The AI supercomputer from Nvidia and the University of Florida is said to be the most powerful of its kind in the academic environment.
Nvidia and the University of Florida plan to build the world’s fastest academic supercomputer in the academic world. The project will cost more than $ 70 million.
UF alumni and Nvidia co-founder Chris Malachowsky has provided $ 25 million, with an additional $ 25 million for hardware, software, training and other services coming directly from Nvidia. The University of Florida also plans to invest significantly more than $ 20 million to integrate the supercomputer into its campus data center.
The new supercomputer comes as an upgrade of the existing super computer “HiPerGator” from the University of Florida. The “HiPerGator 3” runs after the conversion to Nvidias DGX SuperPOD architecture. According to Nvidia, the new supercomputer should offer “700 petaflop AI performance” and will be launched in early 2021.
video-container">—
The system is with 140 DGX A100 Systems equipped by 1.120 A100 Tensor Core-GPUs be driven at four petabytes of high-performance storage. An Nvidia Mellanox HDR 200Gb / s InfiniBand network is designed to ensure high data throughput with low network latency.
30,000 young AI researchers are supposed to solve the big problems
The AI researchers at the University of Florida are to tackle the major challenges that await the world with the supercomputer: Artificial intelligence is to find new approaches and solutions for topics such as climate protection, rising sea levels, aging populations, data security, personalized medicine, and urban transport or food supply.
The University of Florida wants to do so by 2030 30,000 graduates with AI know-how have trained. The supercomputer plays an important role in this endeavor: it serves as a lighthouse that attracts talented youngsters with large AI projects. OpenAI’s new Text-AI GPT-3, for example, proves that more computing power can lead to better results in AI training – and the performance threshold has not yet been reached.
Nvidia also benefits from the cooperation: the young researchers learn how to use Nvidia hardware and software during their time at the university.
