willing to give the Itanic a try. Myriad software funds have been started up to encourage the likes of Oracle and BEA to embrace Itanium. HP has also made its own $3bn pledge to Itanium but never clarified if this money has already been earmarked as part of the overall $10bn ISA commitment.
You can also look at "customer wins" such as the massive Columbia system at NASA Ames. The street value of the SGI Itanium-based cluster with more than 10,000 processors goes well past $200m. Yet, NASA paid only $50m for the box, as Intel and SGI fronted the rest of the costs hoping to brag that Columbia was the fastest computer on the planet. They were crushed when IBM's BlueGene system took the honor instead.
Can you really believe that the ISA "pledge" will improve Itanium sales that much given these massive, past investments?
IDC would do its clients a service by reading over some of the information in Clabby's report and then maybe regurgitating the highlights. Another suggestion would be for IDC or Clabby to investigate Itanium as an expensive but worthwhile sideshow. Imagine where IBM's Power processor might be today if the company didn't have to worry about Itanic. IBM could possibly have pushed its Power.org agenda forward earlier and tried to march Power into a broader ranges of servers, computers and devices. And let's not forget that Itanic killed off PA-RISC and Alpha.
We can't help but wonder sometimes if Intel didn't trick HP into taking quite the wild ride with Itanium. Intel, after all, has largely survived the Itanium era in style, making money off Centrino and Xeon. Meanwhile, HP has struggled to align its Unix business around Itanium. We'd love to see IDC or Clabby tackle these issues.