Intel says Tulsa will reduce single, dual-CPU servers in enterprise

Another week, another new CPU.

That’s how it seemed this week when Intel announced its latest top-of-the-line Xeon dual core processor, the 7100 series.

The chipmaker says the family of processors, for four- to eight-way servers, offers up to twice the performance and nearly three times the performance per watt of the Xeon MP series it replaces.

With each CPU including multithreading for allowing several tasks to be processed at once, a four-way server would have 16 threads, according to Doug Cooper, manager of Intel Canada. “We’re talking pretty hefty computing power in a very capable system,” he said.

Servers powered by these processors will likely be used to reduce the number of single- or dual-CPU servers in an organization, he said, a market that will appeal to system integrators and system builders.

The eight processors revealed this week range in price from US$856 to US$1,980. “It’s definitely not a small business play,” said Cooper, “but for medium-sized businesses it will definitely be attractive.”

It was only two weeks ago that AMD made a splash by revealing its next generation dual core Opteron processors, not too long after Intel pulled the wraps off its 5100 series of Xeon dual core CPUs for two-way servers.

Tuesday’s announcement was called the continuation of “an historic ‘summer of servers’ for Intel” by Tom Kilroy, vice-president and general manager of Intel’s digital enterprise group.

That claim was immediately contested by John Fruehe, worldwide marketing manager for AMD‘s servers. “I wouldn’t call this release historic at all,” he said in an interview.

He noted that the 7100 series, code-named Tulsa by Intel, is built on the company’s older NetBurst architecture, while the 5100 series has a different platform. Fruehe said Intel had introduced four processor platforms this year. By comparison, he said, AMD’s newest Opteron line – which has single and mulitprocessor series – share the same architecture, as will the quad core CPUs it will bring out next year.

“When we introduced our new platform we did it all at once, to minimize the impact on customers,” Fruehe said. “Intel seems to be dragging this out through the entire summer.”

In addition, he estimated the 7100-series chips consume more power than AMD’s equivalent, the new 8000-series, 150 watts to 120. The 8000-series CPUs “use significantly less power and have better performance,” said Fruehe. Like the latest Opterons, the 7100-series come in several flavours including ones that run in low-power. Intel says the low power option chips only consume 95 watts.

Cooper replied that performance per watt, not merely power consumption, is the real benchmark for this category of processors. As for using the Netburst architecture, he said Intel wanted to take advantage of the 65-nanometer manufacturing process it has – and, although he didn’t say it, which AMD doesn’t have yet.

Each of Intel’s 7100-series CPUs share up to 16 MB of cache, compared to 1MB of cache in the AMD 8000-series. However, AMD Opteron processors have a built-in memory controller, whereas Intel memory controllers are built into separate motherboard chipsets.

IBM, Fujitsu, Hewlett-Packard and Dell were among the original equipment manufacturers that announced servers with 7100-series CPUs this week.

As often happens when a new server processor is introduced, Intel said a number of benchmark tests have been broken by servers built around the new CPUs. It said a Fujitsu-Siemens Primergy RX6300 S3 server using a Xeon 3.4 Ghz 7140M processor and running Windows Server 2003 Enterprise Edition with 64-bit extensions set a record running the SPECjbb2005 business operations test, while a Dell PowerEdge server broke a database performance test.

Comment: <a href=mailto:info@itbusiness.ca

Share on LinkedIn Share with Google+
More Articles