World’s largest AI chip installed in Scotland

The University of Edinburgh has launched a supercomputing cluster installed with the largest AI chip ever built.

Operated by the EPCC, the supercomputing centre at the university and part of the Edinburgh International Data Facility, the cluster is made up of four Cerebras CS-3, third generation wafer scale engine processors, making it the largest in Europe.

The system is set to enable a “more democratic approach to training AI models” allowing for “groundbreaking” research and innovation, the university said.

It is understood the technology allows for scientists and machine learning practitioners from other disciplines, not only computer science, to start building, training and using models with no need for complex parallel programming. It also provides the ability to scale linearly, which allows for predictability of the efforts that is unmatched by other technologies.

The new service is part of the AI service provided to the Edinburgh & South East Scotland City Region Deal as part of the Data Driven Innovation programme, which the university has led since 2018.

Professor Mark Parsons, EPCC director, said: “AI is transforming all of our lives, and this investment will help universities, public sector organisations, and companies – large and small – to train and use AI models at speeds and with ease no other AI technologies can match.”

With the new service, the supercomputing centre will be able to train models of up to one trillion parameters, and fine tune 70 million models in a day.

Parameters are internal variables that an AI model adjusts during training to improve its ability to make accurate predictions, meaning that those with higher parameters can perform better at complex tasks.

The cluster will also allow the centre to continue its AI research into parallelism (a process that divides large compute tasks into smaller ones that multiple processors can solve simultaneously) and energy efficiency.

Andrew Feldman, chief executive of Cerebras, said: “EPCC has done pioneering work in enabling the next generation of AI breakthroughs. With our largest installation in the European continent to date, we’re excited to be part of such an important initiative that will enable researchers and institutions to drive innovation and shape the future of AI and HPC at a scale previously thought impossible.”

The launch of the system builds on previous work by the EPPC with the Cerebra system.

The CS-2 systems at EPCC have already allowed researchers to develop highly optimised wafer-scale inference software for large language models to explore their use for biomedical AI, and for local company, smartR AI to explore fine-tuning of LLMs on the system.

The EPCC systems have also helped researchers in India to develop an LLM for materials science and, in Switzerland, to adapt LLMs to better support Swiss German dialects.