Intel’s Nehalem EX Server Chip Spells Doom For RISC?

In addition, the Nehalem EX chip—which will hold 2.3 billion transistors—will have up to nine times the memory bandwidth of the current high-end Xeon 7400 chips, and greater memory capacity through support for up to 16 memory slots per socket.

Other features will include Machine Check Recovery, which is found in Itanium and essentially can detect and correct errors within the CPU, memory or interconnect without having to bring the system down. This is particularly important in virtualised environments, where a single physical system can have many virtual machines running atop it, Davis said. Having to shut down a machine to find and fix an error means having to shut down or migrate all the VMs running on it.

The chip supports live migration of VMs through the company’s FlexMigration which supports VMware’s ESX and vSphere offerings.

He said the combination of Nehalem EX and Itanium will give RISC customers — who use such platforms as IBM’s Power/AIX and Sun Microsystems’ SPARC/Solaris — a less expensive and more efficient options. Davis said he also expects that it will enable Intel technology to muscle its way deeper into the HPC (high-performance computing) space.

Intel officials have been encouraged by the rate of adoption of the Xeon 5500 Series, which was introduced March 30 and which will account for more than half of all Xeon dual-socket shipments by August, Davis said.

The preview comes as rival Advanced Micro Devices readies its six-core “Istanbul” Opteron processor for launch in June.

IBM backing

Despite with Intel’s aggressive bid to do away with IBM’s RISC business, the Intel briefing featured IBM’s Alex Yost, vice president and business line executive for the System x and BladeCenter business. IBM’s next generation X-Architecture and upcoming x3850 M2 and x3950 M2 four-socket servers, all of which will be based in large part on Nehalem EX, he said.

Yost said the combined capabilities of the Intel technology and IBM X-Architecture will help enterprises that are strapped by space and power limitations in their data centers and the increasing complexity of managing these environments.

Jeffrey Burt, eWEEK.com, contributed to this article