Intel Nehalem-EX to be its highest-performing chip

Santa Clara, Calif.-based chip-maker Intel Corp. on Tuesday unveiled details of its upcoming eight-core 2.3-billion transistor processor, the Nehalem-EX, for the expandable high-end server market that offers a performance boost compared to its Xeon 7400 series, its highest-performing offering today.

Designed for servers with multiple processors, Nehalem-EX will feature up to eight cores inside a single chip supporting 16 threads and 24Mb of cache, and it will offer twice the memory capacity, 2.7 times the number threads, and 1.5 times the cache of the Xeon 7400 platform. “It’s a real beast of a processor in terms of its size, 2.3 billion transistors, certainly one of the most complex pieces of technology ever delivered by humans,” said Boyd Davis, general manager of Intel server platforms group marketing.

Intel wants the boost in performance levels of Nehalem-EX to help move customers away from the costly proprietary RISC (Reduced Instruction Set Computing)-processor based systems. “We expect this to accelerate the conversion off of proprietary architectures … onto a more building block approach like Xeon,” said Davis.

The offering is ideal for server consolidation, virtualized applications, data-demanding enterprise applications and technical computing environments, said Davis.

Built using a 45nm process, Nehalem-EX shares some elements of the Xeon 5500 series platform, codenamed Nehalem-EP and released last March, such as QuickPath Interconnect, Turbo Boost, Hyper-Threading and Integrated Memory Controllers.

One distinction in the platform is buffered memory so capability levels required by high-end workloads can be attained, making it compelling for high-performance computing, said Davis. “Today’s high performance compute clusters typically use dual processor servers … (but) the level of memory bandwidth and floating point performance of this class of platform we think will have a dramatic effect on a lot of the HPC market, and that will be a new area for this family to make some strong gains in,” he said.

Another capability in the platform is Machine Check Archive (MCA) Recovery so errors in the data of a server can be observed and reported. Typically, a system will require a restart upon discovery of an error, but for some systems, that downtime can be costly, said Davis. “It allows for a number of previously uncorrectable errors,” said Davis. “They can be observed, and negotiate with the operating system or the virtualization software to contain, correct or recover from that error without forcing a system restart.”

The hardware is based on Nehalem, a new microarchitecture announced by Intel last year. In addition to its Xeon 5500 server processors, the company has also released its Core i7 desktop processors last year based on Nehalem.

According to Warren Shiau, senior associate with Toronto-based consultancy The Strategic Counsel, while the competitive focus of Xeon 5500 was Sunnyvale, Calif.-based rival chip-maker AMD Inc., the focus of Nehalem-EX is “taking that RISC-installed base and moving it over.”

Shiau expects the conversion to industry standard-based architecture to surely accelerate based on performance levels of Nehalem-EX.

But no matter how you cut it, Shiau said Nehalem-EX is a good product all around with significant performance advantages across different areas compared to AMD offerings. “There are no product weak points really,” he said. “Combine that with pricing levels that make it competitive on a line by line basis, there is nowhere to run. This is going to be a big, big problem for AMD.”

Nehalem-EX will be in production later in 2009 and available in systems early 2010.

Davis also provided an update on Xeon 5500 server chips since the release two months ago, describing customer acceptance as “quite strong.” Intel expects that the Xeon 5500 processor line will represent more than 50 per cent of shipments by August for the dual-processor server market. “It’s off to a pretty compelling start … and is at or ahead of where we expected to be at this point,” said Davis.

Related Download
3 reasons why Hyperconverged is the cost-efficient, simplified infrastructure for the modern data center Sponsor: Lenovo
3 reasons why Hyperconverged is the cost-efficient, simplified infrastructure for the modern data center
Find out how Hyperconverged systems can help you meet the challenges of the modern IT department. Click here to find out more.
Register Now