Code-named “Fermi” next-generation CUDA architecture is the history of the world’s most advanced GPU computing architecture. Fermi has more than 30 million transistors, with a maximum of 512 CUDA core, enabling supercomputing performance, features and terror. With only the CPU-based compared to traditional servers, while its cost is only one-tenth, and the power consumption is only one-twentieth.
UPDATE 3 Nov 09: We just learned that whilst the information is seemingly correct, the pictures below aren’t…
Our apologies for this. I think the information can be tracked to here.
NVIDIA Fermi (GT300) is currently the official release of the next generation of the revolution by the structure of Fermi: real-time rendering of the ray tracing demo of the game screen. Fermi’s CUDA will be used to achieve high-performance general-purpose computing, such as the physical effects, ray tracing, and more levels of CGI special effects movies and other traditional GPU architecture difficult to achieve the top visual content. This all will be CUDA general-purpose calculations to achieve alone, will work with DirectX graphics API version of the no direct relation. Regardless of the API environment, such as DirectX9, DirectX10/10.1, DirectX11, will be the perfect support for Fermi’s CUDA general-purpose high-performance computing, to achieve a more realistic, shocking the senior visual effects.
Fermi by addressing a wide range of computing applications, so that GPU (graphics processor) and CPU co-processing can be widely applied. In the video game industry, visual packaging industry, movies, animation, architecture and industrial design, high-end medical imaging, military simulation, visualization, scientific computing, NASA space exploration simulation widely used in various industries worldwide. Fermi specifically designed for C + + and comes with VisualStudio development environment, it allows parallel programming easier and can accelerate the performance of a wide variety of applications, the application scope of its coverage than ever before a structure is more wide. Fermi in ray tracing, physical effects, finite element analysis, high-precision scientific computing, sparse linear algebra, sorting and search algorithms in areas such as dramatically improve performance.
Source: http://bbs.pczilla.net
