Intel Corp. says two brains are better than one when it comes to crunching numbers and serving up Web pages.
At an event Monday in San Francisco, the chip maker unveiled its first dual-core Xeon server chip, code-named Paxville DP, months ahead of its original schedule, and officials said a dual-core version of its Xeon processor for servers with four or more chips will follow within 60 days.
Intel has put the bulk of its chip design efforts into engineering multicore processors, which package two or more processor cores inside one chip.
The results allow computer manufacturers to boost the performance of servers with two, four or more processors without large increases in price or power consumption.
The focus on dual-core chips also changes Intels focus from all-out speed to performance per watt, a measure it has said is more appropriate for businesses that maintain large numbers of servers.
The company, which has already been shipping dual-core Pentiums for the desktop, will add dual-core notebook chips in early 2006.
“Todays focus, and the focus of this year, has been on the transition to dual core,” said Kirk Skaugen, general manager of Intels Server Platform Group.
The Paxville DP Xeon chip, with its twin processor cores running at 2.8GHz, is arriving early. Intel had originally promised to deliver the dual-core Xeons, which will come with up to 2MB of cache per core and an 800MHz front-side bus, in early 2006.
However, Intel pulled in the Paxville DP launch, due to what it says is the health of the new product. The move shaved several months off the lead Intel rival Advanced Micro Devices Inc. had with its dual-core Opteron server chip, which came out last April.
In response to a question at the press conference, Skaugen said Intel has leveled the playing field and will aggressively push dual-core technology in the future.
“Were confident that were competitive and Q1 and Q2  look very promising,” Skaugen said.
The first of the dual-core Xeon MP processors, called the Xeon Processor 7000, will arrive within 60 days, and will run at speeds up to 3GHz. Unlike the Paxville DP chip, which runs with the current E7520 chip set, the Xeon Processor 7000 will come with a new chip set, the E8501, and will support the existing E8500 chip set.
That will be refreshed next with the “Truland” platform, which will include the “Tulsa” processor. Later this year, Intel, of Santa Clara, Calif., will release the first of its dual-core Itanium processors, codenamed “Montecito,” and early in the first quarter next year will release a refresh of the dual-core Xeon DP platform, code-named “Bensley.”
That platform will include the “Dempsey” processor and “Blackford” chip set, and also will include other features designed to drive performance, such as Intels virtualization and management technologies.
A low-power processor aimed at systems such as blade servers, code-named “Sossaman,” also is scheduled to come out in the first half of 2006. Sossaman will have a power envelope of 31 watts, with a 15 watt model for telecommunications devices, Skaugen said.
Processors code-named “Merom,” “Conroe” and “Woodcrest,” for notebooks, desktops and servers, respectively, are due in the second half of 2006. The first of Intels four-core processors will appear in 2007, he said.
“Our basic message is, all the products were developing … are dual core or multi-core,” Skaugen said.
By the end of 2006, Intel said, it expects that 85 percent of its Xeons shipped will be dual-core, with that number jumping to almost 100 percent by the end of 2007. For Pentium chips for PCs, those numbers look to be 70 percent by the end of 2006 and 90 percent by the end of 2007. Much of that will be driven by Intels transition to the 65-nanometer manufacturing process starting next year, Skaugen said.
In all, Intel has 15 dual- or multi-core projects underway, he said.
The chip maker is targeting the new dual-core Xeon DP chips at several areas, including traditional back-end applications, such as databases and business intelligence, as well as server consolidation projects, virtualization technologies and greater density in the data center.
The chip also features Demand-Based Switching, a technology that automatically powers down the processor when its not being fully utilized by the application. Intel officials said the new chip offers up to 51 percent performance improvement over the current processor.
Whos Adopting Xeon DP
The Intel Xeon DP chip is expected to be widely adopted by server manufacturers, including Dell Inc., Gateway Inc., Hewlett-Packard Co. and IBM.
The companies will use the new chips to their marketing advantage by touting their extra computing power. Companies like Gateway, whose largest system now contains four single-core processors, will use the chips to extend their reach.
With dual-core chips on board, Gateway can seek out customers looking for boxes that can do the work of eight-processor servers, said Tim Diefenthaler, senior director of product marketing at Gateway.
Businesses are expected to take advantage of the new chips by buying dual-processor servers, for example, which will work like four-processor systems, but sell for prices closer to todays dual-processor, single-core systems.
Dell, which began gathering the orders in late September, said it would offer the dual-core chips on a mix of single and dual-socket PowerEdge servers, including PowerEdge 1850, 2800 and 2850 models, as well as its PowerEdge 1855 blade server and its Precision 470 and 670 workstations. Pricing for the servers starts at $2,448. They began shipping Monday, the company said.
Dell, along with companies like IBM, has already been selling dual-core Pentium D-based machines for small and midsize businesses.
IBM will also be among the first to offer the new Xeons. It will get its start by offering the dual-core chip in two of its X Series server models.
The Armonk, N.Y., computer giants two-socket x346, a 2U (3.5-inch) rack-mount server, will be its first to offer the dual-core Xeons at speeds of up to 2.8GHz. The machine, which can be configured with up to 2GB of RAM, dual-gigabit Ethernet links and a single hot-swap power supply, will start at $2,969—the same price as its single-core predecessor—a measure IBM has said it hopes will speed adoption of the dual-core chips.
IBM will follow up by adding the dual-core Xeon to its two-socket X Series x336 server in November. Pricing and specs on the x336 will come as the shipping date gets closer, the company said.
Still other server makers will join the fray in November.
Gateway of Irvine, Calif., said it plans to ship a pair of servers based on the new dual-core Xeons at that time.
The company will ship the two-socket E-9510T tower server and the E-9415R rack system armed with the dual-core Xeons. Its E-9510T server tower will offer two of the dual-core chips, along with up to six SATA (Serial ATA) drives or up to 10 SCSI drives, as well as redundant and hot-swap fans and power units. Its dual-processor, rack-mount E-9415R will offer up to three SATA or SCSI drives, and also have redundant power supplies, the company said.
HP will put the dual-core Xeons in its ProLiant DL380, a dual-socket system that will ship in November. Later the same month, HP will start shipping the four-socket DL580 and ML570 systems with the dual-core Xeon chips. Both will have up to 64GB of memory support, the company said.
Despite having a number of brand-name server makers signing on to offer its chips, Intel will still face competition from AMD, as many dual-core Opteron servers—some of which come from Sun Microsystems Inc. and others which are made by IBM and HP, which will also offer the dual-core Xeons—have already been on the market for months.
For its part, HP said the demands of the software its customers are running may well dictate the choice between Intel and AMD processors.
HP has been pleased with the customer adoption of its Opteron-based ProLiants, particularly among those running massive back-end applications and working in high-performance computing.
But the company said it expects to see a mix of demand for both platforms, as those customers who are concerned about processor cache sizes will gravitate to the Intel systems, while those more focused on memory latency will more likely adopt the AMD processor-based systems.