by Robert Lemos

Intel, AMD Compete for Data Center Refresh

Apr 08, 2010
Data CenterGreen ITServer Virtualization

The processor companies release their latest server chips, looking to nab a greater share of the data center as companies' budgets recover from the recession.

Processor makers Intel and AMD are back to tit-for-tat competition in the server market, and data centers looking to replace their computing hardware stand to gain.

Last week, AMD released its Opteron 6100 Series processor, its top-of-the-line server CPU for two- and four-processor systems. A day later, Intel capped its latest line of processors, adding the Xeon 7500 chip as its alpha dog. Both manufacturers argue that companies are ready to buy again and cutting data center costs by consolidating servers makes sense as firms worldwide emerge from a depressed economy.

“We are all seeing signs of economic recovery,” says Kennedy Brown, Xeon product line manager for Intel. “The IT managers can’t wait any longer. If they wait longer, it will cost them more—whether in terms of downtime or lack of productivity.”

The centerpiece of Intel’s rollout is a sales pitch: That a single quad-processor server today can replace 15 to 20 such servers bought in 2005. Five years ago, those servers sported processors with a single core, or perhaps two cores at the most. Intel’s Xeon 7500 combines eight cores, yet still consumes approximately the same energy as previous generations.

“Just by running less servers, you are saving a lot of energy,” Brown says. “The energy efficiency comes from the consolidation story.”

AMD’s pitch is similar: Its Opteron 6100 comes in 8- and 12-core configurations. Yet, the company emphasizes its favorable price-performance benchmarks against rival Intel. AMD contends that customers’ math has changed with regards to processors. Rather than looking for raw performance, customers are looking for a combination of performance, energy efficiency and price from which they can derive the most value.

“Last year, people’s IT budgets got slashed,” says Gina Longoria, director of Opteron product management for AMD. “What we saw is that people got smarter. If they were originally buying a $6,000 server for their e-mail workloads, maybe they only needed a $3,000 server.”

AMD has split its processors into two lines, the energy-efficient 4000 series and the higher-performance 6000 series. The Opteron 4000 Series, with its 4- and 6-core processors, will serve the markets for one-processor (1P) and two-processor (2P) systems. The Opteron 6000 Series, with 8 and 12 cores per chip, will drive higher-performance two-processor and four-processor (4P) servers.

The two-processor market is the sweet spot for both Intel and AMD, with nearly three-quarters of sales coming in that segment, says Longoria.

“Our big focus is on serving the 2P market in a different way,” she says. “One focus is on performance and one focus is on power and cost.”

AMD has made a concerted effort to undercut Intel’s chips, pricing its 6100 series processors 30 percent lower than Intel’s workhorse, the Xeon 5600. The company has made a great deal of its strategy of not charging extra for higher-processor configurations, eliminating what it refers to as the “4P tax”—the extra charge for each chip in a quad processor system.

How Data Centers Can Benefit from the New Server Chips

Such trends are good for data center managers. The combination of an improving economy and competition between AMD and Intel for their business could amount to sizable savings, says Nathan Brookwood, principal analyst for Insight64.

“If you are running a whole bunch of four-year-old servers based on a single-core product technology, and you haven’t upgraded, you are missing a big opportunity,” Brookwood says. “Both companies have made dramatic strides, both in terms of performance and performance per watt, in the last couple of years.”

The new processors also offer more than just performance and energy efficiency. Both companies’ CPUs increase memory capabilities, with AMD adding a fourth memory channel and Intel increasing the size of its addressable memory by a factor of four, up to 1 terabyte per processor. Such memory improvements are an advantage for data centers that make significant use of virtualization, Intel’s Brown says.

“For virtualization, the memory requirements are demanding, which is one reason why we drove up our memory capacity,” he says.

Intel also boasts that it added 20 reliability, availability and scalability (RAS) features to the Xeon 7500, including a corrupt data containment mode and corrected machine check interrupt (CMCI) to minimize downtime and data loss.

“The reliability features we have in there—when you combine it with the scalability and perfo