Pioneering a new era of UK supercomputing

Reading time: 4 minutes

Bulletin speaks to Professor Mark Parsons, EPCC director, about how Edinburgh is pioneering the UK’s first next-generation supercomputer and the ways in which it could advance the fields of artificial intelligence, drug discovery, climate change, astrophysics and advanced engineering.

Supercomputers are already used in a variety of ways across the world such as in weather forecasting, aircraft design and the oil and gas industry. However, the University is now leading a new era of supercomputing after it was selected to host the UK’s first next-generation supercomputer, which is 50 times faster than any of the country’s existing machines.  

The exascale supercomputer is a high-performance computing (HPC) system that can perform one billion calculations per second. Once operational, it will provide supercomputer capability for key research and industry projects across the UK. 

Supercomputers were first introduced in the 1960s by Seymour Cray at Control Data Corporation (CDC) and have been used for science, technology and engineering ever since. Currently, the most powerful computer system in the UK is the ARCHER2 service, also housed in Edinburgh. The new exascale system will be approximately 50 times faster. 

The key to performance

Supercomputers tackle problems that are too large or too complex for a desktop computer to solve and use a process called parallel computing, which means carrying out many calculations at the same time on many processors at once.

Professor Mark Parsons, the director of the University’s EPCC supercomputing centre, explains: “Modern supercomputers are parallel supercomputers, which just means they’ve got lots of CPU cores in them. 

“The average laptop has four cores in it nowadays. ARCHER2 has got 750,080 cores in it so three-quarters of a million. If you want to use the power of that parallelism, you need to write parallel codes and that’s the tricky bit.” 

Is a supercomputer hard to use? Professor Parsons thinks it just requires some practice: “It is harder to use than a normal computer, although I have to say any researcher that’s used to a Linux workstation would find logging on to what we call the login nodes very familiar because it just looks like a Linux workstation. 

Usually the first time a researcher will use a big supercomputer is when they are doing their MSc or PhD and they will attend a bit of training. We do a lot of teaching around the UK and we’ve also been heavily involved in European training programmes.” 

Photograph by Paul Zanre

A new age of AI

Professor Parsons says the main use of supercomputers is for modelling and simulation but that exascale will also deliver an exciting step forward for AI research. 

“We’re at a very interesting time in the development of AI,” he explains. “Most AI researchers use processors called GPU accelerators and what we’re seeing around the world with this explosion of AI is people buying huge numbers of GPUs to do training.  This has coincided with the first generation of exascale supercomputers.” 

Professor Parsons continues: “The exascale supercomputer will have the same number of servers inside it as ARCHER2, about 6,000, but each server is 50 times more powerful than what’s in ARCHER2 and ARCHER2 is only a couple of years old. The difference is that an exascale server has four GPU accelerators in it. ARCHER2 doesn’t have any – all of the increased power comes from GPUs. 

“AI training uses GPUs so an exascale system, which will have about 24,000 GPUs, is the perfect type of supercomputer to do very large-scale AI training.”

Creating skilled jobs

Funding for the new supercomputer was announced by Chancellor Jeremy Hunt in the Spring Budget as part of an investment of around £900 million in both exascale computing and a separate AI Research Resource. 

The Treasury said it will “allow researchers to better understand climate change, power the discovery of new drugs and maximise our potential in AI”. 

Professor Parsons says one of the aims of the EPCC department is inspiring the next generation: “We’re currently looking at this huge demand for people who know how to run these big systems and develop applications for use by industry and by UK universities.  

“It’s very difficult to hire people with the necessary skills so one thing EPCC’s done for the last decade now is we bring in graduate apprentices from Napier University who are with us a year or longer. We train them up to be system management staff. At the same time, we also run an MSc in High Performance Computing and many of our applications developers have joined EPCC from this route. 

“We are currently thinking about creating an MSc in system management for high-performance computing systems for supercomputers and because a lot of the UK’s big organisations and big government labs simply can’t hire people with these skills. 

“I’m immensely proud that Edinburgh through EPCC has been chosen to host the UK’s first exascale system.” 

Find out more

EPCC

Advanced Computing Facility

Edinburgh Impact: Rise and rise of the machines