ESiWACE3’s Post

#Supercomputing is our thing - that's why we're also linking to the latest #Quartl here: The newsletter is a joint product of the competence network for technical-scientific high-performance computing in Bavaria, #KONWIHR, and TUM and compares the Grace Superchip from NVIDIA with the Xeon Platinum from Intel Corporation in the latest issue. Also in the magazine: a recap of the HPC Days in Durham and research reports with #HPC. Worth more than just a look: https://lnkd.in/dhimtWSa (in German and English) #supercomputing, #highperformancecomputing, #HPC4Germany, #IT4Science, #discoverLRZ, #CPU, #GPU, Gauss Centre for Supercomputing, EuroHPC Joint Undertaking (EuroHPC JU), ETP4HPC European Technology Platform for HPC

The #Technovision trend of this Friday is ‘OK QOMPUTE’. Yes with a 'Q' because Quantum computing is still there emerging and developing. Join our #ExpertInResidence Julian van Velzen in their explanation of this trend. New horizons of more organic computing are emerging, driven by Quantum and neuromorphic chips — breaking the spell of bits and bytes, opening up entirely unexplored opportunities. To address some of the most significant challenges of our time, such as improving healthcare and the race to net zero, there is a demand for much more computational power. With Moore’s law finally outpaced, the time has come to explore alternative computing methods. Welcome to the realm of bits, neurons, and qubits. We have already moved beyond the classic CPU processor architecture — with GPUs, TPUs, NPUs, LPUs. But their roles will be shaken up by the new kids on the block — Quantum and neuromorphic chips. They operate on fundamentally different principles, more closely modeling our own, organic reality. That promises extraordinary, quite OK capabilities. Scan the QR code or reach out to learn more.  #TechnoVision2024  AUGMENT ME! Read all about this trend and the 36 other trends & principles in: https://lnkd.in/gjNhdPcs Sjoukje Zaal Ron Tolido Bernd Wachter #Technovision #Architecture #LifeatCapgemini #TechnovisioncardFriday #TechnovisiontrendFriday #TechnovisionprincipleFriday

Scientists at Oak Ridge National Laboratory have just rocked the world of supercomputing by training a whopping one trillion parameter model using just a tiny fraction of their powerhouse Frontier supercomputer. Here’s the jaw-dropping bit: They achieved this feat using only 8% of Frontier's computing power. 3,072 GPUs out of a staggering 37,888 AMD GPUs were harnessed to train this colossal model. By utilizing techniques like shuffled data parallelism and tensor parallelism, these scientists sliced through barriers, optimizing communication and handling memory constraints like seasoned pros. Talk about efficiency redefined! Their performance results? A stunning 100% weak scaling efficiency for models of colossal sizes! It’s not every day you witness such extraordinary computational prowess in action. Have you read their research paper?

The #Technovision trend of this Friday is ‘OK Qompute’. Quantum, quantum, quantum. Not for all tasks the right solution, but be sure to understand for which tasks it is... Join our #ExpertInResidence Julian van Velzen in their explanation of this trend. New horizons of more organic computing are emerging, driven by Quantum and neuromorphic chips — breaking the spell of bits and bytes, opening up entirely unexplored opportunities. To address some of the most significant challenges of our time, such as improving healthcare and the race to net zero, there is a demand for much more computational power. With Moore’s law finally outpaced, the time has come to explore alternative computing methods. Welcome to the realm of bits, neurons, and qubits. We have already moved beyond the classic CPU processor architecture — with GPUs, TPUs, NPUs, LPUs. But their roles will be shaken up by the new kids on the block — Quantum and neuromorphic chips. They operate on fundamentally different principles, more closely modeling our own, organic reality. That promises extraordinary, quite OK capabilities. Scan the QR code or reach out to learn more.  #TechnoVision2024   AUGMENT ME! Read all about this trend and the 36 other trends & principles in: https://lnkd.in/gjNhdPcs Sjoukje Zaal Ron Tolido Bernd Wachter #Technovision #Architecture #LifeatCapgemini #TechnovisioncardFriday #TechnovisiontrendFriday #TechnovisionprincipleFriday 

