Intel Drives Xeon Servers Toward Truland

By Michael Singer (Send Email)
Posted Mar 31, 2005


Intel is rounding out its single-core 64-bit Xeon processor family as it directs future product lines toward a dual-core architecture. The company's 64-bit, multiprocessor platform is setting the stage for next year's dual-core models.

The chipmaker announced five new Xeon chips for multi-processor servers under its Truland group of cores and chipsets. "Think Centrino brand for Xeon MP," Carol Barrett, Intel director for enterprise platforms, said.

The latest Xeon processors are the last step on Intel's road map for single-core multi-processor configurations. Intel and rival AMD have published road maps that show both companies switching to an exclusive dual-core architecture for client PCs and enterprise servers in 2006.

The new Truland line is made up of Xeon processors formerly known as Potomac and Cranford and will power Intel's E8500 chipsets, code-named Twin Castle. The components are sold separately, but Intel will end-of-life its current 32-bit Gallatin Xeon chips.

"This is the completion of the 64-bit offerings through the Xeon MP space," Pat Gelsinger, senior vice president at Intel, said during a press briefing here. Gelsinger was one of the original architects of the Pentium Pro, which is celebrating its tenth anniversary. "We have increased sets of features every year and now literally much of the world operates on this platform that was made 10 years ago."

The Truland platform also includes a faster system bus, support for PCI Express and DDR2-400 memory. The chips will also support demand-based switching with enhanced Intel SpeedStep technology.

Other computer makers and software vendors are already expressing support for Intel's new Xeon family. Dell got the jump on its competition with the new Intel processors. Last week the computer maker said it would use the new Intel chips in its latest PowerEdge 6800 and PowerEdge 6850 servers. IBM, HP, Dell, Unisys, and Microsoft are also among the other top-tier vendors expressing full support for Truland.

Intel said the three chips formerly known as Potomac run at speeds between 3.33 GHz with 8MB of a third-level "L3" cache memory reservoir. There is also a less powerful 2.83 GHz version with 4 MB of L3 cache.

There is a huge price gap between the top and middle of the road Potomac chips. The high-end sells for $3,692 in mass quantities. The next fastest processor in this class runs at 3.00 GHz for $1,980 — nearly 70 percent of the high-end model.

With a mere 1 MB of L2 cache, Intel is slashing the price to $963 for its 3.66 GHz chip previously known as Cranford. A 3.16 GHz version in bulk orders goes for $722.

"Believe me, that is a bargain," Gelsinger said. "The benefit is a performance enhancement that financial services and telcos are asking for. This is fairly consistent with current pricing models."

Gelsinger also said that its future dual-core package codenamed Paxville will be socket-compatible with the Truland configurations.

But the chips eat up more power than their AMD Opteron counterparts. Gelsinger said the Truland family runs at about 110 watts per chip. The AMD dual-core product is expected to consume 95 watts, which includes the I/O and memory, as opposed to 110 watts for a single core chip plus another 20 watts power consumption for the I/O and memory.

In a related story, Intel consolidated its Intel Software Network to give developers one place to go for all of their code needs.

The initial rollout includes tools, tips and training for software developers grappling with multi-core architectures, 64-bit extensions, occasionally connected mobile applications, and advanced system controllers.

This article was originally published on internetnews.com.

Page 1 of 1


Comment and Contribute

Your name/nickname

Your email

(Maximum characters: 1200). You have characters left.