Argonne National Lab Polaris Pre-Exascale Supercomputer

U.S. Department of Energy’s Argonne National Laboratory and Hewlett Packard Enterprise (NYSE: HPE) unveiled a new testbed supercomputer to prepare critical workloads for future exascale systems that will deliver up to four times faster performance than Argonne’s current supercomputers.

Polaris will enable scientists and developers to test and optimize software codes and applications to tackle a range of artificial intelligence (AI), engineering and scientific projects planned for the forthcoming exascale supercomputer, Aurora, a joint collaboration between Argonne, Intel and HPE.

The $500+ million Exaflop Aurora was planned for 2021 but it has been delayed until 2022-2023. Aurora has been delayed waiting for Intel’s Sapphire Rapids server chips. The first plan was for a 180 petaflop Aurora for 2018 but delays in earlier Intel chips caused the need for a new plan.

Polaris will deliver approximately 44 petaflops of peak double precision performance and nearly 1.4 exaflops of theoretical AI performance, which is based on mixed-precision compute capabilities. Polaris 1.4 AI ExaFLOPS does not use standard FP64 (64 bit floating point) for standard supercomputer performance metrics.

It will be built using 280 HPE Apollo Gen10 Plus systems, which are HPC and AI architectures built for the exascale era and customized to include the following end-to-end solutions:

* Powerful compute to improve modeling, simulation and data-intensive workflows using 560 2nd and 3rd Gen AMD EPYC processors
* Supercharged AI capabilities to support data and image-intensive workloads while optimizing future exascale-level GPU-enabled deployments using 2240 NVIDIA® A100 Tensor Core GPUs, making it ALCF’s largest GPU-based system to date.
* Addressing demands for higher speed and congestion control for larger data-intensive and AI workloads with HPE Slingshot, the world’s only high-performance Ethernet fabric designed for HPC and AI solutions. HPE Slingshot will also be featured in Argonne’s Aurora exascale system.
* Enabling fine-grained centralized monitoring and management for optimal performance with HPE Performance Cluster Manager, a system management software solution.

The delivery and installation of Polaris is scheduled to begin this month. It will go into use starting early 2022 and will be open to the broader HPC community in spring of 2022 to prepare workloads for the next generation of DOE’s high performance computing resources.

SOURCES- ANL
Written By Brian Wang, Nextbigfuture.com