Danish supercomputer to drive innovation
www.computerweekly.com
The Gefion AI Supercomputer (GAIS) project, which delivers Denmarks first artificial intelligence (AI) turbo-charged supercomputer, has positioned Denmark as the most advanced of the Nordic regions quantum computing investing nations.It also serves to accelerate the use of AI to drive innovation across Denmarks business and industrial sectors.Built on the Nvidia DGX SuperPOD AI supercomputer, GAIS is powered by 1,528 Nvidia H100 Tensor Core graphics processing units (GPUs) and interconnected using Nvidia Quantum-2 InfiniBand networking.GAIS will be used to tackle high-performance computing (HPC) workloads using a full-stack datacentre platform.The Danish Centre for AI Innovation (DCAI), the company established to operate Gefion, is drawing a significant part of its funding from two principal investors and owners, the state-controlled Export and Investment Fund of Denmark (EIFO); and Danish philanthropic institution Novo Nordisk Foundation (NNF). The NNF acquired an 85% equity holding in DCAI and the EIFO owns the remaining 15%.This groundbreaking private-public partnership has committed 95m in startup development capital to the GAIS project over its initial stage. The NNF is contributing 81m and the EIFO the remainder. The former, a charitable foundation, has a net worth of almost 163bn. Last year, it distributed grants of over 1.3bn to qualifying research and innovation projects.The DCAIs goal is to have the GAIS resource generally available to researchers, industry and other selected users by June 2025, said Nadia Carlsten, the companys CEO. Gefions launch is a remarkable achievement given it took only six months from disclosure of the project to completion for what is a complex state-of-the-art supercomputer, she said. We now have a leading capability that is ready for customers to start testing new use cases on. Gefion is the game-changer that many innovators in Denmark had been waiting for.Under an agreement reached between the owners and stakeholders, all revenues generated from GAIS will be reinvested in the DCAI managed project. The turbo-charged computer will be operated on 100% renewable energy sources.The GAIS was built by Eviden, the high-performance computing and systems integration subsidiary of French IT firm Atos.The pilot phase of GAIS will focus on making it ready-to-use for selected customers that are using AI to accelerate innovation in a broad range of areas; including quantum computing, drug discovery, societal challenges, deep learning and the transition to green energy.The Danish Meteorological Institute (DMI) is among the first scientific agencies selected to have access to GAIS during the pilot test stage. The DMIs meteorologists are using GAIS to develop a Danish weather model, based on 33 years of recorded data, that can predict weather and climate events with greater accuracy.The superpower components in GAIS will present Danish enterprises and organisations with entirely new opportunities, said Morten Bdskov, the Danish industry, business and financial affairs minister.Gefion has the potential to drive innovation and advance the green transition, he said. It will enable tailor-made solutions, and strengthen the competitive standing of Danish companies in global markets. At a time of heightened international competition, this AI supercomputer, and its groundbreaking technology, can propel Danish businesses into the future. Significantly, Gefion will help Danish companies and researchers develop innovative solutions to future challenges.Read more about supercomputersDenmark had no GPU-accelerated supercomputers in use before the launch of GAIS. This lack of access to supercharged computing power was widely regarded by Danish legislators and technology sector leaders as a significant roadblock to further bolstering the countrys AI-based research ecosystem.GAIS will provide fresh momentum to drive future AI-based research, innovation and growth opportunities for enterprises of all sizes in Denmark, said Peder Lundquist, CEO at EIFO. Denmark is already an important player in AI-based research and business development, he said. The power and innovation behind Gefion, and the possibilities it brings, can accelerate development many times over. EIFOs investment in DCAI will support Danish businesses in key areas like green transition and life science. It will also accelerate the commercialisation of research.Collaboration between the EIFO, NNF and the DCAI is happening at a time when the state agency is further scaling-up its partnership-based investments in innovative deep tech projects, particularly within its strategic focus areas, AI; quantum computing; the green transition; life sciences; and the commercialisation of research.The Gefion AI supercomputer is officially ranked 21st on the Top 500 list of the worlds most powerful supercomputers. It will enable research and innovation at levels previously unattainable in Denmark , said Carlsten.Gefions number-21 ranking on the Top 500 list is a milestone for DCAI and Denmarks global standing in AI and supercomputing, she said. This recognition is not only proof of the advanced technology we are able to provide to our customers, but it also reflects our ambitions to drive leading edge scientific research and innovation that was previously not possible to do in Denmark.As part of the lead-in to the central project, the DCAI ran a competition in 2024 to identify innovative AI use cases that had the capability to fully harness the potential of GAIS. The DCAI selected six out of a total of 50 proposals received to run during the pilot phase.Among the approved pilot projects is a proposal from the University of Copenhagen to test large-scale distributed simulation of quantum algorithms for quantifying molecular recognition processes.The Technical University of Denmark will use GAIS to explore CO2 reduction methods in non-metal formate dehydrogenase using machine-learned force fields. In addition, the Danish Meteorological Institute will employ GAIS to develop proficient atmospheric prediction with intelligent environmental network models.
0 التعليقات ·0 المشاركات ·12 مشاهدة