14 DAY TRIAL // Almost immediately after the official launch of the quad core server intended central processing units from Advanced Micro Devices, a number of rumors started circulating about their performance level relative to similar solutions from Intel. While the general consensus was that AMD's products are offering better performance when reported at the energy consumption level or at the pricing policy, hard numbers were sadly lacking.
Until now, that is, when Scali announced the release of the first benchmarking data results using the newly released AMD Opteron 2350 processor based on the quad core internal architecture. The Opteron processor also used the Mellanox ConnectX interconnect system and as a whole the benchmark results showed the increasing importance of the high performance MPI and InfiniBand interconnects when the number of processing cores is rising.
The two companies, Scali and Mellanox, compared the performance of several hardware platforms when running the same application with the LS-Dyna neon_refined_revised TopCrunch benchmark. The results were in favor of the Scali MPI Connect based-version version of the benchmark as it returned a huge 30 percent difference when compared to similar results posted by TopCrunch. Paul Gilmore, the acting director of the Global Software Alliances group from AMD, cited by the news site money.cnn, said that the Scali benchmarking solution is the one able to ''unleash the true power of our Quad-Core AMD Opteron processor'', and that the superior technical design and specifications of the AMD native quad core approach is making itself felt.
As the AMD processors are integrating a number of incompatible technologies from the Intel family of products, while maintaining general compatibility with the x86 standard, it is often difficult to make a direct comparison between the products of the two companies. As Paul Gilmore also said, ''AMD's native core design, integrated memory controller, and Balanced Smart Cache provide the required low latency and high bandwidth needed to feed all four cores on the processor. Scali's MPI Connect techniques for communication between cores connected by shared memory, shows excellent results running the most demanding applications''.