Statistics of platform usage
Modern customer video cards that can be used for general purpose (non-graphic) applications are provided mainly by two vendors, nVidia and AMD. The specifications of the manufactured GPU are collected on the Hardware page of hgpu.org. Introduction of the CUDA proramming language for nVidia graphics processing units, which had made GPGPU calculations easier than before, cause appearence of the large number of GPU applications. Some common libraries (e.g. BLAS, FFT) have also been developed for GPU with the use of CUDA. Most of the application reported up to date are realized on nVidia platform. The statistics page provides information about the platforms (hardware and software) on which the works posted on hgpu.org have been performed.
Statistics is based on 13055 available papers
Hardware percentage of all papers posted on hgpu.org. Unknown platforms include all other parallel processing units (Cell processors, etc.) or information about the platform is not available.
The number of the papers per year describing the applications developed for various graphics processing units.
Full-history use of established GPU platforms. The final year is year to date.
Full-history use of emerging GPU platforms. A separate scale makes newer platforms visible.
Full-history use of NVIDIA data-centre accelerators.
Full-history use of AMD data-centre accelerators.
