Last month of June, NVIDIA has presented the second generation of Tesla, the name indicates that the proposals developed by NVIDIA hardware for processing GPU Computing. This is architecture based on NVIDIA GPU series GT200 specifically targeted at programmers and developers who need to use the GPU to speed up processing time of applications, provided that the code fits good way to the GPU and that Cuda is used to translate the program to be executed in a form that is usable by the GPU.

NVIDIA has chosen to extend the context of Tesla solutions to various OEM partners presenting the concept of Tesla Personal Supercomputer: a PC, typically in tower and then equated to a workstation graphics space, where there are 4 Tesla cards on PCI Express. Such a system provides a total of 960 core parallel with a power that can reach a theoretical peak of 4 Teraflops.

The cards used are those known as Tesla C1060, PCI Express solutions with cooling system at 2 Slot equipped with clock frequency of stream processors of 1333 MHz, combined with a 4 GB of local video memory clock of 1,600 MHz effective with GDDR3 modules.

Other than as implemented by NVIDIA Tesla system in its S1070, a 1 rack unit in which they were inserted cards 4 Tesla, the Tesla Personal Supercomputer is in fact a complete PC itself, with which the developer can then work so ordinary. NVIDIA recommends a system of this type at least a quad-core processor and 16 GB of system memory, and it is expected that the various partners involved in this initiative, among which we remind Dell, Asus, Lenovo, Velocity Micro, Concordia Graphics, and Scan Boxx, will adopt a configuration of this type.

Such an approach takes advantage of the availability of motherboards with 4 PCI Express 16x, able to mount up to 4 video cards with cooling system with two slots as are the proposals of the Tesla family. The model developed by NVIDIA is actually the son of some systems developed in some areas of research independently: the need to use GPU to speed parallel computing areas has led some researchers to mount more video cards equipped with NVIDIA GPU programmable within the traditional desktop PC, building a genuine personal workstation dedicated to processing GPU Computing.