In recent years, the processor clock-speed race has pushed the on-die thermal envelope.
While there is still room to increase clock speed to boost processor performance, the heat generated in higher-clocked processors will eventually be too difficult to contain effectively.
Putting two processor cores in the same die is a more efficient way to increase performance.
But while dual-core processors can definitely bump performance up by a significant margin, a dual-core processor will not be twice as fast as its single-core counterpart.
This is because dual-core processors must share the same memory controller and bus.
In addition, not all software is SMP-aware or designed to utilize multithreading (where two or more processors run a task simultaneously).
Software developers will have to re-architect their code base to take full advantage of the additional processing power of dual-core chips.
New software will have to be aware of the extra physical (or, in the case of Intels HyperThreading processors, virtual) processors available to run extra process threads.
Nevertheless, dual-core processor technologies have generated much buzz in the industry recently, with AMD leading the pack and releasing its first dual-core Opteron processor earlier this year.
Intel isnt far behind, with plans to release dual-core products later this year or by early 2006.
Dual-core processors are especially appealing to the SMP server market.
Major server vendors such as IBM, HP and Sun have already embraced the dual-core Opteron in their SMP systems and are noting impressive performance.
(Dell currently does not have plans to offer AMD processors in its system lineup, so customers will have to wait for Intels release before they can upgrade to dual-core technology.)
This month, IBM announced the availability of server blades outfitted with AMDs 2GHz dual-core Opterons.
IBM has also pledged support for dual-core processors in high-end servers such as the xSeries 460 (to be reviewed in a coming eWEEK issue), which are “dual-core ready,” according to the company.
This means these servers effectively enable scaling of the x86 system to new heights—as many as 64 processors within an eight-node cluster will be possible in the not-too-distant future.
Intels push for dual-core Xeons is driven in part by rival AMDs release of dual-core Opteron processors for server systems.
In fact, Intel has canceled its next-generation single-core Xeon processors in order to push its dual-core chips out sooner.
Intel has announced that its dual-core Xeon MP Processors, code-named “Paxville” and “Dempsey,” will be available in the first quarter of 2006.
AMD is also on track to release dual-core Opteron processors for server systems this year.
AMD is also slightly ahead of Intel in terms of interoperability.
Current motherboards that support 90nm (or 940-pin-socket) Opteron processors can be easily upgraded to support dual-core Opterons, requiring only a BIOS update.
Intels dual-core architecture, on the other hand, will likely require new motherboards and chip sets, which could be a more expensive proposition.
The availability of dual-core processors will likely have a strong effect on the SMP server market, and software vendors will have to find ways to optimize their products. The cost of dual-core processors should not be much higher than single-core processors, but architectural changes in the dual-core variety will require new server hardware and possibly new memory. Software licensing might also become an issue with vendors that charge by the cores.
At the end of the day, dual-core servers will allow IT shops to scale out better, consolidate hardware more effectively and reduce system management costs.
Environments such as high-performance computing and server virtualization will likely reap the greatest benefit from the forthcoming dual-core systems.
Technical Analyst Francis Chu can be reached at [email protected].