The quad-core Opteron is big news for AMD. Not only is it the world’s first natively x86 quad-core processor with tri-level memory shared equally with all four cores, it represents a huge invitation to the corporate server feeding grounds.

Servers are big business and represent serious profits for microprocessor companies.

According to AMD information, the company began engineering the quad-core Opteron during the July-September quarter of 2004 in Austin, Texas with help coming from AMD’s India engineering works as well as its center in California.

Around 400 engineers and 1.5million man-hours went into the quad-core design with production efforts coming from around the world including fabrication in Dresden, Germany, testing in Singapore and assembly in Malaysia.

In total, AMD has released nine new quad-core Opterons plus details on how to tell the chips apart.

The new models are: 8350, 8347, 8347-HE, 8346-HE, 2350, 2347, 2347-HE, 2346-HE and 2344-HE.

The first digit represents the scalability of the processor design so 8xxx series processors can be scaled up to eight-way designs, 2xxx series up to two-way designs.

The second digit is the generation, these all being third-generation Opteron CPUs, all quad-core, second-generation Opterons (x2xx) CPUs were dual-core.

The last two digits give a degree of performance information so that an 2347 CPU would perform faster than a 2344-HE but slower than a 2350.

Importantly, the new quad-core Opterons remain Socket F (1207) compatible so with careful BIOS updates, current Opteron boards should support these new processors.

AMD is pricing the 2xxx series very competitively with the 2344-HE at $US209, the 2346-HE $US255, 2347-HE, $US377, the standard 2347 at US$316 and the 2350 at $US389.

AMD is promising performance gains of 50% plus a host of new features including CoolCore technology, designed to turn off parts of the processor logic when they’re not being used to save power and reduce heat.

The chips also incorporate AMD’s wide floating point accelerator that adds in 128-bit SSE floating point calculations, offering up to four calculations per clock cycle.

The company has also optimised the memory controller for greater throughput of data up to 50%, improving multi-threaded application speed.

Specifications of the chips follow: