Nvidia to release completed CUDA SDK in June
After releasing the beta version of its Complete Unified Device Architecture (CUDA) software development kit back in February, Nvidia is now ready to introduce a finished version of the kit. As ExtremeTech reports, Nvidia plans to release the 1.0 version of the CUDA SDK some time next month. For those not in the know, CUDA is an effort to allow programmers to tap into the floating-point computing resources of Nvidia's GeForce 8800 graphics processors via a high-level, C-like application programming interface. Our counterparts over at Beyond3D have more in-depth coverage of how CUDA works, but essentially, it's a way for programmers to write general-purpose computing applications that run on Nvidia's cards.
In preparation for the 1.0 SDK release, ExtremeTech says Nvidia has showcased applications from early adopters of its CUDA tools. Those early adopters include the Massachusetts General Hospital, which reportedly replaced a 35-node workstation with G80-based hardware for digital tomosynthesis ("a continuous low-power X-ray that can be combined to form a real-time image"). CUDA has also been used by Acceleware for electromagnetic field simulation and by the University of Illinois' Theoretical and Computational Biophysics Group for molecular dynamics computations.