✨ 𝗥𝗲𝗳𝗹𝗲𝗰𝘁𝗶𝗻𝗴 𝗼𝗻 𝘁𝗵𝗲 𝗠𝘂𝗻𝗶𝗰𝗵 𝗤𝘂𝗮𝗻𝘁𝘂𝗺 𝗦𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗙𝗼𝗿𝘂𝗺 (𝗠𝗤𝗦𝗙) 𝟮𝟬𝟮𝟰! Last week, Aqarios joined the 𝗠𝘂𝗻𝗶𝗰𝗵 𝗤𝘂𝗮𝗻𝘁𝘂𝗺 𝗦𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗙𝗼𝗿𝘂𝗺 (𝗠𝗤𝗦𝗙), where our CEO, Michael Lachner, introduced 𝗟𝘂𝗻𝗮 as a practical, end-user platform for quantum optimization. The forum brought together quantum software leaders, such as BMW Group presenting their 𝗤𝗨𝗔𝗥𝗞 benchmarking framework for evaluating quantum solutions on industrial tasks, NVIDIA with 𝗖𝗨𝗗𝗔-𝗤 to program QPUs, GPUs, and CPUs in one system, and Q-CTRL with 𝗙𝗶𝗿𝗲𝗢𝗽𝗮𝗹, a framework enhancing quantum algorithm performance, which is also available in our Luna platform. The event fostered a vibrant community atmosphere, featuring pitches, presentations, and a poster session with various innovative software companies. Luna, positioned as an end-user platform, resonated strongly within this ecosystem, highlighting synergies with other quantum software technologies and tools. Insightful discussions underscored the importance of accessible, practical solutions as the quantum industry moves from research to real-world applications. Thanks to Robert Wille and his team from Technical University of Munich, the European Research Council (ERC), and Munich Quantum Valley for organizing this impactful event and fostering collaboration across the quantum software community. #QuantumSoftware #MQSF #QuantumOptimization #QuantumComputing #Aqarios #Luna #Innovation

ChemicalQDevice research & development of Ni, B., et al. 2024 Science Advances notebook. A) The number of generations based on highest pull forces was increased to eight proteins. B) Perplexity decreased with pulling force, likely associated with less alpha helix structural motifs. C) GPU Cost efficiency study highlighted appropriateness of NVIDIA V100 High Ram GPU, shown above. The conclusion is on slide 06/08.

University of Innsbruck adds quantum computer with HPC cluster The University of Innsbruck and its spin-out company AQT have successfully integrated a quantum computer into a high-performance computing (HPC) environment for the first time in Austria. They connected the university's LEO5 cluster with AQT's IBEX Q1 quantum computer. The IBEX Q1, which fits into two custom racks, operates at room temperature with less than 2kW power consumption. Launched last year, LEO5 has 63 nodes, including 36 featuring Nvidia A30, A40, and A100 GPUs, providing 300 teraflops at FP64 and 740 teraflops at FP32. This integration is part of the FFG-funded HPQC project aimed at demonstrating hybrid quantum solutions. AQT's ion trap-based quantum computers, which use laser-cooled ions to create qubits, can be housed in standard 19-inch rack cabinets and have achieved a quantum volume of 128. #lureit #wannabelured #hardware #network #base #data #it #project #development #company #digital #power #launch #solution #platform #launch #campus #austria #university #quantum #computer #hpc

NVIDIA cuQuantum and PennyLane are accelerating #quantum simulations at groundbreaking speeds on the Perlmutter #supercomputer at the National Energy Research Scientific Computing Center (NERSC). Find out how this collaboration is achieving its remarkable results on #supercomputing simulations for the first time at this scale.

Along with neuromorphic computing, our group has been working on another kind of emerging computing paradigm, Ising computing/ Ising machines, for quite some time now. In this paper which just got published in IOP Nanotechnology journal, we show a significant advantage of oscillator Ising machines (OIM) over a conventional approximate algorithm (GW), meant to be run on CPUs or GPUs, to solve the same problem (MaxCut). We show that while time complexity for GW grows as square of problem size, that of OIM (including spintronic OIM) grows only as logarithm of problem size. We have also carried out the same comparative analysis between electronic ring oscillator Ising machines and GW and obtained similar trends (this will get published later this year in the proceedings of IEEE International Conference on Electronic Circuits and Systems (ICECS). https://lnkd.in/dzwEAX-4

Efficient computation of molecular and materials properties on heterogeneous HPC architectures with the Quantum ESPRESSO suite. Heterogeneous architectures have become a de facto standard in HPC, with the majority of the world's leading exascale and pre-exascale machines being equipped with GPUs. Material science is advancing alongside developments in computer science, with scientific codes being continually refined to exploit the enhanced capabilities of HPC. Quantum ESPRESSO (QE), a suite of Open-Source computer codes for electronic-structure calculations and materials modelling, allows to compute a broad range of chemical and physical properties of molecules and condensed matter systems. Enabling quantum materials discovery and design towards exascale and extreme scaling performance on present and future HPC machines is one of the pillars of the development of the QE suite. In the last years, an extensive work has been done to enable deployment of the suite on modern HPC supercomputers, allowing to compute challenging materials properties, at higher resolution, longer time scales, and greater accuracy. In this contribution the main scientific and technical challenges encountered in this work will be presented, illustrating with example applications the potentialities of HPC for realistic simulations. 2024 Congress of Società Italiana di Fisica #hpc #cpu #gpu #hybridcomputing #quantum #simulation #supercomputer https://lnkd.in/dCN6yfNR