NEW YORK—IBM later this quarter will begin shipping a blade server powered by Advanced Micro Devices Inc.s 64-bit Opteron processor.
The LS20 will represent the third platform in IBMs BladeCenter portfolio, and the first Opteron-based system not specifically aimed at the HPC (high-performance computing) space.
David Turek, IBMs vice president of deep computing, made the announcement at an event here Thursday celebrating Opterons second anniversary and kicking off the launch of AMDs dual-core processors.
IBM, of Armonk, N.Y., was the first major OEM to support AMDs Opteron when the chip—the first in the x86 market to run both 32- and 64-bit applications—was rolled out in 2003. The company unveiled the eServer 325, a 1U (1.75-inch) rack-mount system aimed at the HPC market. Since then, IBM has added a workstation to its Opteron-based offerings.
IBM officials this week said they will put AMDs new dual-core Opterons into the new IntelliStation A Pro 6217 next month, and into the eServer 326 after that.
They held back the news about the blade system for the event itself.
“We know the HPC market is a place where things start, not where they finish,” Turek said, talking from the stage to more than 200 industry officials, analysts and reporters. The acceptance of Opteron “has grown beyond the HPC niche.”
Turek then held up an LS20 blade, announcing the new system to a round of applause. “This is a powerful blade,” he said.
IBM already offers BladeCenter systems powered by Intel Corp.s Xeon chips and its own Power5 processors.
Blade systems will be one of the key benefactors of dual-core computing, according to Marty Seyer, corporate vice president and general manager of AMDs Microprocessor Business Unit.
In an interview before Thursdays event, Seyer said dual-core computing “is going to bring blades to the forefront.”
Dual-core computing puts two processing cores onto a single piece of silicon, essentially turning a two-way system into a four-way. Blade systems are dense form-factor systems that slip vertically into a chassis, where they share such components as power supply and I/O with other blades. When they were introduced in 2001, they were little more than stripped-down versions of traditional rack systems designed to save space and for front-end tasks. However, with major players like IBM and Hewlett-Packard Co. rolling out more powerful and sophisticated systems, theyve since become a greater presence in data centers.
HP, of Palo Alto, Calif., announced on Thursday a new blade—the ProLiant BL45p—powered by AMDs dual-core Opteron chips, and that other blades using the processors will follow in the middle of the year.
Sun Microsystems Inc., of Santa Clara, Calif., said Thursday that it will bring the dual-core Opterons to its entire line of AMD-based systems, including blades.
A host of smaller systems makers, including Cray Inc., Egenera Inc., Rackable Systems Inc. and Supermicro Computer Inc., also announced support for the dual-core Opterons.
Dual-core computing has been commonplace in the RISC market for several years, through IBMs Power architecture and Suns SPARC chips. However, AMD, of Sunnyvale, Calif., and its much larger rival, Intel Corp., this year are bringing dual-core capabilities to the x86 space.
Intel, also of Santa Clara, announced last week that has begun shipping a dual-core version of its Pentium Processor Extreme Edition, and next month will roll out the next chips in its dual-core offerings, the Pentium D for PCs. The company will launch dual-core versions of its Itanium and Xeon server chips later this year and in early 2006.
AMD on Thursday launched its dual-core Model 800 Opterons for servers and workstations with four or more processors, and next month will roll out its Model 200 line for systems with one or two chips. It will introduce its dual-core Athlon 64 processors, for desktops and notebooks, in June, under the banner Athlon 64 2X.
Next Page: Server space made for dual-core.
Page 2
AMD officials said the server space was the natural place to launch dual-core, since most operating systems are tuned for multiprocessors and many applications already are multithreaded.
AMD officials said their dual-core design is aimed at fueling rapid adoption. The dual-core chips fit into the same space and same 95-watt power envelope as their single-core counterparts, and putting them into systems takes little more than switching out the single-core chips and updating the BIOS.
“Its a very elegant design,” Seyer said.
A key hurdle will be how software makers decide to license their products in dual-core environments. Currently most vendors license their software on a per-CPU basis. Having two processing cores on a single piece of silicon muddies that model. Microsoft Corp. already has announced that it will license its Windows OS on a per-socket basis, rather than charge for each core. Linux vendors have said the same.
However, some ISVs have yet to say how they will license their software. James Mouton, vice president of platforms for HPs Industry Standard Server Group, said the issue will sort itself out over the next year, and that he expects ISVs—with pressure coming from chip and systems makers and with the top OS vendors already deciding on a per-socket basis—will follow suit.
During the event, AMD CEO Hector Ruiz said the key for his company going forward is operating in an environment of “fair and open and free competition.” AMD officials for years have complained about what they saw from Intel as anti-competitive business practices.
AMD got a boost late last month when Intel accepted the Fair Trade Commission of Japans earlier ruling that it was using its monopoly power to hold down competition. Ruiz pushed on this issue, saying an open market is important to AMD and end users.
“Without it, we would stifle innovation and youd be left with someone holding you hostage, and you dont want that,” he said.