As you know, in 2007, NVIDIA introduced a closed standard CUDA, which involves the use of GPU power for specialized calculations. In those roles can be a video decoding, Mining cryptocurrency and much more. This reduces the load on the CPU. Such technology is already quite widespread in popular media- Media Player Classic, VLC, and other. It is also used in supercomputers to improve the calculations speed - Tesla modules .
CUDA, as mentioned above - a closed standard, which means its attachment to the GPU NVIDIA, and the need for licensing. Open analogue OpenCL appeared only in 2009, which is why its popularity has appeared regularly below. But it seems that the situation may change soon.
Recently, AMD company was presented to the compiler HCC (Heterogeneous Compute Compiler), and a layer of software compatibility with NVIDIA CUDA. Due to the features HIP (Heterogeneous Compute Interface for Portability) system can automatically convert up to 90% of CUDA code in C ++. Of course, this is not to say that CUDA will just work on AMD cards , as the company is investing resources in the development of OpenCL, and the licensing of the NVIDIA yet has not been canceled, but it will allow the company to strengthen its market supercomputing (HPC). Related Products :
|