AMD sets new goals for AI and HPC, Increase energy efficiency by 30 times by 2025

AMD announced an ambitious plan with the goal of using artificial intelligence (AI) training and high-performance computing (HPC) applications running on accelerated computing nodes by 2025, the energy efficiency of AMD EPYC series processors, and AMD Instinct computing cards will increase by 30 times.

AMD Executive Vice President and Chief Technology Officer Mark Papermaster said that AMD is very focused on these important market segments, as well as leading companies’ propositions on strengthening environmental management. Compared with the previous five years, setting a target of 30 times means that the energy efficiency performance in the field of artificial intelligence training and high-performance computing exceeds 150% of the industry. This reflects AMD’s technical strength and emphasis on environmental sustainability.

The accelerated computing node is the most powerful and advanced computing system in the world, used for scientific research and large-scale supercomputer simulations, it provides scientists with computing power to achieve breakthroughs in multiple fields, including materials science, weather prediction, medical research, and alternative energy development. It is also an indispensable part of artificial intelligence network training, currently used for activities including speech recognition and language translation. AMD’s 30-fold target means that the power required for these systems to complete a calculation will be reduced by 97% in the five-year period from now to 2025.

Improving the energy efficiency of accelerated computing applications is part of AMD’s new environmental, social, and governance (ESG) goal, which covers its operations, supply chain, and products. Due to its recent achievements in product energy efficiency, AMD was included in Fortune’s 2020 “Change the World” list in recognition of its social work efforts.