Having hit the processing-power limits of traditional single-core chips, Intel Corp. and Advanced Micro Devices Inc. are moving the battle to the multicore processor arena. Although eWEEK Labs welcomes multicore technologys potential to handle many compute-intensive tasks at once, we do not see the technology making huge inroads into enterprises without significant operating system and application support.
Dual-core chips integrate two-processor engines into a single package and are making their debut in the x86 world. On-chip cache can be shared between the cores or provided separately for each (as in both AMDs and Intels designs), reducing design time rather than optimizing hardware performance.
Managing power consumption—which is essentially identical to heat production—of these complex processors will challenge both chip makers and system builders, not to mention server-farm-facility architects. Comparisons based on benchmarks may ultimately prove less significant in the marketplace than measurements made with thermometers.
AMD on Thursday released its dual-core 64-bit Opteron processor. AMD, which is already developing quad-core processors, will broaden its dual-core line to desktops with a dual-core 64-bit Athlon in June.
Intel, which earlier this month released its dual-core Pentium Processor Extreme Edition 840 chip, will introduce a dual-core version of its Itanium server chip—code-named Montecito—later this year. Intel is expected to introduce dual-core Xeon processors next year. Both AMD and Intel are expected to add dual-core technology to their mobile chip lines by next year.
In tests, Sun Microsystems Inc.s Sun Fire V40z server, armed with dual-core 64-bit Opteron processors, provided better scalability within the same server footprint.
Organizations should keep in mind, however, that a two-socket, dual-core server might not be able to outperform a four-socket, single-core server, due to limitations of the dual-core architecture. Single-thread applications will not see the same kind of performance boost from a dual-socket, dual-core server. Threads are primarily run through a single processor, thus negating the efficiencies of having two processors.
Organizations looking to deploy dual-core processors will likely find the greatest advantage on the server and workstation side because there are more multithreaded applications available. And as operating system schedulers get smarter and the use of multithreading in applications increases, that advantage will only increase.
Enterprise users will applaud the boost in performance from dual-core chips, but they will likely be disappointed by the complicated licensing options for the technology. Vendors such as Oracle Corp., for example, will require one license for each core on the processor, doubling the price of its software for some organizations.
Microsoft Corp. officials recently confirmed that the company will keep its per-processor software licensing model when dual-core and multicore processor technology becomes available for Windows. This licensing plan extends to several products in the Windows Server System line, including BizTalk Server and SQL Server.
AMD executives suggested that enterprises ask their vendors to count processors based on the number of physical sockets used. In such a scenario, a dual-core chip would count as a single socket.
We urge enterprises looking to deploy dual-core processors to take software vendors licensing requirements into consideration before purchasing. To avoid overpaying, we also suggest that organizations negotiate software licenses that count a single-chip device as one processor, regardless of the number of cores it carries.
Senior Writer Anne Chen can be reached at [email protected].