At the latest supercomputing conference, Huang Renxun, founder and CEO of NVIDIA, announced a reference design for building GPU accelerated arm server, which greatly expanded GPU accelerated support for supercomputers.
Arm architecture can be seen everywhere in smart phones and Internet of things devices. It has already become the most popular processor architecture in the world. Huang Renxun predicted that arm now covers more than 100 billion computing devices, and the number will exceed 1 trillion in the next few years.
NVIDIA GPU accelerated arm server reference design includes hardware and software basic modules, which are supported by high-performance computing and main members of arm ecosystem. Almost all works that can run in high-performance computing should be able to run on any CPU. At present, 30 applications have been started and started to run.
For example, several of the world's top supercomputing centers have begun testing, including the Oak Ridge National Laboratory and Sandia National Laboratory in the United States, the University of Bristol in the United Kingdom, and the Japan Institute of science and chemistry.
At the same time, NVIDIA is cooperating with arm, ampere, Fujitsu, Meiman electronics, HPE and its Cray and other ecological partners. Many high-performance computing software companies have begun to use NVIDIA cuda-x library to bring GPU based management and monitoring tools for arm ecology.
Earlier this year, NVIDIA announced that cuda-x software platform was launched into arm ecosystem. According to previous commitments, NVIDIA is providing a preview version of arm compatible software development suite, including NVIDIA cuda-x library and accelerated computing development tools.