NVIDIA releases CUDA 5
We know that in December this year, will launch its new cGPU Nvidia Tesla K20 kernel-based second-generation kepler “GK110″, and which promises to bring GPU-accelerated computing (GPGPU) to new levels with its improved support and features ( compute capability 3.5 ).
But as we all know there is no point throwing a lot of promises surrounded hardware without the software that will take advantage, and as a prelude to the launch of Tesla K20, today introduces its new Nvidia CUDA Developer Production Release 5, which comes loaded with new features designed to make the most of their recent GPUs based on Kepler architecture of first and second generation (and GK11x GK10x nuclei), among which are:
- Dynamic Parallelism (New algorithms focused to simplify parallel programming).
- CUDA BLAS (libraries that allow interaction between CUDA and third-party libraries).
- RDMA for GPUDirect (“Remote Direct Memory Access” Better communication between multiple GPUs).
- Nsight Eclipse Edition (Lets develop, debug and profile applications within the Eclipse IDE GPGPU).
Also Nvidia has launched a new section called CUDA Resource Center, which contains documents that will help developers to take full advantage of Nvidia technologies, and can be downloaded free of charge , as well as the new CUDA 5 .
As a curious note, with Nvidia CUDA 5 promotes its new slogan: “The Future Is Heterogeneous” slogan very similar to those used by its competitor: “The Future is Fusion” (Slogan of the Fusion APU) and “The Future Is Heterogeneous Compute “(Slogan of Heterogeneous System Architecture ).
Link: NVIDIA Releases CUDA 5 (TechPowerUp)Tags: API, cGPU, CUDA, CUDA 5, developers, geforce, GK10x, GK11x, GPGPU, gpu, kepler, Nvidia, OpenACC, quadro, Tesla