HPE moves to AI supercomputer in France
Reminiscent of the French linguist who deciphered the Rosetta Stone 150 years ago, Hewlett Packard Enterprise today introduced a tool to untangle its customers’ toughest problems.
The Champollion AI supercomputer takes its name from Jean-François Champollion (1790-1832), who decoded the hieroglyphs that opened a door to the study of ancient Egyptian culture. Like Champollion, the mega-system resides in Grenoble, France, where it will search for patterns in massive datasets at HPE’s Center of Excellence.
The work will include the development and training of AI models, as well as advanced simulations for scientific and industrial users.
Among the system’s global user community, researchers from France’s AI for Humanity program will use Champollion to advance industries and drive economic growth through machine learning.
Inside an AI supercomputer
Champollion will help HPE customers explore new opportunities through accelerated computing. The system is based on a cluster of 20 HPE Apollo 6500 Gen10 Plus systems running HPE Machine Learning Development Environment, a software stack for building and training AI models at scale.
It’s powered in part by 160 NVIDIA A100 Tensor Core GPUs, delivering 100 petaflops of peak AI performance for the cluster. They are linked on a high-speed, low-latency NVIDIA Quantum InfiniBand that integrates network computing.
The system can access NGC, NVIDIA’s online catalog for HPC and AI software, including tools like NVIDIA Triton Inference Server that orchestrates AI deployments, and application frameworks like NVIDIA Clara for healthcare. health.
Users can test and compare their own workloads on Champollion to speed up their work in production. It is the ideal tool for Grenoble, which is home to a dense cluster of research centers for energy, medical and high-tech companies.
The system could help find molecular models for a new, more effective drug or therapy. It could create a digital twin to explore more efficient ways to route logistics through a warehouse or factory.
The possibilities are as varied as the number of industries and research fields harnessing the power of high-performance computing.
It is therefore fitting that the Champollion system makes its debut in front of ISC, the largest European gathering of HPC developers. This year’s event in Hamburg will offer an in-person experience for the first time since the pandemic.
Whether you’re in Hamburg or online, join NVIDIA and watch the conference keynote, Supercomputing: the key to unlocking the next level of digital twins, to learn more about the potential of HPC+AI to transform all fields.
Rev Lebaredian, who leads NVIDIA Omniverse and simulation technology at NVIDIA, and Michele Melchiorre, senior vice president at BMW Group, will show how supercomputing can unlock a new level of opportunity with digital twins.
Feature image credit: Steven Zucker, Smarthistory.