Aurora: Architecting Argonne's First Exascale Supercomputer for Accelerated Scientific Discovery
Abstract
Aurora is Argonne National Laboratory's pioneering Exascale supercomputer, designed to accelerate scientific discovery with cutting-edge architectural innovations. Key new technologies include the Intel(TM) Xeon(TM) Data Center GPU Max Series (code-named Sapphire Rapids) with support for High Bandwidth Memory (HBM), alongside the Intel(TM) Data Center GPU Max Series (code-named Ponte Vecchio) on each compute node. Aurora also integrates the Distributed Asynchronous Object Storage (DAOS), a novel exascale storage solution, and leverages Intel's oneAPI programming environment. This paper presents an in-depth exploration of Aurora's node architecture, the HPE Slingshot interconnect, the supporting software ecosystem, and DAOS. We provide insights into standard benchmark performance and applications readiness efforts via Aurora's Early Science Program and the Exascale Computing Project.
Cite
@article{arxiv.2509.08207,
title = {Aurora: Architecting Argonne's First Exascale Supercomputer for Accelerated Scientific Discovery},
author = {William E. Allcock and Benjamin S. Allen and James Anchell and Victor Anisimov and Thomas Applencourt and Abhishek Bagusetty and Ramesh Balakrishnan and Riccardo Balin and Solomon Bekele and Colleen Bertoni and Cyrus Blackworth and Renzo Bustamante and Kevin Canada and John Carrier and Christopher Chan-nui and Lance C. Cheney and Taylor Childers and Paul Coffman and Susan Coghlan and Tanima Dey and Michael D'Mello and Ashok Emani and Murali Emani and Kyle G. Felker and Sam Foreman and Olivier Franza and Longfei Gao and Marta García and María Garzarán and Balazs Gerofi and Yasaman Ghadar and Subrata Goswami and Neha Gupta and Kevin Harms and Väinö Hatanpää and Brian Holland and Carissa Holohan and Brian Homerding and Khalid Hossain and Xue Hu and Louise Huot and Huda Ibeid and Joseph A. Insley and Sai Jayanthi and Hong Jiang and Wei Jiang and Xiao-Yong Jin and Jeongnim Kim and Christopher Knight and Panagiotis Kourdis and Kalyan Kumaran and JaeHyuk Kwack and Janghaeng Lee and Ti Leggett and Ben Lenard and Chris Lewis and Nevin Liber and Johann Lombardi and Raymond M. Loy and Ye Luo and Bethany Lusch and Nilakantan Mahadevan and Beth Markey and Victor A. Mateevitsi and Gordon McPheeters and Ryan Milner and Jerome Mitchell and Vitali A. Morozov and Servesh Muralidharan and Tom Musta and Mrigendra Nagar and Vikram Narayana and Marieme Ngom and Anthony-Trung Nguyen and Nathan Nichols and Aditya Nishtala and James C. Osborn and Michael E. Papka and Scott Parker and Saumil S. Patel and Julia Piotrowska and Adrian C. Pope and Sucheta Raghunanda and Esteban Rangel and Paul M. Rich and Katherine M. Riley and Silvio Rizzi and Kris Rowe and Varuni Sastry and Adam Scovel and Filippo Simini and Haritha Siddabathuni Som and Patrick Steinbrecher and Rick Stevens and Xinmin Tian and Peter Upton and Thomas Uram and Archit K. Vasan and Álvaro Vázquez-Mayagoitia and Kaushik Velusamy and Brice Videau and Venkatram Vishwanath and Brian Whitney and Timothy J. Williams and Michael Woodacre and Sam Zeltner and Chuanjun Zhang and Gengbin Zheng and Huihuo Zheng},
journal= {arXiv preprint arXiv:2509.08207},
year = {2025}
}
Comments
40 pages, 10 figures. Submitted to J. Supercomputing