In-Depth
Itanium Dealt Setbacks by Microsoft, HP
Intel is abandoning the low-end 64-bit space to compete against RISC chips from IBM and other vendors in the high-end
Intel Corp.’s Itanium aspirations were dealt another blow last week when Microsoft Corp. confirmed that its forthcoming Windows Server 2003 Compute Cluster Edition operating system would not be available on Itanium 2.
Microsoft’s announcement, coupled with Hewlett Packard Co.’s (HP) recent decision to backtrack from the Itanium 2 workstation market, are two of the most recent setbacks for the world’s largest chipmaker.
It wasn’t supposed to turn out this way. When Intel first trumpeted its Itanium plans a decade ago, the chip giant outlined a vision of a mainstream 64-bit chip that would be comparable in sales volumes and ISV acceptance to its seminal Pentium microprocessor. “Certainly the original hopes that I think Intel had are pretty much in the dust at this point, because originally Itanium was posited as an industry standard processor that was basically going to take over the RISC market,” says Charles King, a principal analyst with consultancy Pund-IT Research. “I don’t see that happening at this point.”
Not that Intel is going down without a fight. Last week, the chip giant announced new Itanium 2 processors designed for the high-performance computing (HPC) and large-scale database markets, among others. The new Itanium 2 chips, six in all, support faster bus speeds, boast beefier Level 3 (L3) caches, and include a low-power Itanium 2 processor designed for blade servers. The centerpiece of Intel’s announcement was a new Itanium 2 MP CPU that runs at 1.6 GHz and is equipped 9 MB of integrated L3 cache. This chip is designed specifically for one market in which Itanium has had some notable success—namely, the large database market on large (16 processor and up) SQL Server 2000 systems.
Yet, in many of the markets in which Itanium was slated for superstardom, the 64-bit RISC-slayer simply hasn’t panned out.
Ironically, Itanium has been stymied by the dazzling success of the x86 architecture, which has spawned two 64-bit offshoots—the AMD64 architecture from Advanced Micro Devices Inc., and Intel’s so-called Extended Memory 64-bit Technology (EM64T), a series of proposed 64-bit extensions for the Pentium and Xeon microprocessors. As Microsoft’s announcement last week made clear, there simply isn’t much demand for Itanium 2 in the low- and midrange HPC space the software giant plans to target—not with AMD’s Opteron and Intel’s EM64T-enabled Xeons available at a lower price point.
“It certainly indicates that Microsoft didn’t expect to see a lot of volume on that product on Itanium,” confirms Gordon Haff, a senior analyst with consultancy Illuminata. “Microsoft isn’t doing anything in the high-performance space right now, and with this Cluster Edition it obviously hopes to start doing better. But assuming that they’re halfway realistic, they’re still not going to be expecting huge volumes in the near term.”
Itanium has had some limited success in the HPC space, Haff says, but as a high-end alternative to RISC microprocessors.
“Itanium is actually doing pretty well in HPC, if you look at things like the Top 500 Supercomputer list. That said, Itanium, like Power, and like other types of high-performance sort of RISC-like or RISC-replacement chips—from a volume perspective, these are a lot lower volume than the x86 systems.”
Last month, HP backed away from plans to deliver Itanium 2-based workstations. Like the HPC and large-scale database markets, the technical workstation space was pegged as a likely bread-and-butter market segment for Itanium. As Itanium cedes market after market to the volume 64-bit plays, some industry watchers believe it’s destined to be a niche-oriented offering.
“The sort of modest numbers for Itanium in HPC and other spaces isn’t really a commentary that Itanium isn’t doing well. It’s just a commentary that Itanium is more of a high-end, high-function processor as opposed to the large-volume little two-way systems,” Haff argues.
Nevertheless, analysts say, Intel has been forced to conclusively reassess the way in which it positions Itanium.
