KU Leuven and HPE Advance AI Capabilities With New Supercomputer

KU Leuven and HPE Advance AI Capabilities With New Supercomputer
HPE

Hewlett Packard Enterprise announced a new supercomputer installation at KU Leuven, a Flemish research university consistently ranked as one of the five most innovative universities in the world.

HPE collaborated with KU Leuven to develop and deploy Genius, a new supercomputer built to run AI workloads. The system will be available to both academia and the industry to build applications that drive scientific breakthroughs, economic growth and innovation in Flanders, the northern region of Belgium.

Supercomputers increasingly play a crucial role in generating actionable insights from vast amounts of data. Applications such as AI, for which HPC is a foundational technology, have evolved rapidly in recent years and are introducing new powerful methods in scientific research. AI has the promise to radically change economics, digital humanities, biomedical sciences and cosmology.

The new system at KU Leuven will be part of the Flemish Supercomputing Center (VSC) and available for KU Leuven and UHasselt researchers, as well as other VSC users, from academia and industry. It will support a very diverse research community in scientific areas like molecular modelling, engineering, physics, chemistry, climate studies, astronomy and astrophysics or psychology.

The university, together with HPE, also will provide training for researchers, equipping them with the knowledge and skills required to successfully and productively work with HPC systems in the future. Training activities will include CPU optimization techniques as well as GPU code modernization and optimization techniques. Specialized Machine Learning trainings also will be offered.

This HPE solution is based on new technologies that are purpose-built for HPC workloads, including HPE Apollo k6000 systems and HPE Gen10 compute servers designed to deliver 300 teraflops of computational power. They come with support of NVIDIA Tesla P100 GPU accelerators and emerging technology servers including the HPE Apollo sx40, which supports deep learning and HPC workloads.

For this system, each GPU could make the computational equivalent of two standard dual-socket servers. With four GPUs in one server, the compute power transforms from one server to 9. Genius is comprised of a cluster with 96 CPU compute nodes and 20 GPU machines with 4 GPUs per server; the result is a blazing fast system, particularly with AI workloads. Designed, built and supported by HPE, the deployment is located at the university’s main campus in Leuven in a state-of-the-art green datacenter.