AMD will introduce in volume the industry’s first native quad- core processors, meaning that four processing cores will reside on a single silicone die. This contrasts with the competition’s current dual/dual core approach, where two dual- core chips are connected in a quad-core multi- hip module. The Quad-Core AMD Opteron” processors, which will be available for IP to 8P servers and workstations, will also feature micro-architectural enhancements. Indeed. AMD’s new quad-core technology has led experts to suggest that AMD’s approach to multi-core processing is more efficient than that of the competition.

“Intel has pursued a strategy of optimizing time-to-market, whereas AMD has pursued a strategy of optimizing performance,” says Nathan Brookwood, founder of Insight64, a semiconductor and IT analyst firm in Saratoga.
Quad-Core AMD Opteron processors will deliver the same 95-watt electrical and thermal envelope (with low power options) as its current single- and dual core processors, AMD expects that quad-core will advance the performance-per-watt characteristics of AMD Opteron processor-powered systems. Once available, data center managers can benefit by being able to deploy quad-core servers using the same space and infrastructure as existing servers powered by AMD Opteron processors. Current DDR2-based AMD Opteron processor-based platforms will be able to seamlessly upgrade to native Quad-Core AMD processors without compromising performance, power or heat.

POWERING DESKTOPS
AMD also plans to introduce quad-core processors for desktop systems, though the demand will likely be greatest for quad- core servers. The benefit of upgrading to quad-core is that you can ideally fit more processing power in a small space. These days, many IT managers are operating data centers where they are maxed out for physical space or incoming power and cooling capabilities, for them, the name of the game is consolidating workloads, and replacing all of their less powerful machines with newer ones that can do more computations and data processing without using more power or gene rating more heat.

MAXIMUM PERFORMANCE-PER-WATT
AMD expects that the native Quad-Core AMD Opteron processor will offer significant efficiencies to its customers. AMD Opteron processors reduce bottlenecks found in legacy architectures through the innovative Direct Connect Architecture, which includes an integrated memory controller per processor and fast Hyper Transport™ technology* links connecting processors and I/O. With AMD’s native quad-core, each core has a 64KB L1 data cache, which supports two data loads per clock cycle, and an independent 512KB L2 cache. When data must be shared between cores, it is transferred within the processor. Quad-Core AMD Opteron processors also add a shared 2MB L3 cache across all four cores, which is especially useful in database applications. “Having an intelligent native quad-core, as opposed to a dual dual-core, translates to efficient power usage and outstanding performance,” says Brookwood. “Because AMD has four cores on one chip, communication between any of the cores happens on the chip itself, whereas competitive quad-core design requires each dual-core chip to communicate with the other by going out over the front-side bus.” Also conserving power is AMD’s use of power-efficient DDR2 memory. And AMD PowerNow!™ technology with Optimized Power Management can reduce power based on workload, for savings of as much as 75 percent at CPU idle.
On the performance side, Quad-Core AMD Opteron processors will feature IPC enhancements which impact all workloads and some focused enhancement upgrades to the floating-point execution units, which is especially valuable in computer intensive applications .And there are enhancements to AMD Virtualization™ technology which will improve the performance in virtualized environments.

This entry was posted on Monday, May 21st, 2007 at 2:45 pm and is filed under Computers. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.