Lawrence Livermore National Laboratory’s (LLNL) next-generation supercomputer, Sierra, is a third-fastest computing complement in a world, according to a TOP500 list announced now during a International Supercomputing Conference in Frankfurt, Germany.
Expected to be entirely operational this fall, Sierra will support a National Nuclear Security Administration’s (NNSA) core goal of ensuring a safety, confidence and efficacy of a nation’s chief stockpile. It is approaching to go into prolongation for personal work in 2019 with capabilities in appurtenance training and high-resolution displaying and simulation, and will perform formidable multi-physics calculations during speeds distant over any other supercomputer deployed by NNSA, a semi-autonomous group within a Department of Energy.
“NNSA and a predecessors have led a republic and a universe in growth of supercomputing record given World War II,” pronounced NNSA’s Advanced Simulation Computing Program Director Mark Anderson. “Sierra is a latest instantiation in this unapproachable tradition and is optimized to yield limit potency for a NNSA mission.”
Running a standardised benchmark exam employed by a TOP500 classification on a apportionment of a system, Sierra clocked in during scarcely 72 petaFLOPS (quadrillion floating-point operations per second) in limit performance, fixation it behind China’s Sunway TaihuLight, a complement grown by China’s National Research Center of Parallel Computer Engineering Technology (NRCPC) though forward of a subsequent fastest Chinese supercomputer, a Tianhe-2 (MilkyWay-2). Sierra is approaching to tip out during 125 petaFLOPS of rise performance, about 6 to 10 times some-more absolute than LLNL’s stream flagship complement Sequoia.
“Sierra will be indispensable to accommodate a perfectionist mandate of a Life Extension Programs (LEPs) and other goal deliverables into a 2020s,” pronounced LLNL’s Program Director for Advanced Simulation Computing Mike McCoy. “Our basic work in modernized architectures on Sierra is already temperament fruit by softened make-believe capability.”
Sierra’s TOP500 outcome means a U.S. now has dual of a 3 fastest computers in a world, with a tip container assigned by Oak Ridge National Laboratory’s Summit system, that available 122 petaFLOPS on a benchmark test. Summit and Sierra are both a deliverables of CORAL (the Collaboration of Oak Ridge, Argonne and Livermore) partnership, and were done probable by a public-private partnership with vendors IBM, NVIDIA and Mellanox. Red Hat is provision Sierra’s open source handling system.
Like Summit, Sierra is built by IBM and is a extrinsic supercomputer, comprised of a sum of 8,640 IBM Power9 executive estimate units (CPUs) and 17,280 NVIDIA Tesla V100 Tensor Core graphics estimate units (GPUs), that could be used to try cognitive make-believe capabilities for save stewardship applications. It is a initial complement of a kind constructed for NNSA.
“Summit and Sierra are really identical systems,” pronounced Livermore Computing’s Chief Technical Officer Bronis de Supinski. “However, by usually deploying 4 GPUs per node, Sierra offers aloft CPU to GPU bandwidth per GPU, that we have found is vicious to grasp opening for the applications.”
“The Sierra appurtenance goes serve than any prior appurtenance to capacitate the save stewardship goal with clever opening opposite a far-reaching operation of applications,” combined Adam Bertsch, Livermore Computing’s formation manager for Sierra.
Comment this news or article