‘NVIDIA’s Cambridge-1 supercomputer will be a force for life sciences’
While NVIDIA awaits regulatory approval for its $40bn bid for Arm, the US technology company is progressing its vision for a starring role in the Cambridge ecosystem by engaging with the local community.
The first two major goals are creating an AI centre of excellence in the centre of the city, and launching the UK’s most powerful supercomputer, the Cambridge-1.
Among the first companies to harness Cambridge-1 for research will be GSK and AstraZeneca. Additionally, researchers from Guy’s and St Thomas’ NHS Foundation Trust, King’s College London and Oxford Nanopore Technologies plan to use the system.
The inward investment will start to yield results in the first half of this year, says David Hogan, VP Enterprise EMEA (Europe, Middle East, Asia) at NVIDIA. David discussed the company’s plans at an online Cambridge Network event on January 28.
“We had a very, very positive response,” he says of the occasion – NVIDIA’s first event as a member of Cambridge Network.
“The ecosystem really outreached to us and for us that’s part of the journey.”
The next step will be the Cambridge-1 launch.
“The Cambridge-1 is for grand challenges, where a lot of research needs to be done – Covid research for instance, which has huge power requirements: we expect to be operational in the first half of the year.”
The supercomputer’s data-crunching capabilities will see it ranked number 29 on the latest TOP500 list of the world’s most powerful supercomputers, and in the top three for energy-efficient supercomputers on the Green500 list. The data centre that handles the mechanics – data centres being a key NVIDIA revenue stream – will be located in Harlow.
“The data centre there is hosted by KAO Data Centres – they host the system and we run the platform,” says David.
Yet no quantum involvement?
“Quantum is a different technology,” replies David.
“They can use NVIDIA technology and a lot of work to enable quantum is being done with NVIDIA GPUs [graphics processing units], like modeling quantum computers. It is an area of research for us – and others – looking at what quantum could and couldn’t be.”
Along with the supercomputer comes a proposed new AI centre of excellence featuring a new Arm-based supercomputer.
“If the deal is approved this will be an international centre of artificial intelligence using NVIDIA and Arm architecture,” says David, “and around that will be built a centre of excellence to attract world-leading talent around the development of AI, especially for start-ups.”
NVIDIA’s Inception, its AI acceleration platform, data science and high performance computing start-ups, will provide go-to-market support, expertise, and technology.
“It’s a big opportunity to upskill people in terms of knowledge and expertise,” David notes.
“So, for example, the US and China are doing more in this space, so we are looking to create real breakthroughs in science and technology in the UK, for wider use.
“For Cambridge, one of the central points is that it’s a key location – with Arm based in the city – as we go forward.”
It is as well to consider that NVIDIA may see buying Arm as the jewel in the crown of its booming data centre business because an NVIDIA-owned Arm brings challenges for the two companies’ very different licensing models. Arm’s success is based on its open licensing model, which allows it to sell its chips designs to technology and automotive companies – including ostensible competitors – across the globe.
The challenge is for the engineers in the individual companies to make best use of the architecture. So would NVIDIA’s chip arm adopt an open licensing model to become compatible with Arm’s strategic reach?
“One reason to acquire Arm is to invest in that model, to enable as many people as possible to build on their success, and to infuse NVIDIA’s intellectual property to enhance Arm’s existing architecture,” David says of the arrangement.
“Arm has been fantastic and, with NVIDIA, there is a lot more capability – the Internet of Things, for example, has not been realised to its full extent at the moment. What’s starting to happen is that machines are talking to machines, with not thousands but billions of interconnections every day, traffic control for instance, or 5G, virtual reality and augmented reality.
“Some systems need to be independent, such as autonomous cars, so all these things need to be able to interact continuously: Arm and NVIDIA, together, transforms how that happens. So we see value at the edge – it’s already there on mobile – but also for data centres.”
Edge computing is a distributed computing paradigm that brings computation and data storage closer to the location where it is needed, to improve response times and save bandwidth. It is crucial to the development of autonomous cars, but the big question is how to ensure this functionality is available on the cloud?
“It takes place on as many locations as possible – cloud providers, or data centres, it depends on the workload. There’s lots of data flowing back from the cloud but some data comes from an on-premise data centre so, with certain data, that can be challenging.”
NVIDIA is making significant progress on autonomous driving. New software-defined architecture built on the NVIDIA DRIVE platform will be incorporated as standard in Mercedes-Benz’s next-generation fleet. A primary feature will be the ability to automate driving regular routes from address to address.
“NVIDIA’s DRIVE platform for Mercedes is a great example of what we’re doing,” says David. “It involves multiple AI for a smart system in vehicles. In addition, when you’re not driving you have the opportunity to have experiences in that vehicle, it could be for entertainment or for work, so the car becomes a very different platform, it’s not just enabling autonomous driving.”
The autonomous era is already kicking in.
“Level 2 is where the industry is going right now,” explains David. “That’s assisted driving, the car guides you, and takes proactive action to keep you safe.
“Levels 3 and 4 could be autonomous, including motorway driving or city driving where well mapped. Level 5 is no steering wheel.
“We’ll be in level 3 or 4 for the next two or three years, and level 5 from 2025 – that’s dependent on regulatory approvals as well as building the car.”
One of the more intriguing possibilities of the Arm/NVIDIA compact – assuming regulatory approval – will be whether Arm becomes more games-friendly. Under Softbank the company seemed to be steering away from games, but NVIDIA has had a deep games pedigree since it started trading in 1993 – and its recently launched GeForce platform lets you play games anywhere, on any device.
“Our gaming heritage has enabled us to do some amazing things,” notes David, “but in 2012 our founder Jensen Huang took a big decision to pivot the company towards AI, so for us this [Cambridge development] is part of that journey.”
Indeed, in 2020, NVIDIA’s data centre sales surpassed its gaming revenue for the first time since it was founded – but the future still awaits the outcome of the regulatory process.
“We suspect the whole regulatory process probably has another year to run,” David concludes. “We’ve estimated 18 months, and meanwhile we’ll continue our engagement and in the next few months we have the Cambridge-1 launch and the GTC Conference in April, where delegates can learn about AI ...we’re expecting 100,000 people to attend.”
Register here for the GTC Conference.