NVIDIA Updates CUDA Toolkit Version
NVIDIA Updates CUDA Toolkit Version
NVIDIA has rolled out the latest version of its CUDA Toolkit and software development kit (SDK). The new releases follow closely in the steps of NVIDIA’s partnership with Dell and its announcement of a new preconfigured cluster.
CUDA is NVIDIA’s parallel computing engine for its GPUs. The new 2.2 version of the toolkit and SDK come with improved performance and a hardware debugger in addition to support for the Windows 7 operating system. Also included in this version is the CUDA Visual Profiler for profiling C applications running on the GPU.
CUDA 2.2 also offers better performance for OpenGL applications such as medical imaging. By enabling CUDA functions to read and write directly from pinned system memory, the new version reduces the volume of data exchanged between the GPU and CPU. By doing so, it boosts performance for streaming media, video transcoding, image processing and signal processing applications.
Video processing applications will only require half the bandwidth under the new engine, while multiple GPUs will perform better and use less memory. Asynchronous copying of memory in the toolkit also betters application performance. Overall, the new toolkit and SDK promise to bring several improvements for the GPU.
Discuss in our Forums