The plans for the new supercomputer have been announced by the Oak Ridge National Laboratory which would use the GPU architecture of the next generation by Nvidia and is codenamed Fermi. The announcements by Oak Ridge and Fermi were made at the conference of the GPU Technology of NVIDIA which was held in San Jose, Calif. This chip named Fermi would integrate the three billion of the transistors, which is almost three times of the number of the transistors in the most powerful graphics chip which is present in the market. In the future time this chip would also find its way into the GeForce product line of Nvidia which is intended for the PCs.

Oak Ridge’s supercomputer

The supercomputer of Oak Ridge would be used for the purposes of research in energy as well as in climate change and is expected that it would 10 times more powerful as compared to the fastest supercomputers of today as according to one joint statement which was made by Oak Ridge and Nvidia. This architecture would make use of both GPUs i.e. the graphics processing units that are manufactured by Nvidia and the CPUs i.e. the central processing units as according to Nvidia. The central processing units are made by Intel and Advanced Micro Devices among many of the others.

GPUs

The high end GPUs today usually have hundreds of the processing cores which allows them to accelerate specific types of the computational tasks in a much more efficient manner and therefore much faster as compared to the CPUs. According to Jen-Hsun Huang, who is the CEO of Nvidia, the new chip called Fermi would have a little greater than twice as many cores as the current high end GPUs by Nvidia and it has jumped from 240 cores to 512 in Fermi.

Jeff Nichols comment

Jeff Nichols, who is the associate lab director of Oak Ridge for the Computing and the Computational Science, has said in one statement that the GPUs named Fermi would allow substantial scientific breakthroughs which would not be possible without the new technology. He further said that with the help of the Nvidia technology, Oak Ridge has proposed to create a platform for computing which would be delivering the exascale computing in a time period of ten years. It has been proposed by Exascale computing that it would go beyond the performance of the petaflop, which is one thousand trillion, to one million trillion of the operations in one second.

It has also been announced by Oak Ridge that it would be creating the Hybrid Multicore Consortium that would be focused on computing with different types of the architectures for the processors. The goals of this consortium would be to work with the developers in order to run the applications on the super computers of the next generation that are built with the CPUs as well as with the GPUs.

AMD – the rival of Nvidia, has recently announced one of its fastest chips for graphics yet which packs 2.15 billion transistors and it also supports the upcoming programming interface DirectX 11 of Microsoft for accelerating the graphics as well as the computing of general purpose which is in Windows 7 operating system.