"Getting 40% to 50% performance gains is a very significant advancement over the standard processor frequency increases that have been part of the history of the computer industry," says Colin Lacey, director of platform marketing for industry-standard servers for Hewlett-Packard.
Lacey says performance gains from dual-core enabled servers versus traditional single-core systems will vary depending on the application the system is running. However, HP already has seen significant improvements from systems introduced earlier this year based on dual-core Opteron processors from Advanced Micro Devices Inc. In one specific case, HP was able to demonstrate greater than 80% performance improvement using a dual-core Opteron system to run a SAP application environment.
In the past, performance advancements have been limited to much smaller jumps attributed primarily to small, generational increases in clock speed, or gains from chip-structure enhancements like faster buses. And the large-scale performance boost from the dual-core introduction won't be the last time server manufacturers and enterprise customers can expect such a major increase in capability.
"The next time that we will see such a significant performance boost is when we move to a four-core environment," Lacey says.
With Intel accelerating its road map to bring dual-core Xeon's to market as much as two quarters in advance of its original projection, Tim Golden, director of PowerEdge server marketing for Dell, says he wouldn't be surprised to see four-core processors hit the market late next year, well in advance of original projections of a second-half 2007 availability.
"Two-core, four-core, eight-core, 16-core -- all of these are on the road map," Golden says. "This is how we're going to get the greater computational power of processors going forward. Four cores is the next step, and yeah, that could be happening in 2006."
By placing two or more processing cores within a single chip, processor manufacturers such as AMD and Intel are able to significantly increase total processing capabilities while either dialing back clock speed to decrease the amount of heat generated by the chip, or even more greatly increasing performance by boosting clock speed and associated generated heat in small increments.
Chip and equipment manufacturers expect to see rapid adoption of the systems by enterprises, citing the ability to gain a major performance boost at little additional investment. AMD and Intel have projected that more than 90% and 85%, respectively, of their processors shipped will be dual-core by the end of next year.
The only gating factor on adoption could be how software providers decide to charge for use of their applications on multicore systems. Microsoft previously has said it will treat dual-core processors as single processors for software licensing, while virtualization software company VMware Inc. this week confirmed that it will price its products on a "per-socket" basis rather than charging by the individual core.
Oracle has announced a more-complicated formula in which it will charge for each core in a multicore processor implementation as if it were 75% of a CPU. In a dual-core system, two cores would be charged at licensing rate of 1.5 processors.
If Microsoft and others decide to adopt similar licensing schemes as processors move to four or more cores, then the promised performance gain of multicore processing would be offset by the increased cost for software licenses for applications running on the servers.