Nvidia has unveiled next generation computational graphics processor architecture codenamed Fermi. Jen-Hsun Huang, Nvidia CEO, stated that Fermi GPU architecture represents the foundation for the world's first computational graphics processors in Nvidia GPUs family - GeForce, Quadro, Tesla. Fermi architecture incorporates three billion transistors and 512 CUDA (Compute Unified Device Architecture) cores to deliver super computing features.
Nvidia's next generation 40nm process Fermi architecture also carries codenames of GT300 aka GF100 that succeeds GT200 GPU architecture. GT stands for GeForce Tesla while GF is presumed to be GeForce Fermi. With Fermi architecture, Nvidia confirms its interest in High Performance Computing, GPU Stream Computing and Super Computing business.
Huang said, "The Fermi architecture, the integrated tools, libraries and engines are the direct results of the insights we have gained from working with thousands of CUDA developers around the world. We will look back in the coming years and see that Fermi started the new GPU industry."
Fermi architecture based GT300 GPU has three billion transistors and has 512 shader processing cores which are organized as in 32-core configuration in 16 shader processors. For memory, Nvidia GT300 GPU will have six channels of 64-bit GDDR5 memory controllers with totally 384 bit memory bus width and supporting up to 6GB of video memory. Entire GPU has 768KB L2 cache and acts as common bridge between 16 stream multiprocessors.
Nvidia remains quiet about the power requirement of this Stream processing monster and probably working out to make it less power hungry. Being manufactured from 40nm manufacturing process, this GPU would be cheaper and we can expect to consume as much energy or little more than GT200 chips.
Touted as world's first computational GPU, Fermi architecture is designed to support current programming tools in Fortran, C, and C++ for optimal performance to be extracted in parallel computing tasks. And with Nexus, Nvidia brings Visual Studio 2008 computing application development environment.
Nvidia also took a big leap ahead by adding memory technology such as ECC memory support
The new set new technologies that arrive in Fermi architecture are:
* C++, complementing existing support for C, Fortran, Java, Python, OpenCL and DirectCompute
* ECC, a critical requirement for datacenters and supercomputing centers deploying GPUs on a large scale
* 512 CUDA Cores featuring the new IEEE 754-2008 floating-point standard, surpassing even the most advanced CPUs
* 8x the peak double precision arithmetic performance over NVIDIA's last generation GPU. Double precision is critical for high-performance computing (HPC) applications such as linear algebra, numerical simulation, and quantum chemistry
* NVIDIA Paralle DataCache - the world's first true cache hierarchy in a GPU that speeds up algorithms such as physics solvers, raytracing, and sparse matrix multiplication where data addresses are not known beforehand
* NVIDIA GigaThread Engine with support for concurrent kernel execution, where different kernels of the same application context can execute on the GPU at the same time (eg: PhysX fluid and rigid body solvers)
* Nexus, the world's first fully integrated heterogeneous computing application development environment within Microsoft Visual Studio
Nvidia didn't share any performance levels of Fermi for leaving its existing graphics cards business unaffected. Now, the question that rises here is about competing with AMD's freshly announced Radeon HD 5800 series DirectX 11 supporting graphics cards. However, Huang emphasized that Nvidia had little interest in replacing the CPU (DirectX 11 based) and is more focused on advancing the computing industry.
Huang said, "No one likes it when competition has a product out in the marketplace. But we have a different vision and it is not only about market share. Of course, we don't want to keep our fans and enthusiasts waiting for the next generation GPU. But we want to take the industry forward and won't change the way we are doing things even by a little iota. We want to make sure the product is ready when it ships in a few months. And it will be - modular, flexible and powerful."
We aren't sure when Fermi architecture based consumer graphics card would come out. Though the company refused to comment on the availability of Fermi graphics cards, it's said that GT300-based Tesla cards would be out in first quarter of 2010. In another note, a Wikipedia entry about GeForce 300 series reveals some GeForce 300 graphics models but there's no authentic information in it. Hence, we count it as speculation.
SOURCE: techtree.comClick to discuss this news item in the forums