2 min

Tags in this article

, , ,

The first details are known about the new generation of supercomputers. The new project uses exasceral computing and AI to accelerate new discoveries. The system should be fully operational by the end of 2021.

The supercomputer, called Aurora, is built at the Argonne National Laboratory and is built by Intel for the Department of Energy (DOE). It’s the very first exascal supercomputer in the US.

It remains to be seen whether it will be the very first worldwide. Since last year, the Chinese have lost their top spot in the Top 500 ranking of supercomputers. Supercomputer Summit (143.5 petaflops) holds the top position and later in 2018 Sierra (94.6 petaflops) took second place. According to Science, however, China would work on an exasculate system that should be ready by 2020.

China

Despite the fact that China is expected to break open the exascale port first, Aurora is expected to take the top position in the supercomputers list by the end of 2021. However, these are all speculations, because the DOE only knows what developments are happening in the US.

Aurora has been planned as a classic supercomputer since 2015, but was scaled up to exascale in 2017. An exascale supercomputer can do 1018 calculations per second. The system is specifically optimised to analyse the data flow of the many DOE devices. These include telescopes, particle accelerators and various other detectors.

Given the recent growth in machine learning and additional benefits for science, Aurora is being set up as a solid platform for deep learning, says Rick Stevens, specialist in computing, environment and life sciences at Argonne. A total of 100 different AI applications are currently being developed across all national labs.

Cray & Intel

Aurora is built by Intel in collaboration with Cray and Argonne. The total contract is estimated at 500 million dollars. Intel doesn’t want to give details about Aurora’s architecture yet. The chips that are processed in Aurora are still being developed today and will probably not be finalised until 2020. As a result, no technical details about Aurora, such as computing power and power consumption, are known yet.

The DOE expects exascale computing to be an accelerator for scientific discoveries and machine learning methods. AI is becoming more and more commodity, making the exascal step crucial to look beyond what is possible today. Remarkable: according to The Verge, Aurora will be DOE’s first supercomputer that will not be used to develop weapons for the US.

Related: IBM unveils the world’s fastest supercomputer

This news article was automatically translated from Dutch to give Techzine.eu a head start. All news articles after September 1, 2019 are written in native English and NOT translated. All our background stories are written in native English as well. For more information read our launch article.