NVIDIA has released Version 3.0 of its CUDA Toolkit. Version 3.0 provides developers with tools for preparing their codes for Fermi-based GPUs due to hit the market in a few weeks time. The CUDA Toolkit v3.0 is freely available at www.nvidia.com/getcuda.
Key features of CUDA Toolkit v3.0 include:
- Support for new GPUs based on Fermi architecture: including ECC, optimized double precision, complete BLAS and LAPACK libraries, cuda-gdb debugger and Visual Profiler
- C++ support delivers improved productivity with class and template inheritance
- GPGPU/Graphics interoperability, critical for the GeForce/Quadro market delivering Direct3D 9,10 and 11 and OpenGL for both CUDA and OpenCL
- Improved developer tools for Linux, including the new CUDA Memory Checker that reports misalignment and out of bounds errors
- Tesla Compute Cluster (TCC) support.
One area that is going to see real benefits from the improvements in CUDA Toolkit v3.0 is the field of quantum chromodynamics (QCD). To date, three Nobel laureates have been recognized for their contributions to this field which helps us understand the nature of our universe. The study of QCD is so important that up to 20% of the compute cycles across several national labs and other facilities in the U.S. Japan and Europe are dedicated to it.
"QCD codes need all the compute cycles they can get and we're really excited about the results we're getting by using GPUs -- we've already reduced the cost of our calculations by a factor of 5," said Richard Brower, U.S. Software Coordinator for USQCD, Boston University Physics and Electrical Engineering Departments. "With support for more complex linear algebra routines and a GPU-accelerated BLAS library, the new CUDA Toolkit 3.0 will help us to further extend our use of GPGPU architectures to advance research on QCD and related strongly coupled quantum field theories."