Graphcore IPUs will feature in a new supercomputer designed to advance cutting-edge scientific research projects in the USA.
ACES (Accelerating Computing for Emerging Sciences) is described as a ‘holistic’ computing system, bringing together a range of state-of-the-art technologies in a single platform, with users selecting the hardware that best suits their specific application.
Graphcore IPUs will deliver high-performance AI computation, alongside other forms of compute, running on hardware from Intel, NEC, NextSilicon, and others.
ACES will be built by researchers from Texas A&M University, the University of Illinois at Urbana-Champaign and the University of Texas at Austin, and has been made possible by a $5m grant from the National Science Foundation.
Researchers from a broad range of disciplines will have access to ACES, including fields as diverse as genomics and bioinformatics, life sciences, climate modelling, imaging and quantum computing.
As the quantity and complexity of data continues to grow, scientists are increasingly adopting AI techniques in their research applications. Whereas traditional processors were not built to efficiently process AI, novel hardware such as Graphcore’s IPU systems has been designed to accelerate machine learning algorithms and, as a result, is often better suited to modern research workloads. Thanks to its multiple instruction, multiple data architecture (MIMD), IPU systems can also accelerate the algorithms behind many traditional HPC research applications. Further examples of scientific research accelerated by Graphcore IPUs is available on our Research Papers portal.
The ACES supercomputer is expected to be deployed and ready for use by September 2022. Until then, scientific researchers anywhere in the world can request access to IPU systems by applying for Graphcore’s Academic Programme (university researchers only) or contacting our Sales team.
More information on the ACES project can be found on the NSF website.
Sign up for Graphcore updates:
Jan 31, 2023 \ Healthcare, Drug Discovery, Scientific Research, GNN, Mixture of Experts