Monday began shipping the second-generation of its Xeon Processor MP platform, code-named ‘Gallatin.’
The Xeon MP is Intel’s 32-bit processor for the 4-way and up processor market. The previous version, known as Foster, was released in March.
Santa Clara, Calif.-based Intel claims a 19 percent to 38 percent (depending on the application being run) performance improvement for the new chip. It also claims a performance advantage of up to 85 percent over similar dual processor configurations.
The main difference between the two versions lays in Gallatin’s souped-up cache. The Gallatin processor contains 2 GHz of 2 MB Level-3 cache, a 100 percent increase over Foster’s 1.6 GHz of 1 MB Level-1 cache, which Intel said it will phase out over time.
Intel said Gallatin is best suited for typical server workloads such as databases, customer relationship, and supply-chain management. Hardware OEMs like IBM
; operating system OEMs like Microsoft
and various Linux makers; and application server OEMs like BEA
and SAP AG
are expected to announce their support for the new chip either today or in the next few days.
“We work with our OEMs months in advance to give them tuners and compilers to prepare them for that eco-system,” Intel Enterprise Marketing Manager Shannon Poulin told internetnews.com. “That way, they can take the old processors out of their boxes and put the new ones in… and the beat goes on.”
But while faster chips are considered a good thing for server makers, analysts confer that the new chip is hardly earth shattering.
Gallatin is manufactured on 0.13 micron processor; Foster used 0.18 micron processor. The size difference allows for the larger cache, but cuts into frequency, something that Illuminata analyst Gordon Haff believes is of little impact for a high-end server chip like the Xeon MP.
“The name of the game in the high-end server space is not frequency but rather overall system design,” Haff told ServerWatch.
He adds that Gallatin’s greatest strength is that it corrects Foster’s greatest flaw, which is its too-small cache.
In Stat/MDR analyst Peter Glaskowski agreed. “This is strictly a steady evolution of its Xeon processor,” he said. “Gallatin merely gives you more performance for the same price as a Foster. I’m not aware of any new applications that would take advantage of the faster speeds.
Glaskowski did say that the closest competition to the new Xeon comes from IBM’s POWER4 or AMD’s high-end Athlon chips. But, Glaskowski says Intel should remain at or near the top of the heap for some time.
And as for Intel’s speed comparisons with Sun Microsystems
, Glaskowski said Ultra SPARC can’t compete with the same level for integer and floating point use.
“Ultra SPAC processors have not been competitive for about 10 years, but that hasn’t killed Sun. They build their systems on the performance and reliability, not chip speeds,” said Glaskowski.
Like Foster, Gallatin is optimized for the Intel NetBurst microarchitecture and Hyper-Threading Technology.
The Gallatin chip also offers investment protection, as enterprises can now replace the Foster chips with Gallatin chips in a configuration. Future Xeon designs will continue to allow for such upgrades (what Intel refers to as “growing room”). The two versions cannot be mixed on the same board, however.
John Enck vice president and research director, Server and Directory Strategies at Gartner Group noted, “The release of ‘Gallatin’ is a good indicator of the strength of the IA-32 server market. The OEM designs for 16-way, 32-way, and beyond show how IA-32 computing continues to push upward in the market, extending IA computing in the midrange server market.”
Enck told ServerWatch that he believes the only downside of Intel’s further entrenchment in the 32-bit space is potential overlap with its still-emerging Itanium Processor Family (IPF) market. “Intel will need to be very clear on where it sees the line between high-end IA-32 computing and IPF computing. The danger of this overlap is not to IA-32 computing, which remains incredible strong, rather the danger applies to IPF computing because the market for IPF is still in the early stages of evolution.”
Haff sees this strategy as less cannibalistic, however, believing that the IA-32 space has become more important than Intel had anticipated when it launched Itanium. He believes that this space has become critical to Intel retaining its hold on the high-end server market.
Haff believes that while Intel dominates the 32-bit space, its 64-bit Itanium offerings don’t have the same draw. Therefore, by powering up its 32-bit chips, the vendor is able to keep customers who might otherwise migrate to non-Intel 64-bit chips within the fold.
Editor’s note: Amy Newman is Managing editor for Server Watch. Michael Singer is an Associate Editor for internetnews.com