Tino EberlApr 4, 2025

In #Jülich startet der #Supercomputer Jupiter, Europas erster #Exascale-Rechner mit über einem #ExaFLOP/s, in den Testbetrieb.

Das modular aufgebaute #Rechenzentrum in Containern nutzt rund 24.000 #Nvidia GH200-Chips für Simulationen und #KI-Training.

Ziel ist nicht nur Leistung, sondern auch #Energieeffizienz mit #Wärmerückgewinnung. Ab 2027 ergänzt das KI-Modul #Jarvis das System.

https://www.heise.de/news/Juelicher-Supercomputer-Jupiter-startet-Probebetrieb-10340419.html

#ForschungszentrumJülich #HPC #GreenIT #Containerrechenzentrum #Green500 #EuroHPC

Jülicher Supercomputer Jupiter startet Probebetrieb

Es kommt Leben in das Jülicher Containerdorf, in dem bald der Supercomputer Jupiter in Betrieb gehen soll – als Erster in Europa mit mehr als einem ExaFLOP/s.

heise online
Benjamin Carr, Ph.D. 👨🏻‍💻🧬Jan 30, 2025
#Aurora #supercomputer is now fully operational, available to researchers
Scientists use Aurora to drive breakthroughs in cancer, materials discovery, energy technologies and many other fields.
Aurora is one of first #exascale supercomputers, along with Frontier at #DOE’s Oak Ridge National Laboratory and El Capitan at Lawrence Livermore National Laboratory. Exascale refers to systems performing at least an #exaflop — a quintillion (a billion billion) calculations/s.
https://www.tomshardware.com/tech-industry/supercomputers/aurora-supercomputer-is-now-fully-operational-available-to-researchers
Aurora supercomputer is now fully operational, available to researchers

Another ExaFLOPS supercomputer available for researchers.

Tom's Hardware
Stefano PiccoSep 5, 2024
396 Exaflops! ... den Begriff #Exaflop kannte ich bisher noch nicht bzw. noch nirgends aufgeschnappt ... https://www.heise.de/news/xAI-hat-offenbar-den-schnellsten-Supercomputer-der-Welt-fertiggestellt-9857511.html #colossus
Colossus ist fertig: Womöglich schnellster Supercomputer der Welt

Der Supercomputer Colossus verwendet 100.000 H100-Beschleuniger von Nvidia. Die Rechenleistung ist enorm, aber auch der Stromverbrauch.

heise online
Matt WillemsenDec 19, 2023
World's First Human 'Brain-Scale' Supercomputer Will Go Online in 2024
https://www.sciencealert.com/worlds-first-human-brain-scale-supercomputer-will-go-online-in-2024 #supercomputer #BrainScale #online #DeepSouth #exaflop
World's First Human 'Brain-Scale' Supercomputer Will Go Online in 2024

Our brains are remarkably energy efficient.

ScienceAlert
Benjamin Carr, Ph.D. 👨🏻‍💻🧬Dec 3, 2023
#China Publishes Its Own #Top100 #Supercomputer List
China's lack of #Top500 submissions underscores separation of Chinese #supercomputing from global #HPC. But even China's own list does not include country's fastest #ExaFLOP computers; upper echelons of China's supercomputing remains shrouded in secrecy. 21 out of Top 50 were deployed in last 2yr, indicating China's supercomputesr continue to develop rather rapidly despite US efforts to contain the country's advances.
https://www.tomshardware.com/news/china-supercomputing-top-100-list-us-vendors
China Publishes Its Own Top 100 Supercomputer List: No Exaflops Machines Listed

Chinese supercomputing sector further separates itself from the world.

Tom's Hardware
Erik JonkerNov 9, 2023

Interesting article to read about the infrastructure used for #AI ,
"It leverages 10,752 GPUs strung together using NVIDIA's Infiniband networking (moving a petabyte of data a second) and 860 terabytes of high bandwidth memory (36PB/sec aggregate bandwidth and 1.1PB sec interconnected) to deliver 40 exaflops of AI processing power. " ...the largest supercomputer in the Netherlands is in petaflops..🤔
https://www.engadget.com/nvidias-eos-supercomputer-just-broke-its-own-ai-training-benchmark-record-170042546.html

#AI #nvidia #infrastructure #supercomputer #exaflop

Engadget is part of the Yahoo family of brands

Benjamin Carr, Ph.D. 👨🏻‍💻🧬Nov 4, 2023
21 #ExaFLOP #Isambard #AI #Supercomputer Uses 5,448 #NVidia #GH200 #GraceHopper Superchips
U.K.'s most powerful supercomputer boasts 200 #PetaFLOP #FP64 compute.
The machine, which is set to be built by #HPE, cost £225 million ($273.38 million)
https://www.tomshardware.com/pc-components/gpus/21-exaflop-isambard-ai-supercomputer-uses-5448-gh200-grace-hopper-superchips
21 ExaFLOP Isambard-AI Supercomputer Uses 5,448 GH200 Grace Hopper Superchips

U.K.'s most powerful supercomputer boasts 200 PetaFLOPS FP64 compute.

Tom's Hardware
Benjamin Carr, Ph.D. 👨🏻‍💻🧬Jun 13, 2023
#AMD #MI300 – Taming Hype – #AI Performance, Volume Ramp, Customers, Cost, IO, Networking, Software
Amazing engineering, but what of the path to market?
AMD delivered #HPC #GPU silicon in 2021 for world's first #ExaFLop #supercomputer, Frontier. Now, everyone is looking forward MI300 which will be delivered later this year to El Capitan, 2nd Exascale supercomputer win. There hasn’t been this much buzz for a #datacenter chip since #Nvidia’s #Volta GPU and AMD’s Rome CPU.
https://www.semianalysis.com/p/amd-mi300-taming-the-hype-ai-performance
AMD MI300 – Taming The Hype – AI Performance, Volume Ramp, Customers, Cost, IO, Networking, Software

Amazing engineering, but what of the path to market?

SemiAnalysis
Tero Keski-ValkamaMay 30, 2023

#NVIDIA Announces #DGX GH200 #AI #Supercomputer | NVIDIA Newsroom

"New Class of AI Supercomputer Connects 256 #GraceHopper Superchips Into Massive, 1-#Exaflop, 144TB #GPU for Giant Models Powering #GenerativeAI, #RecommenderSystems, #DataProcessing"

https://nvidianews.nvidia.com/news/nvidia-announces-dgx-gh200-ai-supercomputer

NVIDIA Announces DGX GH200 AI Supercomputer

NVIDIA today announced a new class of large-memory AI supercomputer — an NVIDIA DGX™ supercomputer powered by NVIDIA® GH200 Grace Hopper Superchips and the NVIDIA NVLink® Switch System — created to enable the development of giant, next-generation models for generative AI language applications, recommender systems and data analytics workloads.

NVIDIA Newsroom
Show threadTero Keski-ValkamaNov 14, 2022
"#Cerebras said #Andromeda can perform 1 #exaflop worth of #AI #computing - or at least one quintillion (10 to the power of 18) operations per second - based on a precision of 16 bit floating point format."