Intel officials are changing some of the goals they have for the next generation of their high-end Itanium platform, a move brought on by declining demand in the Unix server market and changing requirements of OEMs using Itanium, which essentially means Hewlett-Packard.
Intel issued a brief, one-paragraph notice on its Website Jan. 31 announcing that the next version of the Itanium—code-named Kittson and which is still two to three years from being released—will be built on the 32-nanometer manufacturing process rather than 22nm, which had been the earlier plan.
At the same time, Kittson will not be socket-compatible with Intel’s x86-based Xeon server chips. Instead, it will be socket-compatible with the existing Itanium 9300 Tukwila and 9500 Poulson processors. The 9500 Series was introduced in November 2012, and HP introduced enhanced Itanium servers at the same time.
“The modular development model, which converges on a common Intel Xeon/Intel Itanium socket and motherboard, will be evaluated for future implementation opportunities,” Intel officials said in the notice.
The future of the controversial Itanium platform has been debated for years—most recently during a heated legal dispute between HP and enterprise software giant Oracle—and the changes to the plans for Kittson will only fuel further debate, though an Intel spokesman told eWEEK that changes to some of Kittson’s characteristics don’t mean changes to the road map; the chip is still expected to hit the market within two to three years.
Poulson is a good chip—it brought significant improvements in such areas as energy efficiency and scalability—and Kittson will be, too, he said, declining to elaborate on any plans for Itanium after Kittson. He also noted that Intel engineers could continue to make improvements to the platform even at 32nm, and that Kittson’s socket compatibility with Poulson will make upgrading easier for end users.
Intel and HP engineers began developing Itanium in the 1990s, with the idea that the processor would become the de facto 64-bit platform. That changed when rival Advanced Micro Devices in 2003 unveiled Opteron, a 64-bit x86 chip. Since then, Intel has built up the capabilities of its Xeon line, pushing Itanium into more of a niche role for particular high-end workloads and competing with the likes of IBM’s Power systems.
The future of Itanium came into the spotlight again in 2011, when Oracle in 2011 announced it would no longer support the platform in its enterprise software, including its databases. Oracle executives said they had learned from Intel engineers that the chip maker was planning to end development of Itanium, drawing a sharp rebuke from both HP and Intel officials who said the chip’s road map stretched out through the decade.
HP, which has based its high-end Business Critical Systems—including its Integrity and NonStop servers—on Itanium, sued Oracle, claiming the move breached its agreement to support technologies shared by their 140,000 or so joint customers. A judge last year agreed with HP and ordered Oracle to continue porting its software to Itanium.
Intel Changes Direction on Next-Generation Itanium Platform
While HP won the court case, the situation took its toll. Sales of its Itanium-based servers fell sharply since Oracle made its announcement, which also highlighted the shrinking ecosystem around Itanium. (Other software makers, including Microsoft and Red Hat, had already pulled support for the platform.) In addition, it was revealed that HP, which buys more than 90 percent of the Itaniums made, over the years had paid Intel hundreds of millions of dollars to keep developing the platform.
Analysts at Clabby Analytics, in an eight-page report in December 2012, outlined a list of 10 reasons HP will continue to see stagnant—at best—sales of its Itanium-based systems, ranging from increased competition and the migration of workloads from HP-UX to Linux on x86 servers to the growing costs of development, growing customer indifference and Intel’s end-of-life plans for Itanium.
The top reason was what the analysts said was a “broken ecosystem.”
“With major vendors pulling support for Itanium, combined with fewer independent software vendors (ISVs) signing on to host their applications on Itanium-based systems, fewer and fewer key industry solutions are being made available on Itanium-based servers,” they wrote, noting not only the defections of Oracle, Microsoft and Red Hat, but also VMware’s decision to never offer its technology on Itanium. “Fewer solutions = fewer sales.”
In addition, they said they “do not see the next-generation business analytics and big data applications being hosted on Intel’s Itanium architecture (but we do see these applications being hosted on Intel’s x86-based Xeon multi-core architectures, Power Systems, and mainframes).”
The Clabby analysts pointed out that the high point for HP’s Itanium systems came in 2007, when the company sold more than $1 billion of these systems. That number is hovering around $400 million now, they said. HP officials in November said that in the company’s fiscal 2012 fourth quarter, revenues for its Business Critical Systems fell 25 percent from the same period in 2011, following a pattern that officials said had been exacerbated by Oracle’s move.
The declining demand in the overall Unix market mentioned by the Intel spokesman also has been ongoing. According to Gartner analysts, third-quarter 2012 revenue in the RISC/Itanium Unix market fell 18.5 percent from the same period the year before. Within that market, HP has been losing share to IBM.
Intel Changes Direction on Next-Generation Itanium Platform
Intel for several years has been working to bring the Itanium and Xeon platforms closer together—what Intel calls the Common Platform strategy—giving Xeon more of the RAS (reliability, availability and serviceability) features of Itanium, and Itanium the economics of Xeon. Currently, Xeon and Itanium share common chipsets, memory and interconnects. With Kittson and the upcoming “Haswell” Xeon chips, the shared capabilities will extend to the silicon, including design (memory, I/O and RAS features) and sockets.
Meanwhile, HP has launched Project Odyssey, designed to enable businesses to run their mission-critical workloads on either Itanium or x86 servers within a single platform. Part of that is a project called “Dragon Hawk,” an Integrity Superdome system that will run Xeon and Itanium blade servers in the same enclosure. However, the Clabby analysts said in their report that they saw little benefit to this program outside enabling HP to continue to protect the investments it has in the Itanium systems.
“We don’t see the customer demand for an HP-based converged, combined Itanium/x86 infrastructure (where is the lengthy list of HP customers demanding this?),” they wrote. “We see HP’s “converged infrastructure” as a way to keep various HP software programs alive. If Integrity servers are eventually withdrawn from the market, then much of HP’s intellectual capital—and associated revenue stream—in operating systems such as HP-UX and NonStop, much of HP’s availability/resiliency software, and all of HP’s virtualization software investments that are tied to Integrity infrastructure/systems software go down with the ship.”
The analysts, pointing to Intel’s efforts to more closely merge the Itanium and Xeon platforms as part of its Modular Development Model push, said they don’t expect Itanium to be killed off, but rather simply disappear.
“The way we see this new approach … is that Itanium essentially fades into a joint Itanium/Xeon design sometime after the Kittson chipset arrives,” they wrote. “At this juncture Itanium can be expected to just ‘fade away.'”