Europe's Leonardo pre-exascale system is set for its official inauguration this week following its confirmation as the fourth most powerful supercomputer on the Top500 list at the recent SC22 conference.

Leonardo is the second of the [1]EuroHPC Joint Undertaking's pre-exascale supercomputers to enter into service following the [2]LUMI system in Finland back in June. It is expected to be capable of a performance level of more than 249 petaflops when fully operational, and will be officially opened up for access by scientists and engineers at the Bologna Technopole in Italy on November 24.

The move means that Europe now has two of the four most powerful supercomputers as ranked in the [3]Top500 list , behind the US Frontier exascale system and Japan's Arm-based Fugaku system.

[4]

As with most supercomputers, it will be used for a variety of the most demanding applications including materials science, biomedicine, climate change, engineering, modeling of the human brain, and AI development.

[5]

[6]

The [7]Leonardo system itself is built by French IT outfit Atos and based on its BullSequana XH2000 architecture. It comprises two main computing modules, named Booster and Data Centric, to enable it to cover a range of different workloads.

According to [8]Atos , the system is equipped with approximately 3,500 Intel Xeon processors and 14,000 Nvidia A100 GPUs. The Booster has 3456 Intel Ice Lake computing nodes, while the Data Centric module has 1536 nodes. The later is based on BullSequana X2140 three-node CPU blades with Intel's 4th Generation Sapphire Rapids Xeon Scalable processors, presumably coming in 2023 when the chips are set to be available.

[9]

Leonardo will also be augmented by the integration of quantum processors as accelerators in future, according to EuroHPC.

[10]HPC's lost histories will power the future of tech

[11]Nvidia H100-based Henri supercomputer tests AMD's claim on Green500

[12]Azure-based 'AI supercomputer' to use Nvidia GPUs and software

[13]Cerebras's supercomputer has more cores than world's fastest iron – with a big catch

This is set to be joined next year by another pre-exascale supercomputer, the MareNostrum 5 at the [14]Barcelona Supercomputing Center in Spain. This is also being built by Atos and will pair Intel's Sapphire Rapids processors with Nvidia Grace chips.

MareNostrum 5 is being designed more for general-purpose compute and AI, according to EuroHPC, and will feature a different architecture to Leonardo and LUMI.

These are stepping stones towards the [15]first European exascale supercomputer , Jupiter, which will be based at the Jülich Supercomputing Centre (JSC) in Germany and is expected to arrive sometime in 2023/2024, according to EuroHPC.

However, the senior European Commission official speaking to the media said that the tender for Jupiter has yet to be awarded, but the system is expected to be capable of "1,000-plus petaflops" – exascale – when fully operational.

[16]

"It will be a bit faster than the current US supercomputer [Frontier], and we're hoping to be the number one machine in the world," the official said. It remains to be seen how it will compare with Aurora, the exascale system based at the Argonne National Laboratory in the US, which has been delayed because of problems Intel has had in delivering Sapphire Rapids processors.

All of the EuroHPC supercomputers will be available via cloud-based services offered through a federated HPC infrastructure with terabit connectivity, according to slides shown by the official.

"That's quite an impressive network, if you consider that in 2018 we had nothing," the official said, claiming that "Europe is now a computing superpower." ®

Get our [17]Tech Resources



[1] https://eurohpc-ju.europa.eu/index_en

[2] https://www.theregister.com/2022/06/13/finlands_amdpowered_lumi_arrives/

[3] https://www.theregister.com/2022/11/14/frontier_top500_aurora/

[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/hpc&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2Y3yrydr1O6B47AVSynjJVgAAAIc&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0

[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/hpc&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Y3yrydr1O6B47AVSynjJVgAAAIc&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0

[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/hpc&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33Y3yrydr1O6B47AVSynjJVgAAAIc&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0

[7] https://leonardo-supercomputer.cineca.eu/

[8] https://atos.net/en/2022/press-release_2022_11_14/atos-delivers-eurohpc-leonardo-supercomputer-ranked-4th-among-the-worlds-most-powerful-supercomputers

[9] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/hpc&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Y3yrydr1O6B47AVSynjJVgAAAIc&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0

[10] https://www.theregister.com/2022/11/21/hpc_opinion_column/

[11] https://www.theregister.com/2022/11/17/henri_green500_nvidia/

[12] https://www.theregister.com/2022/11/17/azure_nvidia_ai_supercomputer/

[13] https://www.theregister.com/2022/11/15/cerebras_supercomputer_frontier/

[14] https://www.bsc.es/

[15] https://www.theregister.com/2022/06/16/first_european_exascale/

[16] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/hpc&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33Y3yrydr1O6B47AVSynjJVgAAAIc&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0

[17] https://whitepapers.theregister.com/