Nvidia Tesla technology will help AMD to break into the server market leaders

Intel's server market. certainly, even in times when they were not prepared processors for servers, and AMD has been successful K8 Intel architecture is owned by an overwhelming advantage. But times change, energy efficiency sometimes comes to the fore, and the ratio of performance per watt has become a determining factor in the world of large computing. More and more servers can require a strong parallelization of computations, even with low productivity in the stream, but consume little power, so the server market began to penetrate processor with ARM architecture. So close, and before the arrival of NVIDIA Tesla technology in the server segment of the mass, according to the. overclockers. ru.

These expectations have another root: August 10, 2011 responsible for the development direction for the High Performance Computing at NVIDIA. Steve Scott has been appointed. (Steve Scott), Doctor of Science in computer architecture, has long served as director of Cray. Almost all experts agree that one of Steve Scott's mission - to make technology-based Tesla microprocessor product of a mass than it is today.

Generally, the NVIDIA Tesla GPUs may not work as usual purpose processors, but as co-processors allow you to effectively solve their cores working in parallel set of resource-intensive tasks. As a result, the narrow profile of these systems compensate for the greater efficiency and lower cost.

However, today NVIDIA Tesla is just an additional component to the cluster server systems, but their basis. For example, for the U.S. Department of Energy. will be deployed. Titan supercomputer system at the GPU NVIDIA Tesla ( they have 85% CPU load), which is three times more economical, and twice as productive of the most powerful at the moment the Japanese supercomputer K ( 20 petaflops ).

The newly appointed Chief Technology Officer of Tesla lineup noticed that this method of computing involving heterogeneous GPU will create a supercomputer in the next decade with a capacity exceeding ekzaflops (1000 petaflops ). Of course, the successes and benefits of NVIDIA Tesla will help create a truly competitive processor. servers.