Santa Clara (CA) – Nvidia has released a new version of its CUDA platform that exploits the GPGPU capabilities of its graphics processors. There are no new features, but the SDK now supports added support for Windows Vista, Red Hat Linux as well as VisualStudio 2008 on Windows XP. And we wonder what may have happened to the x86 version of the software that was promised twice year, but has not surfaced yet.
The standardization of OpenCL may have changed the GPGPU accelerated computing landscape, but Nvidia has no intention of giving up on its CUDA platform that still relies on a proprietary approach to tap into the processing potential of Nvidia graphics processors.
The company now offers version 2.1 beta as a free download, adding Windows Vista support for Tesla products and 32-bit debugger support for CUDA on RedHat Enterprise Linux 5.x, which is a separate download. Also supported is VisualStudio 2008 support on Windows XP and Vista and Just-In-Time (JIT) compilation for applications that dynamically generate CUDA kernels.
According to Nvidia, new interoperability APIs for Direct3D 9 and Direct3D 10 have been included to accelerate the communication with DirectX applications. There is also “a series of improvements to OpenGL interoperability” as well as “support for recent releases of Linux including Fedora9, OpenSuse 11 and Ubuntu 8.04.”
Nvidia, like AMD, recently announced that developers can use its SDK to create cross-platform OpenCL-based GPGPU applications. Proprietary developer solutions like CUDA have remained one of the main roadblocks in mainstream GPGPU software development. Earlier this year, Nvidia had planned to release a version of CUDA for x86 multi-core processors and later mentioned that the software was delayed to Q4 2008. Now our sources tell us that x86 CUDA may not be released at all, which raises questions how future-proof CUDA really is.
At least at this time it seems that the most conclusive development solution may be OpenCL, as the language is supported not only by Nvidia and AMD, but by Intel’s upcoming Larrabee graphics card as well.