Nvidia supercomputer with 80 DGX A100 accelerators should become UK’s most powerful

Spread the love
Nvidia wants to build the UK’s most powerful supercomputer this year, delivering 8 petaflops of computing power. The company will use eighty DGX A100 systems for this. The supercomputer will be available for medical research in Cambridge, and the Cambridge-1 supercomputer will be used for AI research into medical problems, such as the coronavirus, according to Nvidia. The company expects that the supercomputer can be deployed by the end of this year and can then deliver eight petaflops of computing power for the Linkpack benchmark. That would take it 29th place on the Top 500 list of the world’s most powerful supercomputers and third place on the Green500 list. Nvidia says the AI ​​computing power will be more than 400 petaflops. Thanks to the modular DGX SuperPOD architecture, which clusters the DGX A100 systems, Nvidia says it will be able to install the supercomputer and get it up and running in a matter of weeks. According to the company, building such a powerful supercomputer normally takes years. The DGX A100 consists of eight A100 accelerators with a total of 320GB of memory. The 80 DGX A100 systems will be linked via Nvidia Mellanox InfiniBand network equipment. With Cambridge-1 and the Clara Discovery software, Nvidia aims to help researchers investigate medical problems. The software, according to the company, combines ‘the power of radiology, genomics and medical imaging’ to develop AI applications that aid in medical research. The software can support drug and vaccination research, Nvidia writes, with the US company investing $ 44.1 million in Cambridge-1. Nvidia vice president of health care Kimberly Powell tells CNBC it’s not a “commercial plan.” Multiple pharmaceutical companies, start-ups and academics have plans for Cambridge-1, according to Nvidia. GSK, AstraZeneca and King’s College London, among others, want to work with Cambridge-1. Cambridge-1 will become part of Nvidia’s AI Center of Excellence, where Nvidia also wants to build a supercomputer with Arm CPUs.
You might also like