But while the panelists praised Intel's first 64-bit offering as "working better than anticipated" and capable of handling various applications, they offered no specific details on the chip's performance. In addition, some comments -- including an awkward moment of silence following one question -- highlighted some of the hurdles the chip has yet to clear to win industry acceptance.
Since December 1999, Intel has shipped more than 7,000 prototype Itanium systems to software developers and high-end system users around the world.
The large-scale deployment of the prototypes, largely provided for free, is aimed at both sparking rapid development and adaptation of software for the new IA-64 architecture as well as fostering greater acceptance of the chip.
Intel's new chip also has significant industry support, with most top computer makers promising to offer Itanium-based systems, the notable exception being Sun Microsystems, which is the sales leader in high-end Unix servers.
Itanium is Intel's bet to break into the high-end server market currently dominated by RISC-based processors designed by Sun, Hewlett-Packard and IBM. In fact, HP is co-developer of the Itanium, and has committed to adopting the chip in future servers.
While Intel's IA-32 Pentium III Xeon processors have rapidly gained acceptance in workstations and low- to mid-level servers, the chip maker's lack of a 64-bit product effectively shut it out of the more profitable high-end server space.
"Intel could only look and drool at the kind of money those guys were making," said Nathan Brookwood, principal analyst with Insight 64 in the US. "Intel could not touch that stuff with its 32-bit line."
So the company has been working since 1994 on the 64-bit Itanium, which is now set for release in the second quarter, following numerous delays.
Barries to acceptance
But the processor, which features a whole new architecture, must overcome a number of challenges to gain widespread acceptance.
Chief among them: concerns over the stability of the new and yet-unproven design, especially by companies looking to deploy the product in "fail-safe" servers. Another issue is whether the chip's architecture will be compatible with widely used 64-bit applications many companies already rely on.
Panel member Mahlon Stacy, who oversees the use of two Itanium systems at the Mayo Clinic, admitted that the early prototypes he received were unstable.
"The early A2s [among the first prototypes shipped] that we had would all of sudden just hang and we'd have to reboot them," Stacy said, following the panel discussion. "But those kinds of issues have been resolved since last November."
Yet while his systems are no longer plagued by such lock-ups, Stacy admitted they still run into problems running some applications.
"There are still little bits and quirks that have to be flushed out," he said.
Intel representatives say that approximately 400 of the most popular 64-bit server and workstation applications have been or will soon be adapted for use with the IA-64 architecture.
But that still leaves the Itanium well shy of the thousands of applications already used on current 64-bit architectures.
Wait for McKinley?
Indeed, many analysts view deployment of the initial Itanium processor as merely a development tool for adapting software to the new architecture, and are instead looking ahead to Intel's release early next year of the second-generation IA-64 chip, code-named McKinley, which is said to offer far greater performance.
"By and large, this is all about proving the architecture so that when the McKinley hits, people will feel good about buying it," Brookwood said.
Members of Intel's panel inadvertently underscored that perception by their silence after an audience member asked why a customer should invest in a first-generation Itanium system rather than wait for a second-generation McKinley system.
Only Intel's Doug Busch, the chip maker's own IT manager, offered his perspective, contending that the initial version of Itanium would be a worthwhile investment.
But panelists' comments also highlighted factors that may work in Intel's favor in winning over customers.
One panelist said that going with an Intel-based 64-bit chip enabled him to choose from a wider variety of hardware vendors than he was able to with current RISC-based systems.
"We've been a supercomputing facility for about 15 years, and until recently we've been locked into the architectures provided by certain vendors," said Rob Pennington, director of computing and communications at the National Center for Supercomputing Applications at the University of Illinois.
"By moving to the Intel-based system we can move away from having to buy into a particular operating system," he said. "If you look at the Pentium clusters we have now, we've got systems from probably five or six vendors, and they're all operating together just fine."
Other panelists cited Intel's dominant role in providing processors for a majority of corporate PCs and low-end servers.
"Basically, our decision to go with Intel was based on faculty decisions," said Al Stutz, director of high-performance computing at the Ohio Supercomputing Center at Ohio State University. "By watching what the faculty were buying for their laptops and research labs, the clear path for us was to follow the Intel line."
James Pool, executive director of the Center for Advanced Computing Research at the California Institute of Technology, speculated that Intel could eventually dominate the PC and server processor markets.
"One of the reasons that we are interested in this architecture is that it could potentially dominate from the desktop to the terraflop," said Pool.
The Mayo Clinic's Stacy said that while his staff was "quite pleased" with their new Itanium prototype systems, the chip isn't the ideal solution for everybody.
"We're not saying it's a runaway on everything, because it's not," he said. "Some of the tests we give it, like moving 8-bit pixels around, it's not very good at that. But there are other compute-intense things that it just shines on."
One area where Intel admits the Itanium doesn't shine is in handling current 32-bit applications. While the chip maker has sought to make Itanium compatible with applications that currently run on Pentium processors, the performance of the IA-64 will lag behind that of a similarly clocked Pentium chip.
"Generally, your mileage will vary depending on the application," said Michael Pope, director of Intel's enterprise software programs. "But overall it will be slower" than a comparable Pentium or Xeon processor, he said.
