In an impressive feat, Intel has announced that its Aurora supercomputer, still under construction in partnership with Argonne National Laboratory and HPE, has broken the pivotal exascale computing barrier by achieving 1.012 exaflops of performance. Even more remarkably, Aurora also became the world’s fastest AI system dedicated to open science research by hitting 10.6 AI exaflops.
Key Details:
- Aurora is the second supercomputer after Frontier to surpass 1 exaflop, an exceptional milestone.
- This performance was achieved with only half of Aurora’s final configuration currently operational.
- Once fully built, it will consist of 10,624 blades, 21,248 Xeon Max CPUs, and 63,744 Max GPUs across 166 racks.
- Designed for AI research, applications include climate modeling, fusion energy, atomistic simulations, and more.
- Despite being incomplete, Aurora ranked #2 on November 2023’s TOP500 supercomputer list.
Intel and partners aim for Aurora to unlock major scientific breakthroughs across multiple domains by providing exascale-class computing capabilities optimized for AI workloads in the field of open science.
The fact that the partially constructed system has already broken the exascale barrier shows incredible promise for what the complete Aurora supercomputer can achieve once fully commissioned.