“I think even Intel is positioning it as a high-end alternative to RISC. Six months ago, they were saying, ‘For $25,000 and above, it’s going to be Itanium; for below $25,000, Xeon,’” says Nathan Brookwood, a principal with microprocessor consultancy Insight64. “Now they’re saying, ‘If you’re using a RISC system, then you’ll probably want to migrate to Itanium. If you’re using an x86 system, then you’ll probably want to migrate to an x86 system with 64-bits.”
It’s taken some prodding, but even Intel officials now acknowledge that Itanium is a high-end-only play. “If you look at a lot of the RISC vendors today—IBM, HP, even Sun—all of those major high-end RISC OEMs have continued to maintain RISC at the high-end and Xeon for the volume market,” said Jason Waxman, director of multiprocessor marketing for Intel’s Enterprise Product Group, in an interview this summer. “There’s a reason for this. There are two very distinct sets of requirements. It would be nice—it would have been nice—if Xeon could cover both the volume segments and go all the way up, but, fundamentally, there are two different requirements that will continue to coexist.”
Waxman also conceded that Itanium unit shipments aren’t likely to ramp up any time soon—if ever. “By the middle of next year, we expect that more of our enterprise processors that we ship will be 64-bit rather than 32-bit,” he said. “But what we don’t necessarily see is that all of the 64-bit software or infrastructure will be there, so we don’t see that [volume] shifting to [the 64-bit only] Itanium.”
But no one is ready to write Itanium off just yet, with good reason. It’s still a viable alternative to RISC in the high-end. “The RISC systems today comprise only about five percent of the units in the server market, but they still comprise about 50 percent of the revenues, and therefore, Intel still believes they can get a slice out of that 50 percent of server revenues,” Brookwood observes.
Intel has invested more than a decade of research and untold billions of dollars in Itanium’s development, but it isn’t alone. Computing and services giant HP collaborated with Intel on Itanium’s development and has an enormous stake in the 64-bit chip’s future; HP plans to discontinue development of its own PA-RISC CMOS and transition users of its HP-UX operating environment over to Itanium-powered servers. So does Itanium’s volume lethargy spell bad news for HP?
Not everyone thinks so. One of the biggest impediments to Itanium uptake concerns its radically different Explicitly Parallel Instruction Computing (EPIC) technology, which has a prodigious learning curve. Developers are still warming to Itanium development tools, and few—if any—are as comfortable programming for EPIC as they are for x86. Yet these are issues for which HP has had years to prepare. In fact, by all accounts, the computing and services giant ships robust and functional Itanium-ready development tools for HP-UX. The result, says Haff, is that HP has had a lot of success encouraging Itanium uptake among its own PA-RISC and Alpha (formerly Digital) customers.
“If you look at HP and HP-UX specifically, that’s going along quite well. These things never turn over night, but the kind of numbers I’ve seen of PA-RISC customers and Alpha customers adopting Itanium seem to be about where one would reasonably expect,” he says.
More to the point, says Insight64’s Brookwood, Itanium gives HP a migration path for not just HP-UX, but also Tru64 Unix and OpenVMS, both of which it acquired from the former Compaq Computer Corp. (by way of Digital), along with the NonStop Himalaya operating environment from the former Tandem (itself a Compaq acquisition.) “HP will be able to basically turn off the development of PA-RISC and turn off the development of Alpha, and focus all of its efforts on Itanium-based systems,” he points out. “That in itself will be an achievement, because if you look at other computer companies who have merged in the past, they have never been able to get rid of those proprietary product lines.”
Pund-IT Research’s King, however, thinks that Itanium has been a disappointment for HP, especially when evaluated in light of the strategic vision—i.e., an industry-standard 64-bit architecture that would achieve high volumes at relatively inexpensive price point—its original premise.
“This has got to be a disappointment, if for no other reason than for the amount of money that was invested by both HP and Intel and the potential return on that investment over time,” he says. “If what you started with is the idea to take on the market and what you’re ending up with is a processor that’s basically going to be a 64-bit platform for a single vendor, the payback is going to take an enormous amount of time. As far as money paid out and money paid back in, the more niche-oriented Itanium becomes at this point, the further and further out the point where that investment and R&D and marketing are ever going to be paid back.”