Three of the latest Nationwide Nuclear Safety Administration (NNSA) commodity computing clusters not too long ago deployed at Lawrence Livermore Nationwide Laboratory (LLNL) and Sandia Nationwide Laboratories (SNL) are among the many strongest supercomputing programs on this planet, Top500 organizers introduced Monday.
Funded underneath the second Commodity Know-how Methods contract (CTS-2) by NNSA’s Superior Simulation and Computing (ASC) program, the machines sited at LLNL — named Dane and Bengal — started performing key modeling and simulation capabilities for the NNSA’s stockpile stewardship mission in mid-2023. Every system is constructed by Dell Applied sciences and powered by 4th Gen Intel Xeon Scalable Processors coupled to Cornelis Networks Omni-Path high-speed networking.
The debut of Dane, a 7.041 petaFLOP (barely greater than 7 quadrillion calculations per second) cluster at #108 and the 6.134 petaFLOP Bengal at #129, brings the entire of LLNL-sited programs on the Top500 Record to 11, essentially the most of any supercomputing heart on this planet. A 3rd new CTS-2 system, sited at SNL and named Stout, reached 8.987 petaFLOPs on the LINPACK benchmark used to find out the rankings, incomes it 87th place on the record. Every system attained 89-percent or higher effectivity on LINPACK. Top500 organizers unveiled the up to date bi-annual record of the world’s strongest computer systems on the 2023 Worldwide Convention for Excessive Efficiency Computing, Networking, Storage and Evaluation (Supercomputing) in Denver.
Since their deployment for mission-critical work, Dane, Bengal, Stout and a fourth CTS-2 system named Amber at SNL, have displayed improved efficiency and effectivity over the earlier technology of commodity programs (CTS-1), with NNSA researchers reporting preliminary speedups of 4-5X throughout a variety of high-performance computing purposes.
“The deployment of those first CTS-2 platforms gives a big addition to the NNSA labs’ simulation environments in assist of our nationwide safety mission,” mentioned Matt Leininger, CTS venture lead for the NNSA Tri-Labs (LLNL, SNL and Los Alamos Nationwide Laboratory). “We’re proud to proceed the NNSA ASC custom of excellence in high-performance computing and U.S. know-how partnerships at LLNL, SNL and all through the NNSA advanced.”
The commodity-technology-based programs are the “workhorses” of the NNSA and assist the day-to-day simulation workload whereas lowering prices by standardizing {hardware} and software program throughout the NNSA labs, when coupled with the LLNL-led Tri-Laboratory Working System Stack (TOSS) and Tri-Lab Widespread Atmosphere. The CTS-2 platforms permit NNSA’s extra highly effective Superior Know-how System supercomputers, similar to the present Sierra system and future El Capitan exascale system at LLNL, to deal with essentially the most advanced issues important to NNSA’s Stockpile Stewardship Program.
Every CTS-2 system is made from constructing blocks known as “scalable items” (SUs) representing about 1.5 petaFLOPs of computing energy apiece. The SUs design permits the NNSA laboratories to fine-tune system efficiency relying on mission want or programmatic budgets. Every SU is constructed with Dell PowerEdge C6620 and R760 servers that make the most of 4th Gen Intel Xeon Scalable processors.
The 4th Technology Intel Xeon Scalable processors are outfitted with purpose-built accelerators that may elevate HPC workload efficiency and energy effectivity by offloading duties to those acceleration options. Intel Superior Matrix Extensions (Intel AMX), one of many processors’ built-in accelerator engines, can remodel the big matrix math calculations which can be on the coronary heart of deep-learning workloads right into a single operation, making them important for delivering efficiency throughout workloads the place HPC and AI converge.
“These processors are perfect for constructing and deploying general-purpose AI workloads with the preferred AI frameworks and libraries. These capabilities will make it potential for the engineers, researchers and scientists at Lawrence Livermore Nationwide Laboratory and Sandia Nationwide Laboratory to simulate advanced issues important to the nation’s nationwide safety,” mentioned Deepak Patil, company vp and basic supervisor, Accelerated Computing Methods & Graphics, Intel. “We’re proud to assist the labs with options that assist speed up the effort and time to research future and current information for his or her focus areas.”
Every system can also be outfitted with 200Gbps Omni-Path Categorical high-speed networking from Cornelis Networks, offering CTS-2 with a extremely scalable cloth with open-source software program totally built-in with TOSS.
“With CTS-2, the workforce at Cornelis is worked up to as soon as once more assist supply of the workhorse commodity programs of the NNSA, having beforehand delivered the high-performance networks for the second technology of Tri-Lab Capability Clusters (TLCC2) and CTS-1,” mentioned Gunnar Ok. Gunnarsson, vp of options supply and assist at Cornelis. “Working carefully with our ecosystem companions, Dell Applied sciences and Intel, Cornelis is happy to allow main cluster efficiency and effectivity at scale for the Tri-Labs’ mission-critical workloads with our Omni-Path know-how. We look ahead to continued CTS-2 deployments, together with introducing our new 400Gbps CN5000 Omni-Path product household as a part of a future structure refresh.”
A number of extra CTS-2 programs will probably be deployed at LLNL and SNL within the first half of 2024, in assist of varied NNSA applications leveraging the CTS structure and procurement developed by the NNSA ASC program.
