Archive notice: This article was originally published on November 27, 2012. Links and embedded videos are preserved as part of the historical record.
Version 5 of the CUDA computing architecture is intended to provide fresh impetus to the field of parallel computing. Alongside new tools, the CUDA 5 platform is also intended to offer a wide range of libraries and interfaces for developing technical applications on GPUs.
CUDA 5 is intended to make optimal use of the full performance quality of Nvidia graphics processors, including GPU acceleration based on the new Nvidia Kepler architecture. New functions are said to include dynamic parallelism, intended to bring GPU acceleration to new algorithms. GPU-callable libraries are also said to have been added, along with integration of the Nvidia Nsight Eclipse Edition, which is intended to make CUDA code quick and easy to generate.
The CUDA 5 platform is available as a download for Windows, Linux and Mac OS X here.
To help developers increase their potential in application development with CUDA technology, Nvidia has established a free central online resource: http://docs.nvidia.com/cuda/index.html. Here, alongside the latest information about the CUDA platform and development model, developers can access all documentation and technologies, consisting of tools, sample code, APIs and guidance on adapting to and complying with framework guidelines.
Interested parties can also register for the CUDA programme to obtain early access to software releases and tools. More information is available here.
