Stuttgart, December 7, 2023 – Immediately, Lenovo introduced the growth of the Excessive-Efficiency Computing (HPC) system “LISE” on the Zuse Institute Berlin (ZIB) – an interdisciplinary analysis establishment for utilized arithmetic and data-intensive HPC. This growth offers researchers with the required energy to execute compute-intensive functions equivalent to synthetic intelligence and machine studying. The main focus of the growth has been on bettering vitality effectivity, using revolutionary applied sciences, together with Lenovo Neptune liquid cooling, and guaranteeing seamless integration into the present system by HPC companion pro-com.
Know-how and Sustainability
Relying on workload, inlet temperature, and ambient temperature, the system’s effectivity is within the vary of 85-90%. By using the Lenovo Excessive Warmth Restoration answer with water-cooled energy provides and insulated, water-cooled rear doorways of the HPC racks, effectivity is additional enhanced, with over 97% of the warmth directed into the new water circuit as an alternative of the information heart. This enhances the information heart’s vitality steadiness, eliminating the necessity for added room cooling. A feasibility examine is at the moment underway to discover the reuse of waste warmth for the district heating community or native use at ZIB.
“For the Zuse Institute Berlin, the energy-efficient operation of its highly effective HPC sources has been of explicit significance for years, standing on equal footing with excessive computing energy and environment friendly operation,” added Thomas Steinke, Head of the Supercomputing Division at ZIB. “By way of sustainability and vitality price improvement, we have now chosen technological options from Lenovo that enable us to fulfill our necessities for each robust computing energy by the most recent processor generations and excessive vitality effectivity of CPU and GPU platforms as a consequence of direct liquid cooling. We’re happy to have discovered robust companions in Lenovo and pro-com DATENSYSTEME GmbH for the implementation of our plans.”
The LISE system was commissioned on the finish of 2019 and expanded this 12 months with GPU partitions from NVIDIA and Intel as a consequence of elevated demand for GPU sources. The NVIDIA GPU partition consists of 42 computing nodes, every containing 4 NVIDIA A100/80 GPUs. The Intel GPU partition, alternatively, contains 8 computing nodes, every geared up with 4 Intel MAX GPUs (previously generally known as “Ponte Vecchio”). Each partitions make the most of Lenovo {hardware}. The NVIDIA GPU partition has been accessible to customers because the finish of March, whereas the Intel partition is at the moment being ready for operation. Each GPU partitions are deliberate to be utilized for 4 to 5 years and will likely be expanded with new heterogeneous processor architectures from 2025. Moreover, the set up of an energy-efficient CPU partition with 168 nodes is scheduled for the top of 2023/early 2024, using the Lenovo ThinkSystem SD665 V3 with AMD EPYC processors of the 4th technology and built-in Neptune liquid cooling.
“We’re happy that the Zuse Institute Berlin, along with pro-com DATENSYSTEME GmbH, has chosen Lenovo for the growth of their HPC cluster ‘LISE’ to additional advance analysis in Germany,” commented Andreas Thomasch, Director HPC & AI DACH, France, UKI at Lenovo. “This alternative underscores our dedication to technological progress by the warm-water-cooled deployment of the most recent Intel GPUs and NVIDIA GPUs, in addition to AMD and Intel CPUs, mixed with our dedication to sustainable and environment friendly options within the discipline of Excessive-Efficiency Computing. By way of our revolutionary Neptune liquid cooling, we improve the computing energy of the cluster for even sooner insights for scientists whereas concurrently decreasing environmental affect.”
Purposes
The LISE system at ZIB helps numerous functions from varied scientific disciplines, significantly in life sciences, chemistry and supplies science, earth system science, and engineering (CFD). GPU sources play a central position in tasks using Synthetic Intelligence (AI) and Machine Studying (ML) strategies in simulations and technique improvement. Two examples that mix conventional simulation strategies with AI approaches are drug improvement and the examine of the dynamics of protein constructions. The primary venture includes an in-silico seek for new opioids with out dangerous unintended effects, coupling complicated atomistic simulations with AI strategies to successfully determine candidates and their mechanisms of motion. In one other venture in collaboration with the Free College of Berlin, new atomistic-based protein fashions are refined utilizing AI approaches to simulate essential structure-dependent protein features that develop over an prolonged timescale.
Funding and Community
ZIB is one in every of 9 NHR (Nationwide Excessive-Efficiency Computing) facilities within the Nationwide Excessive-Efficiency Computing Consortium, aiming to additional increase high-performance computing capacities for analysis and improvement in Germany. The collaboration and networking of HPC facilities within the NHR consortium enhance effectivity and entry to highly effective computing infrastructure for scientists and researchers all through Germany. The HPC Cluster LISE performs a big position on this community, serving to set up Germany as an innovation power within the discipline of high-performance computing. LISE is now being expanded as a part of this program with monetary assist from federal and state governments.
