HPE dominiert bei Supercomputern – HPE Dominates the Supercomputing Market

Unter den Top10 sind auf Platz 5 und 6 zwei weitere HPE Supercomputer zu finden, und das sogar in Europa. Der für das EuroHPC JU gebaute und vom LUMI-Konsortium betriebene Supercomputer LUMI am CSC in Finnland belegte Platz 5. Die Forscher nutzen LUMI zur Modellierung des Klimawandels und zur Früherkennung von Krebs, indem sie ein KI-Modell auf Millionen von Proben trainieren, um präzisere Diagnosen zu erhalten.

Der Supercomputer Alps, der für das Nationale Zentrum für Hochleistungsrechnen der Schweiz (CSCS) gebaut wurde, schaffte es zum ersten Mal in die Top 10 und erreichte Platz 6. Alps basiert auf dem HPE Cray EX Supercomputer mit NVIDIA GH200 Grace Hopper Superchips und erreicht eine Leistung von 270 Petaflops. Forscher aus einer Vielzahl von Disziplinen werden von der herausragenden Leistung von Alps und seinen beispiellosen KI-Fähigkeiten profitieren. Klima und Wetter, Astrophysik, numerische Strömungsmechanik, Biowissenschaften, Quantenchemie und Teilchenphysik werden alle von der innovativen Architektur von Alps profitieren.

Ein Exascale-Computer schafft eine Trillion Rechenoperationen pro Sekunde. Die Grundlage von Aurora ist das System HPE Cray EX, das speziell für das Exascale-Computing entwickelt wurde. Aurora ist außerdem die größte Installation von HPE Slingshot – einem offenen Ethernet-Supercomputer-Netzwerk – in einem einzigen System. Slingshot verbindet in Aurora 75.000 Rechnerknoten sowie 2.400 Speicher- und Service-Netzwerk-Knoten mit 5.600 Switches. Dieses Hochgeschwindigkeitsnetz steigert die Leistung der 10.624 Rechner-Blades von Aurora, die mit 21.248 Intel-Xeon-CPUs der Max-Serie und 63.744 Intel-GPUs der Max-Serie ausgestattet sind. Aurora ist damit eines der weltweit größten GPU-Cluster.

Aurora wurde von Anfang an als KI-fähiges System geplant und wird unter anderem für die Forschung zu generativer KI eingesetzt werden. Weitere KI-gestützte Forschungsprojekte sind beispielsweise die Kartierung der 80 Milliarden Neuronen des menschlichen Gehirns, die Verbesserung der Hochenergie-Teilchenphysik mit Deep Learning und die Entwicklung von Medikamenten mithilfe von maschinellem Lernen.

“We are honored to celebrate another significant milestone in exascale with Aurora, which delivers massive compute capabilities to make breakthrough scientific discoveries and help solve the world’s toughest problems,” said Trish Damkroger, senior vice president and general manager, HPC & AI Infrastructure Solutions at HPE. “We are proud of the strong partnership with the U.S. Department of Energy, Argonne National Laboratory, and Intel to realize a system of this scale and magnitude that was made possible through our joint innovative engineering, multiple teams, and most importantly, shared value of delivering state-of-the-art technology to fuel science and benefit humankind.”

An exascale computing system can process one quintillion operations per second. Computational power at this scale makes it possible to address some of humanity’s most complex problems. Aurora is built with the HPE Cray EX supercomputer, which is purpose-built to support the magnitude and scale of exascale. The system is also the largest deployment of open, Ethernet-based supercomputing interconnect – HPE Slingshot – on a single system. The fabric, connects Aurora’s 75,000 compute node endpoints and 2,400 storage and service network endpoints along with 5,600 switches to boost performance and achieve AI at scale by enabling high-speed networking across Aurora’s 10,624 compute blades, 21,248 Intel® Xeon® CPU Max Series processors and 63,744 Intel® Data Center GPU Max units, making it one of the world’s largest GPU clusters.

Planned as an AI-capable system from inception, researchers will be able to use generative AI models on Aurora to accelerate scientific discovery. Early AI-driven research that scientists have run on Aurora include brain mapping to better understand the human brain’s 80 billion neurons, high energy particle physics enhanced by deep learning, and machine-learning accelerated drug design and discovery, among others.

An exascale computing system can process one quintillion operations per second. Computational power at this scale makes it possible to address some of humanity’s most complex problems.

“Aurora is a first-of-its-kind supercomputer and we expect it to be a gamechanger for researchers,” said Rick Stevens, associate laboratory director and distinguished fellow at Argonne National Laboratory. “Reaching this milestone with a second exascale system in the U.S. is an incredibly significant achievement that will advance open science initiatives globally.”

The Aurora exascale supercomputer is the result of a strong private-public partnership between HPE, Intel, the U.S. Department of Energy, and Argonne National Laboratory, which requires co-investment and co-development to create the breakthrough engineering that is necessary to advancing science. Partnerships between the private and public sectors are key to achieving scientific progress as demonstrated by work done through the Aurora Early Science Program. As part of the process of optimizing and stress-testing the system, researchers have already successfully run a diverse range of programming models, languages and applications on the system.

“The Aurora supercomputer was designed to support the research and science communities within the HPC and AI space,” said Ogi Brkic, Intel vice president and general manager, Data Center AI Solutions. “Our ongoing collaboration with Argonne National Laboratory and HPE has resulted in promising early science success stories. And we’re excited to see what’s to come as we continue to optimize system performance to accelerate the science and march toward what is next.”

Aurora has achieved exascale on a partial run of the system, tapping 9,234 of the total nodes. Aurora is an open science system housed at the Argonne Leadership Computing Facility (ALCF), a part of the U.S. Department of Energy’s Office of Science user facility.