Compatible with Windows, Linus and Mac OS operating systems

Mar 20, 2010 09:31 GMT  ·  By

It seems that NVIDIA isn't just concerned with consumers when it comes to its latest graphics solutions. It is more than well-known that the GeForce GTX470 and GTX 480 will make their appearance at PAX 2010, on March 26th, but what has not been mentioned so far are NVIDIA's plans for Fermi on the professional level. Aiming to keep hype high until the actual release occurs, NVIDIA has officially debuted its latest version of the CUDA Toolkit.

Besides just offering new gaming experiences, NVIDIA wants Fermi to become widespread on the professional market, the domain of designers, 3D model makers and professionals that use any other sort of graphically-heavy application. In order to make sure that its CUDA Toolkit is up to speed with recent technologies, the Santa Clara GPU maker has updated it to version 3.0, which brings a number of benefits for GPGPU developers.

One thing the new CUDA Toolkit will do is deliver improved productivity with class and template inheritance through its support for C++. Secondly, Direct3D 9, 10 and 11 support, along with OpenGL support for both CUDA and OpenCL will enable GPGPU/Graphics interoperability. Furthermore, the Tesla Compute Cluster (TCC) feature will improve performance and cluster management. The new toolkit will also include better developer tools for Linux, such as the new CUDA memory Checker, which reports misalignments and out-of-bounds errors.

NVIDIA made its CUDA toolkit available in versions compatible for Windows, Linux and Mac OS, all of which will fully support existing and future Fermi cards, whether they are GeForce, Tesla or Quadro. End-users may download either of the three from the company's official website. Of course, to take advantage of the new features, those interested still have to wait for March 26, when the first Fermi-based DirectX 11 cards will make their appearance.