New Lenovo Supercomputer for Potsdam Institute for Climate Impact Research

New Lenovo Supercomputer for Potsdam Institute for Climate Impact Research

Lenovo is installing a powerful high-performance computer (HPC) at the Potsdam Institute for Climate Impact Research (PIK), a world-leading research institution in its field. This is expected to be operational by the end of August as HPC partner pro-com ensures a seamless integration. The inauguration ceremony for the new high-performance computer at the Potsdam Institute for Climate Impact Research is scheduled for early next year.

The new system is intended to support researchers in their work to gain a deeper understanding of the effects of climate change and subsequently develop strategies for the sustainable existence of mankind and nature. Particular focus will be placed on increasing efficiency and performance over the previous system. HPC systems with Lenovo’s Neptune Liquid Cooling technology will be used for this purpose.

The new supercomputer offers higher performance, enabling PIK researchers to perform faster and more comprehensive calculations. An important project of the research institute is the economic shock model Acclimate. This is used to analyze disruptions in over 1.8 million supply relationships affecting more than 7000 economic sectors worldwide. To do this, global supply lines must be recalculated daily to understand the impact of extreme weather events on these relationships. The new HPC system ensures that uncertainties can be estimated more precisely, and predictions improved.

Another area of application for the supercomputer is risk forecasting. For a statistically robust risk assessment, as many extreme events as possible must be considered. The high-performance computer is used to simulate flood events, for example, for possibly affected regions via a combination of climate and hydrological models. With the capacities now available, significantly more experiments can be conducted and thus risks can be predicted more accurately.

In addition, PIK will be able to use the full range of machine learning (ML) methods to solve fundamental problems in climate research. This will allow PIK to continue to be a world leader in contributing to core topics such as climate networks, hybrid or ML-based Earth system modeling, bias correction, and downscaling, and predictions of extreme events and abrupt transitions such as monsoon onsets and their associated socio-economic impacts.

To meet the technical requirements and support the researchers at PIK in the best possible way, the new HPC system at PIK will use the latest version of Lenovo’s ThinkSystem SD665 V3 servers with an integrated AMD EPYC 9004 processor. The combination of 128 cores with a high base clock of 3.1 GHz as well as a bandwidth of 24 memory channels with built-in DDR5 memory was decisive for the decision for this server. The overall system thus has a total number of over 30,000 cores, 240 nodes, and 185 TB of RAM. The GPU used will be NVIDIA’s Hopper H100 in SXM form factor, specifically designed for HPC and AI applications. In addition, for the AI requirements of the researchers at PIK, the GPU partitions are important as part of the overall solution.