On the same day that Intel met with the press to discuss the bright future of its Itanium processor, analyst firm IDC cut its sales forecasts for the chip, adding to a decade long slump for Intel's great 64-bit hope.
Mike Fister, server processor chief at Intel, held a "state of the union" talk about both the Itanium processor and the far more successful Xeon chip, giving journalists an idea of what is coming in 2004 and beyond. Ever the optimist with Itanium, Fister reiterated the belief that Itanium will thrive in the server market and one day eclipse Xeon in sales. IDC, however, undercut such predictions, saying Itanium will face increasing pressure from x86-compatible 64-bit processors like AMD's Opteron - a notion everyone but Intel seems to support.
In the near term, IDC thinks Intel will benefit from the third generation Itanium - Madison - but this boost will be short lived. The company cut its 2007 forecast for Itanium sales down to $7.5 billion from $8.7 billion. Why IDC looks so far out is anyone's guess, as their track record with Itanium isn't terribly spectacular. In 2000, IDC predicted that 2004 Itanium sales would hit the $28 billion mark. Whoops.
But after plugging away at Itanium for so long, it's hard for Intel to give up on the chip now. So what's coming in the Itanium pipeline?
First up, HP will start shipping its MX2 module this year, which puts two Itanium 2 processors together. This mimics a dual-core processor to some degree and helps HP keep up with RISC vendors such as IBM and Sun Microsystems.
In addition, Intel's revamped IA-32 Execution Layer technology will go into production this year and be available in Microsoft's Windows Server 2003. This makes it possible to run 32-bit code on Itanium chips at a rate equivalent to a 1.5GHz Xeon. Intel expects the performance to increase over time, topping out at about 50 percent to 70 percent of native Itanium performance.
This is nice answer for customers trying to move over a few 32-bit applications but does not compare favorably with the performance of AMD's x86-64 line.
In three years time, Intel expects the performance of its Itanium chip to be about 7 times that of the Madison product. With Tukwila - formerly Tanglewood, Intel will place twice as many cores on the chip as it does Xeon chips available at the time. This Tukwila design done by former Alpha folks is Intel's major hope for reviving the Itanium line.
One last bit of interest on Itanic will be the introduction of server blades based on the chip by 2005. It's clearly taking quite a while to shrink and cool the beast.
On the Xeon front, Intel plans to bring out a larger cache version - 4MB of the Gallatin chip this quarter along with a new Xeon targeted for the dual processor server market. As part of its IBM partnership, Intel will also release a four-processor Xeon blade server in the first quarter. IBM recently announced its own blade system.
Overall, Intel is delivering ever faster Xeons and complementary chipsets that increase front side bus performance and continue the chip's strength in the 32-bit market. The 64-bit market, however, continues to plod along at a very non-Intel pace.
Recent data from both IDC and Gartner suggests that the server market is seeing signs of life. In particular, customers are picking lots of one, two and four-processor boxes. This is prime Xeon territory, and upgrades here could begin to harm Intel.
With memory prices falling at a fairly dramatic rate over the past ten years, it's only a matter of time before customers stop upgrading from a 32-bit box to a faster 32-bit box and instead to a 64-bit system. AMD is providing the best means to pulling this upgrade off, which is why the analyst community is rewarding the company with praise and why Opteron sales have shot right past the much more mature Itanium.
Some analysts expect Intel to fight off AMD by releasing its own 64 bit x86 chip this year. Others, however, like Nathan Brookwood at Insight 64 doubt such a chip would arrive so soon.
"Like still waters than run deep, it has been difficult for Intel to demonstrate that Itanium has attained any marketplace momentum," Brookwood wrote in a recent note. "They hope to communicate this message during 2004 through the steady drip-drip of press releases that announce this or that major organization has installed Itanium-based systems as a key part of its computing infrastructure. Unless and until Intel cements Itanium's image as a successful product line within its intended high-end markets, any action that calls attention to alternate Intel 64-bit strategies could be harmful to Itanium's long-term prospects. Insight 64 believes this phenomenon alone will push an Intel 64-bit x86 into 2005 at best . . ."
These squabbles over exactly when Intel will ship an x86-comptiable 64-bit product sidestep the major issue, which is that analysts have largely reached a consensus that Intel can and must do something in the space.
At this stage, Sun and IBM can pick up all the early Opteron adopters they like without any competition from HP and Dell. The vast majority of these Opteron wins will come from customers moving off of Xeon - not those shelving RISC. How long will Intel and its two favorite OEMs allow a squeeze of Xeon from the bottom and a crushing from the struggling Itanium on the top?
It's hard to imagine things continuing along this path for too many months. ®