Cambridge-1 is a supercomputer that Nvidia is going to build in Great Britain … But it’s not the one announced during the ARM buyout. No, this is another power monster, we were told during a call with officials from Nvidia. This second supercomputer has something to dream of British researchers and politicians, and perhaps facilitate the takeover of the famous chip designer.
Ultra powerful and deployed quickly
It is indeed expected to be the most powerful supercomputer in the UK, ranking 29e of the latest Top500 of the most powerful supercomputers in the world, with 8 Linpack performance Petaflops and 400 Artificial Intelligence computing Petaflops. Two different metrics that show both its raw power and its ability to adapt to the needs of the parallel calculations required by the algorithms of machine learning. Better yet, it should come third in the Green500, which lists supercomputers based on their energy efficiency.
To achieve such numbers, Cambridge-1 will bring together 80 DGX A100 systems. Each system carries no less than 1,024 Tesla V100s, Nvidia’s Rolls of Cards for cloud and scientific computing. The American giant will use what it calls its DGX SuperPOD architecture, scalable bricks, interconnected by technologies resulting from the takeover of Mellanox by Nvidia to build Cambridge-1. He will thus be able to deploy this supercomputer in a few months or years, but only in a few weeks. He says he plans to invest around 40 million pounds sterling (almost 44 million euros) in the project.
For science and health …
Nvidia has already announced that British pharmaceutical giant GlaxoSmithKline, also known by the acronym GSK, will be the first to harness this power. But other companies will obviously be able to access it, as will some academics, from King’s College London, for example, or from the Oxford Nanopore company, particularly at the forefront in the fight against Covid-19.
Thanks to Cambridge-1, Nvidia intends to participate in solving public health problems that could not be solved otherwise, because the data to be processed is too large and the available power too low. The giant of Santa Clara will therefore work with manufacturers, hospitals and laboratories.
He also announces that he will provide free access time to this supercomputer to academics who need it to find a specific cure. The University of Oxford is at the forefront of the implementation of a vaccine against covid-19, for example, and modeling the molecules or simulating the effects of the different proteins involved requires an enormous amount of time and / or power. Calculation.
Finally, Nvidia intends to make this second supercomputer available to start-ups and future researchers in artificial intelligence. A good way to be at the heart of innovation and what will emerge. Looks like medical innovation in the UK, thanks to AI, has just gone green …