Bulldozer processors in market by end of month

AMD announced today that it has begun revenue shipments of processors based on new architecture Bulldozer. The first deliveries of the 16-wire, Form 8-processor, code-named Interlagos, go to the super-computers, and there will be 2 – and 4-socket, at the end of the month. The chips were branded Opteron 6200-series.

For more information, there was a shortage, the company did not confirm clock speeds, prices, or performing parts of the Opteron 6200. Gateway leaked from a technical point of a maximum speed of 3 GHz (3.5 GHz The addition of a turbo), parts of 8-threaded, fell 2.3GHz/2.8GHz turbo speed 16-threaded components.

AMD did not provide release dates for 4200 shares Opteron processor, code-named Valencia, which should be 6 – and 8-threaded versions. These are currently scheduled to arrive September 26th day. Even less is known when the first desktop-oriented Bulldozer CPUs will come in the midst of speculation that the FX series processors, codenamed the Zambezi, have slipped in the fourth quarter.

Bulldozer is first company microarchitecture significantly since the introduction of new 2003 core K8 (which generated the core K10 published in 2007). Bulldozer uses a modular approach to processor design, and compare the modules for coarse grains is tricky. Each module can run two threads simultaneously, and each module has two integer pipelines and level 1 cache, one for each thread. However, the floating point pipeline is divided between the two threads, as is the instruction decoder front-end and level 2 cache. Each module has two nuclei in some places, but one heart to another.

Whole, heavy workload, the module will be substantially equivalent to two large-scale of the umbilical cord. Compared to Intel’s hyperthreading, which manages two threads in one core, Caterpillar will offer a lot more resources dedicated to the creation of the two wires. However, the story may have a different floating-point workloads, where the chains have to compete with the resources to implement. Even if the floating point unit is in a better position in the Q10, more resources and support the implementation of new instructions, it may not be as effective as two separate floating-point core.

This design allows AMD to keep each module is relatively small, making it possible to install more modules on a chip. It is also designed to be energy efficient, resource sharing rather than duplicating allows AMD to do more with fewer transistors.

AMD has been shaken since Intel introduced the Core 2 architecture in 2006. Core 2 eclipsed the performance of K8, K10 and K8 although improved, remains behind Intel’s offerings. Intel Nehalem and Sandy Bridge architecture has widened the hole. Bulldozer is a significant move for AMD. The design is not even try to match Intel’s single-thread performance. This is a design suitable for highly threaded server workloads over-heavy: a more lucrative market than the desktop market, but where Intel currently outperforms AMD by about 19 to 1 in 2006 ratio of about 3 to 1 in favor Intel. If AMD Gamble pays off, it will be able to reverse the trend and increasing its share. If so, it is not difficult to see become a major player in the server market again.