Both my kids have developed a passion for science, each in their own unique ways. One of my observations about today’s generation of scientists is that they are both much more interdisciplinary as well as more aware of the computational implications of their work. Much scientific work is already being accelerated today with GPU computing solutions based on HP ProLiant servers. For instance, the Nvidia GPU Starter Kit from HP delivers 10,752 GPU compute cores in less than half a rack of compute space at under $10/core! Factory configured with HP’s Linux HPC Value Pack, HP Cluster Management Utility, and the latest CUDA 4.0 software, there is no quicker way to get started in GPU computing.
CUDA 4.0 makes it easier than ever for developers to take advantage of the disruptive performance/$/watt of GPUs and new GPU accelerated commercial applications are announced every day, such as today’s announcement to develop a future GPU-accelerated release of Gaussian, one of the world’s leading software application for quantum chemistry.
The GPU support in Gaussian will be enabled, in part, through use of The Portland Group’s PGI CUDA Fortran Compiler. With CUDA Fortran, it is almost trivial to modify existing Fortran code to define subroutines that execute in parallel on a GPU.