Aurora (supercomputer)
| Company type | Project |
|---|---|
| Industry | Supercomputing |
| Founded | 2015 |
| Founder | United States Department of Energy |
| Headquarters | Argonne, Illinois , United States |
Key people | Olivier Franza (chief architect and lead investigator) |
| Products | Exascale supercomputer |
| Owner | United States Department of Energy |
Aurora is an exascale supercomputer funded by the United States Department of Energy (DOE) and engineered through a collaboration between Intel and Cray for use at Argonne National Laboratory. It held the position of the world's second fastest supercomputer from November 2023 until June 2024.
The projected expenditure in 2019 was approximately US$500 million. Olivier Franza acts as the chief architect and lead investigator overseeing its design.
History
The DOE introduced plans in 2013 for an exascale supercomputer aiming to reach speeds close to 1 exaFLOP (1018 floating point operations per second) while limiting power consumption to 20 MW by 2020. Aurora was unveiled in 2015 with an original completion target of 2018, designed to deliver 180 petaFLOPS, a performance level similar to Summit. The goal was for it to become the most powerful machine upon launch, constructed by Cray utilizing Intel processors.
In 2017, Intel postponed Aurora's delivery to 2021 and increased its target performance to 1 exaFLOP. By March 2019, the DOE reaffirmed its commitment to deploying the first US exascale supercomputer by 2021. Nevertheless, in October 2020, Aurora encountered an additional six-month delay, resulting in it losing the lead as the initial US exascale system.
Towards the end of 2021, Intel declared that Aurora would surpass 2 exaFLOPS peak double-precision computing power—though this milestone was never achieved. Full installation of the system was completed by June 22, 2023.
As of May 2024, Aurora occupied the second spot on the Top500 list with a measured performance of 1.012 exaFLOPS, making it the second exascale computer featured on that ranking.
Usage
Aurora facilitates scientific investigations into brain mapping, nuclear fusion, sustainable energy technologies, particle physics, oncology, and cosmology. Additionally, it supports innovation in new battery materials and more efficient solar cells. The platform is available for use by the broader scientific research community.
Architecture
The system comprises 10,624 nodes, each equipped with two Intel Xeon Max CPUs, six Intel Max series GPUs, and a unified memory design, achieving up to 130 teraFLOPS per node. Its total memory capacity is around 10 petabytes, coupled with storage reaching approximately 230 petabytes.
Aurora's power consumption is roughly 39 MW. For perspective, the current fastest supercomputer, El Capitan, operates at about 30 MW, while Frontier uses around 24 MW.
See also
- ARM supercomputers
- El Capitan (supercomputer)
- Fugaku (supercomputer)
- List of fastest computers
- TOP500