Nov. 19, 2024 — The new Top500 list of the fastest supercomputers was released Monday at the SC24 conference in Atlanta, Georgia, USA. LUMI, EuroHPC Joint Undertaking’s flagship supercomputer operated by the LUMI consortium, positioned 8th on the Top500 reaching a measured High-Performance Linpack (HPL) performance of 379.7 petaflops on this biannual list. Over 2200 scientific projects have been running on LUMI, and over half of LUMI’s computing resources have been utilized for AI-related research and innovation.
The list was topped by the El Capitan supercomputer located at the Lawrence Livermore National Laboratories in the USA, and Europe’s fastest supercomputer on the list was the HPC6 system installed at Eni S.p.A energy company’s data center in Ferrera Erbognone, Italy, placing 5th. LUMI entered the Top500 list for the first time in June 2022 and kept the position of Europe’s fastest supercomputer for 30 consecutive months.
LUMI losing its position as Europe’s fastest was expected and results from rapid technological development that brings faster supercomputers to the market continuously. This underlines the need for the European Union and its Member States to remain ambitious with supercomputing investments to ensure that European researchers and businesses continue to have access to world-class tools for data-intensive research and innovation that is needed for Europe’s competitiveness.
“LUMI’s remarkable computing power has to date enabled around 2200 scientific projects. What is more important than the power as such, is the impact. According to an independent study made in early 2024, one euro invested in CSC’s HPC services brings back 25–37 euros in return as direct economic impact. Academic and industrial users throughout Europe have been able to make discoveries using LUMI’s computing platform in a wide range of disciplines ranging from large language models, digital twins to quantum technologies and more. It’s also noteworthy that more than half of LUMI’s capacity has been used for AI-related research and innovation,” says Kimmo Koski, Managing Director of CSC – IT Center for Science, on behalf of the LUMI consortium.
“LUMI is a cornerstone of Europe’s digital sovereignty and scientific excellence. As a EuroHPC pre-exascale supercomputer, it has consistently set benchmarks in computational power and sustainability since 2022, driving Europe’s leadership in research, innovation, and tackling global challenges. In this fast-evolving technological landscape, EuroHPC JU is committed to continuously advancing Europe’s supercomputing capabilities. Our goal is to equip researchers and industries with the cutting-edge tools needed for innovation, investing in the next generation of EuroHPC systems and ensuring Europe’s lasting leadership in supercomputing,” says Anders Dam Jensen, Executive Director of the EuroHPC Joint Undertaking.
So far, over 3000 researchers across Europe have used more than 280 million GPU-node-hours and over 5700 million CPU-core-hours on LUMI.
Some highlights of LUMI’s use from this year include SiloAI and the University of Turku’s large language models developed on LUMI, numerous industrial examples such as the Norwegian DigiFarm, which is revolutionizing agriculture using HPC and AI, the development of multiple digital twins such as the Destination Earth Platform plus countless scientific publications and discoveries made using LUMI’s immense computing power. LUMI’s journey is far from over, but we have already started preparations for the EU’s upcoming AI Factories, Koski summarizes.
One of the Leading AI Platforms for Science
In addition to the Top500 list, several other new benchmark lists were published at SC24.
LUMI was listed in the HPL-MxP benchmark as number 3 with its 2.350 exaflops performance. The HPL-MxP benchmark probes the system capability for converged high-performance computing (HPC) and artificial intelligence (AI) workloads, which was one of the design goals of LUMI. This makes LUMI one of the leading AI platforms for science in the world, which has also been proven with the many scientific projects utilizing LUMI’s revolutionary AI capacity.
On the latest HPCG list (High-Performance Conjugate Gradient) LUMI was ranked number 4, with a result of 4.587 pflop/s. The HPCG benchmark provides an alternative metric for assessing supercomputer performance from the memory-bandwidth bound applications point of view, and is meant to complement the HPL measurement.
LUMI Helps to Fight Climate Change
LUMI supercomputer reached the 25th spot on the new Green500 list which biannually ranks supercomputers from the Top500 list in terms of energy efficiency. The GFlops/Watts ratio for LUMI is 53.4, making it one of the greenest supercomputers in the world.
LUMI’s climate and environmental impact is based on the whole lifecycle of the data centre. Located in an existing industrial facility it didn’t produce construction emissions. Using 100% renewable energy, a cooling system allowing for world-class energy efficiency and re-using of waste heat makes the carbon footprint significantly smaller.
In addition, the purpose of the data center is among others to provide a better understanding of climate change and tools for fighting it. This could be thought as a positive carbon handprint, meaning the positive impact on climate and environment through the green research and innovation it enables. LUMI is a key computational platform in the European Commission’s Destination Earth flagship initiative. The projects running on LUMI are related to the Climate Change Adaptation digital twin, the Extreme Weather digital twin, the Biodiversity digital twin, and other digital twins. These digital twins advance researchers’ knowledge of climate change and how to best predict it and adapt to it, including how to sustain biodiversity.
The LUMI system is supplied by Hewlett Packard Enterprise (HPE), based on an HPE Cray EX supercomputer. The supercomputer is owned by the EuroHPC Joint Undertaking and hosted by the LUMI consortium, including eleven European countries.
Source: LUMI
This post was originally published on here
