NVIDIA to Support Healthcare by Creating a Supercomputer- Cambridge-1
Taking into account the need for focusing on healthcare research and resources to do that, NVIDIA has announced the launch of the UK’s fastest supercomputer; Cambridge-1, by the end of the year.
After acquiring ARM, NVIDIA makes the news with another major development for the UK intending to aid AI research for health care, first talked about by Jensen Huang, Chief Executive, at the GTC 2020. In his address, Huang mentioned how powerful resources with the ability to compute AI methods is a necessity to help solve healthcare-related problems, to be made available to scientists, researchers, and startups across the UK. It is intended to support innovation in the UK for inventive work in healthcare and drug discovery. Predicted to be out by the end of the year, it will rank as the world’s 29th most powerful computer and the most powerful in Britain.
As per Nvidia, the supercomputer is proposed to have 8 petaflops(measures the processing speed of a computer) of Linpack performance and 400 petaflops(1 petaflop equals1015 floating-point operations) of Linpack performance being an NVIDIA DGX SuperPOD system, ranking as the third most energy-efficient supercomputer in the world on the current Green500 list. Another merit of this system is the setup, which is presumed to require as little as a few weeks to complete, being powered by a connection of 80 NVIDIA DGX A100TMsystems together with NVIDIA Mellanox® InfiniBand networking. In terms of revenue generation for NVIDIA, it was stated that Cambridge-1 is “not a commercial endeavor”.
Furthermore, the two companies involved with the coronavirus vaccine, AstraZeneca and GSK, are to be the first two to equip the facilities of Cambridge-1. Also, researches from other organizations like the Guy’s and St Thomas’NHS foundation, Oxford Nanopore Technology, and King’s College London project to take advantage of this system.
Areas of focus
Aiming to support and empower health care and life science research in the UK, the four areas of main focus are:
Educate AI aspiring members: Cambridge-1 will be treated as the goal for world-class researchers besides providing great in-the-field experience to the coming generations.
University-granted compute time: To be donated as a resource for particular research projects, NVIDIA GPU time will help find cures.
Support AI startups: NVIDIA plans to come together with startups to provide early access to AI tools and guide the next generations providing great opportunities to explore and learn in this field.
Joint industry research: Some large-scale healthcare problems and data science problems could not be solved due to their size and computation requirements, which is where NVIDIA Cambridge-1 comes into the picture, helping solve these problems and resulting in improved success rates along with reduced healthcare costs.
Cambridge-1 is said to incorporate Clara Discovery by NVIDIA for its operation.
This is a tool suite that has been optimized for NVIDIA DGX that converges various technologies like radiology, imaging, and genomics for developing AI-based applications for computation expensive tasks for healthcare.
Clara’s guardian’s assistance for coronavirus concern is to check people’s temperature, monitoring social distancing using real-time computer vision methods, also helping with patient diagnosing in a contactless manner. This can also help with tasks like analysis of surgeries and tracking equipment during operation. Clara processes using all of the EGX edge AI chips from NVIDIA’s suite ranging from T4 edge inferencing servers to the Jetson chip that is embedded in devices.
NVIDIA Clara used pre-trained AI models that have an application-centered framework to help researchers with the making of new drug discoveries, by finding targets, building compounds, and even developing responses. In light of recent development in NLP, researchers can use biomedical-specific language models for organizing, understanding, and activating huge datasets, researching literature, and sorting through patents, papers, and other existing methods on real-world data.
Clara model’s featured pre-training has been used by a few organizations to determine whether a person coming in with Covid-19 symptoms is going to need supplementary oxygen in a few days or hours after the initial test. This works on health records and imaging to efficiently manage hospitalization in times like today when most countries are expected to see the second wave of Covid-19.
As per James Waterfall, Head of AI and Data Science, AstraZeneca, with the use of AI, supercomputing, and big data, it is possible to transform RnD, starting from target identification to the final release of new medicines.
And according to NVIDIA, its operations with GSK and AI groups integrate the use of genomic and genetic data for the improvement of treatment design, intending to deploy GSKwith GPU optimization and development tools like NVIDIA Clara providing access to Cambridge-1. As NVIDIA VP, Kimberly Powell said, NVIDIA in collaboration with GSK will explore the capabilities of AI using huge data resources for advancing the discovery of new vaccines and medicines.
Cambridge-1 has the power to perform a wide range of tasks ranging from discovering new drugs to diagnosing serious illnesses years in advance. It is believed that if you have access to large amounts of computational power, then focus can be shifted on asking the right question rather than focusing on the technical limitations.
NVIDIA also announced the availability of the NVIDIA DGX SuperPOD solutions which is the world’s first turnkey AI infrastructure. The DGX SuperPOD is ready to be shipped and is predicted to be installed in the UK, Korea, India, and Sweden before the year ends. These will be available in clusters of 20 to 140 individual units of NVIDIA DGX A100 models. When connected in a 20-unit module, interconnected using the NVIDIA Mellanox HDR InfiniBand networking, it provides an AI performance of 100 petaflops scalable up to 700 petaflops for processing the most complex AI computations.