HPE is lifting the lid on a brand new AI supercomputer – the second this week – aimed toward constructing and coaching bigger machine studying fashions to underpin analysis.
Based at HPE’s Center of Excellence in Grenoble, France, the brand new supercomputer is to be named Champollion after the French scholar who made advances in deciphering Egyptian hieroglyphs in the nineteenth century. It was constructed in partnership with Nvidia utilizing AMD-based Apollo laptop nodes fitted with Nvidia’s A100 GPUs.
Champollion brings collectively HPC and purpose-built AI applied sciences to coach machine studying fashions at scale and unlock outcomes quicker, HPE stated. HPE already gives HPC and AI sources from its Grenoble amenities for patrons, and the broader analysis group to entry, and stated it plans to offer entry to Champollion for scientists and engineers globally to speed up testing of their AI fashions and analysis.
Those purpose-built AI applied sciences consult with the HPE Machine Learning Development Environment, a software program platform which kinds a part of the HPE Machine Learning Development System that HPE launched final month.
Coincidentally, the HPE Machine Learning Development System can also be based mostly on AMD-based Apollo laptop nodes fitted with Nvidia GPUs, making it doubtless that Champollion is actually an incarnation of the HPE Machine Learning Development System.
The precise Champollion {hardware} specified by HPE contains 20 HPE Apollo 6500 Gen10 Plus server nodes, with 160 Nvidia A100 GPUs, and Nvidia Quantum InfiniBand networking. The HPE Machine Learning Development System begins at 4 nodes, however clients have the choice to scale up.
If it’s based mostly on that platform, every Apollo node can have 4TB of reminiscence and 30TB of NVMe native storage, with HPE Parallel File System Storage non-obligatory. The HPE Machine Learning Development Environment runs atop this and gives an built-in platform for constructing and coaching fashions, suitable with in style machine studying frameworks similar to TensorFlow and PyTorch.
HPE stated that Champollion is at present accessible to pick customers and might be launched to broader group entry in the close to future for customers to start creating and coaching their fashions.
Earlier this week, HPE additionally unveiled one other AI supercomputer on the Leibniz Supercomputing Center in Munich, pairing an HPE Superdome Flex server with a Cerebras CS-2 specialised AI system. That system can also be designed to speed up functions for the scientific and engineering group. ®
https://www.theregister.com/2022/05/27/hpe_champollion_supercomputer/
