Download the Nvidia CUDA Toolkit 4.1 RC2

It's really nice to see Nvidia coming forward with two nifty releases at once: we're talking about the CUDA Developer Driver 285.86 and the new CUDA Toolkit, now reaching version 4.1.

If you're a graphics developer working with the Nvidia platform, the 4.1 toolkit is most likely to become your new best friend. The present version sports a lot of enhancements and new features designed to improve your overall performance while at the same time letting you become more creative and experiment more.

The CUDA Toolkit 4.1 integrates the following components:

- C/C++ compiler;
- Visual Profiler;
- GPU-accelerated BLAS library;
- GPU-accelerated FFT library;
- GPU-accelerated Sparse Matrix library;
- GPU-accelerated RNG library;
- Additional tools and documentation.

Nvidia claim an 10% speed increase when using the new LLVM-based compiler enclosed in the new CUDA toolkit, and twice as fast sparse matrix vector multiplying when using ELL hybrid format. On the other hand, the NPP library has received more than 1000 new image processing functions and the GPU-Accelerated  libraries seem to deliver better, faster results.

Finally, new, improved SDK code samples have been included in the present release of the toolkit. For more detailed specifications, please read the changelog in the download page of the Nvidia CUDA Toolkit 4.1. A MAC version is also available.