Nvidia’s Grace superchip made waves when launched earlier this yr, as the corporate promised a supercharged Arm-based product that might tackle Intel and AMD’s x86 dominance within the Excessive-Efficiency Computing (HPC) house. Now, as reported by HPC Wire, the corporate has snagged a $160 million contract (~€151 million) to offer the brains and brawn of supercomputing {hardware} for one in every of EuroHPC’s supercomputing initiatives. The MareNostrum 5 (MareNostrum roughly interprets to “our sea”) shall be put in within the Barcelona Supercomputing Centre (BSC) in Spain and shall be operational as early as 2023.
Mare Nostrum 5 is being constructed as a part of the EuroHPC JU mission, and is predicted to supply peak efficiency of 314 Petaflops of FP 64 computing energy throughout each CPU and GPU accelerators, with 200 Petabytes of storage for in-access workloads, and an additional 400 Petabytes of chilly storage. Following traits in HPC structure design and different initiatives throughout the EuroHPC mission, it is anticipated that the 200 Petabyte node shall be stored in a quick, NAND-based storage subsystem, whereas the chilly storage node (additionally referred to as energetic storage, referring to information that is essential however not incessantly accessed) will seemingly make use of less expensive, classical HDD topologies.
The system will make use of Nvidia’s 144-core, Arm-based Grace “superchips” in dual-chip configurations, paired with the corporate’s H100 (Hopper) discrete GPU accelerators (which function 80 billion transistors apiece with 80 GB of HBM3 reminiscence and three.2 TB/s bandwidth). In consequence, MareNostrum 5 is projected to ship greater than 18 Exaflops of AI acceleration (sometimes FP8 8-bit floating-point operations), making it the quickest AI supercomputer within the European Union. Apart from Nvidia’s chip tech, the corporate’s Quantum-2 (aka NDR) InfiniBand software-defined networking (opens in new tab) leveraging the corporate’s BlueField information processing items (DPU) to maintain all parts speaking at low latency with a excessive throughput of 400 GB/s – not in contrast to the efficiency achieved by Cray’s Slingshot interconnect.
Educated hypothesis from The Subsequent Platform estimates that MareNostrum 5 might deploy as many as 4,500 “Hopper” H100 accelerators, which might even be good for round 270 petaflops of FP64 oomph due to the chip’s Tensor Cores. The remaining 44 Petaflops of FP64 efficiency are anticipated to be derived from the dual-Grace CPU techniques, which the publication estimates as offering 3.84 teraflops per Grace chip – amounting to a possible complete of round 5,730 twin Grace modules.
MareNostrum 5 is very focused for medical analysis, chemistry simulations and drug growth whereas supporting purposes like local weather science and environmental engineering. Nvidia’s Omniverse software program bundle (opens in new tab) will energy the event of digital twins for these purposes – primarily enabling large-scale, physically-accurate simulations of industrial-scale property and processes. The digital twin tech permits the combination of autonomous techniques with real-world, real-time information streams, enabling a round suggestions system of simulation, output, and on-the-fly updates on the simulated fashions.
“The acquisition of MareNostrum 5 will allow world-changing scientific breakthroughs such because the creation of digital twins to assist clear up world challenges like local weather change and the development of precision medication,” stated Mateo Valero, director of BSC. “As well as, [BSC] is dedicated to growing European {hardware} for use in future generations of supercomputers and serving to to attain technological sovereignty for the EU’s member states.”
Moreover, as is the case with this newest technology of supercomputers being put in throughout Europe, MareNostrum 5 shall be fully powered by renewable energies, with extra warmth being repurposed – as a substitute of simply being expelled with no vacation spot.
It is at the moment unclear how precisely the surplus warmth shall be repurposed (and in what quantity). It’s going to seemingly comply with the identical design ideas as Europe’s present chief within the supercomputing house, the Kajaani, Finland-installed, all-AMD-powered LUMI supercomputer, which repurposes 20% of its waste warmth in the direction of the encircling civil district, permitting for cuts in additional vitality expenditure in the direction of heating.
