Friday, February 10th 2023
Atos to Build Max Planck Society's new BullSequana XH3000-based Supercomputer, Powered by AMD MI300 APU
Atos today announces a contract to build and install a new high-performance computer for the Max Planck Society, a world-leading science and technology research organization. The new system will be based on Atos' latest BullSequana XH3000 platform, which is powered by AMD EPYC CPUs and Instinct accelerators. In its final configuration, the application performance will be three times higher than the current "Cobra" system, which is also based on Atos technologies.
The new supercomputer, with a total order value of over 20 million euros, will be operated by the Max Planck Computing and Data Facility (MPCDF) in Garching near Munich and will provide high-performance computing (HPC) capacity for many institutes of the Max Planck Society. Particularly demanding scientific projects, such as those in astrophysics, life science research, materials research, plasma physics, and AI will benefit from the high-performance capabilities of the new system.
The system will run without fans thanks to its unmatched hot water cooling (Direct Liquid Cooling) and thus have a significantly improved energy efficiency. The efficiency gained from this Atos' cooling system matched with the latest in silicon architecture innovations from AMD, designed with energy efficiency in mind, brings the Power Usage Effectiveness (PUE) value to less than 1.05 (1 being the ideal ratio), far below the average of other HPC installations.
The installation will feature 4th Gen AMD EPYC processors and for the first time in an Atos-based European system, the upcoming AMD Instinct MI300A accelerator. The system will consist of ten BullSequana XH3000 racks with a total of 768 processor nodes of and 192 accelerator nodes and will be complemented with an IBM SpectrumScale storage solution. The CPU nodes will be delivered in the third quarter of 2023, with full installation of the GPU nodes expected in the first half of 2024.
Prof. Erwin Laure, Director of the Max Planck Computing and Data Facility said "The computing power required by scientific research is ever increasing and we see an unabated need for high-performance computing capacity. We want to provide the best possible support to our researchers in their work and have therefore decided to modernize our high-performance computing complex. With Atos and AMD, we have the right partners for this. The new solution will certainly meet our demands and once again advance science in leaps and bounds."
Emmanuel Le Roux, Group SVP, Global Head of HPC, AI & Quantum at Atos, highlighted, "We are very proud to have been awarded this contract and to have been trusted with our extensive HPC expertise. We are convinced that the combination of our newly developed, powerful and energy-efficient BullSequana XH3000 system with the high-performance AMD Instinct MI300A processors will provide the Max Planck Society with a future-proof system and the computing power required to further continue its ground-breaking research."
Brad McCredie, corporate vice president, Data Center Acceleration Business, AMD said "AMD, with the MI300A APU, continues to deliver breakthrough performance for researchers and the high performance computing industry. AMD is excited by the strong market validation of the performance, power and simplified programming advantages unlocked by the MI300A APU architecture and we are looking forward to working with Atos to advance the specific science and research needs of the Max Planck Society."
The new supercomputer, with a total order value of over 20 million euros, will be operated by the Max Planck Computing and Data Facility (MPCDF) in Garching near Munich and will provide high-performance computing (HPC) capacity for many institutes of the Max Planck Society. Particularly demanding scientific projects, such as those in astrophysics, life science research, materials research, plasma physics, and AI will benefit from the high-performance capabilities of the new system.
The installation will feature 4th Gen AMD EPYC processors and for the first time in an Atos-based European system, the upcoming AMD Instinct MI300A accelerator. The system will consist of ten BullSequana XH3000 racks with a total of 768 processor nodes of and 192 accelerator nodes and will be complemented with an IBM SpectrumScale storage solution. The CPU nodes will be delivered in the third quarter of 2023, with full installation of the GPU nodes expected in the first half of 2024.
Prof. Erwin Laure, Director of the Max Planck Computing and Data Facility said "The computing power required by scientific research is ever increasing and we see an unabated need for high-performance computing capacity. We want to provide the best possible support to our researchers in their work and have therefore decided to modernize our high-performance computing complex. With Atos and AMD, we have the right partners for this. The new solution will certainly meet our demands and once again advance science in leaps and bounds."
Emmanuel Le Roux, Group SVP, Global Head of HPC, AI & Quantum at Atos, highlighted, "We are very proud to have been awarded this contract and to have been trusted with our extensive HPC expertise. We are convinced that the combination of our newly developed, powerful and energy-efficient BullSequana XH3000 system with the high-performance AMD Instinct MI300A processors will provide the Max Planck Society with a future-proof system and the computing power required to further continue its ground-breaking research."
Brad McCredie, corporate vice president, Data Center Acceleration Business, AMD said "AMD, with the MI300A APU, continues to deliver breakthrough performance for researchers and the high performance computing industry. AMD is excited by the strong market validation of the performance, power and simplified programming advantages unlocked by the MI300A APU architecture and we are looking forward to working with Atos to advance the specific science and research needs of the Max Planck Society."
19 Comments on Atos to Build Max Planck Society's new BullSequana XH3000-based Supercomputer, Powered by AMD MI300 APU
Max-Planck-Gesellschaft MPI/IPP
Germany
So yes.
*Today price for exaflop supercomputer like Frontier.
I've heard of all-CPU designs, and all-GPU designs. But this is a weird hybrid between the two.
Its just weird to me to have.... what is effectively two different supercomputers hooked up and networked together. Normally, a supercomputer has the "same node" design across the entire supercomputer, so that there is less to study / easier to understand to code against.
There's a 768-pure CPU cluster + the 192-node CPU+GPU (probably MI300) cluster. A strange design and a headache to load balance against.
CUDA is nice but too closed imho, at least OpenCL can run on CPU too ... :D
NVidia is pushing CUDA. AMD is pushing ROCm / HIP specifically. Intel is pushing OneAPI / SyCL.
Supercomputers have a lot of Fortran, and OpenMP code. There are efforts to port this code over to GPUs, and that's my overall expectation for how things would work in the Supercomputer world. If specific AMD code needs to be written, I'd expect ROCm to be leveraged on this particular supercomputer. (Similarly, CUDA would be used on NVidia-specific supercomputers if portability doesn't matter to that particular team). But I'm not in the supercomputer field at all actually, just passing some rumors / hearsay.
(CPU are OpenCL1 1.2 at max, iirc, which explain what you mention)
anyhow, as long as they have API to run the things ... be it OpenCL or anything else ... and the performances are there : yep, all is good.Crysis is the usual joke in that case ... :D