This article was accepted into the corpus but its outbound wikilinks were never NER-processed — typical at the deepest BFS hop or when the run's entity cap was reached. No expansion funnel to show.
| Tianhe-2 | |
|---|---|
| Name | Tianhe-2 |
| Developer | National University of Defense Technology |
| Release | 2013 |
| Type | Supercomputer |
| Cpu | Intel Xeon E5-2692 |
| Accelerator | Intel Xeon Phi (initially), later models |
| Speed | 33.86 PFLOPS (peak LINPACK) |
| Location | National Supercomputer Center in Guangzhou |
| Os | Kylin (custom Linux), Ubuntu, CentOS |
Tianhe-2
Tianhe-2 was a Chinese supercomputer developed by the National University of Defense Technology and deployed at the National Supercomputer Center in Guangzhou; it achieved world-leading performance in the TOP500 list, surpassing peers such as Titan (supercomputer), Sequoia (supercomputer), K computer, Fugaku (supercomputer), and Summit (supercomputer). The system influenced research institutions including Lawrence Livermore National Laboratory, Oak Ridge National Laboratory, Argonne National Laboratory, RIKEN, and European Centre for Medium-Range Weather Forecasts by setting new targets for high-performance computing for tasks similar to those run on Blue Gene/Q, Cray (company), IBM, HPE (Hewlett Packard Enterprise), and Dell EMC systems. Industry partners such as Intel Corporation, NVIDIA, AMD, ARM Holdings, Huawei, and Inspur observed shifts in procurement and accelerator strategy after Tianhe-2 demonstrated large-scale deployment of manycore accelerators alongside x86 processors.
The architecture combined compute nodes based on Intel Xeon E5 series processors and manycore accelerators drawing on concepts from Intel Xeon Phi coprocessors, with interconnect designs influenced by research at Myricom, Mellanox Technologies, and InfiniBand. System topology used a hierarchical network reminiscent of designs studied at Sandia National Laboratories and Los Alamos National Laboratory for message-passing in MPI (Message Passing Interface), and routing strategies compared to those in Dragonfly topology deployments at Cray Inc. The storage and I/O subsystems integrated parallel file system concepts from Lustre and techniques developed at National Center for Supercomputing Applications, with metadata and checkpointing strategies similar to those used by NERSC and PRACE projects.
Tianhe-2 topped the TOP500 list in June 2013 with a reported LINPACK performance that eclipsed contemporaries such as Titan (supercomputer) and Sequoia (supercomputer), using measured metrics similar to those for HPL (High Performance Linpack), HPCG (High Performance Conjugate Gradients), and other community benchmarks like SPEC CPU and STREAM. The system's sustained throughput for scientific simulations compared to workloads run at Los Alamos National Laboratory, Argonne National Laboratory, and Lawrence Berkeley National Laboratory for climate modeling groups like Hadley Centre, GFDL, and NCAR (National Center for Atmospheric Research). Performance tuning practices employed tools from Intel VTune, Allinea (Now ARM Forge), TAU (Tuning and Analysis Utilities), and profiling techniques used in projects at CERN and Max Planck Society.
Development was led by the National University of Defense Technology with collaboration spanning Chinese institutions such as Tsinghua University, Peking University, Chinese Academy of Sciences, Zhejiang University, and industrial partners like Inspur and Huawei. Deployment at the National Supercomputer Center in Guangzhou followed project management approaches comparable to those at Jülich Research Centre and Riken installations, coordinated with operators from China Mobile, China Unicom, and administrators familiar with governance models seen at NSF (National Science Foundation) funded centers and European Grid Infrastructure sites. The procurement and supply chain involved multinational actors including Intel Corporation, and experienced disruptions analogous to episodes faced by Rosetta (spacecraft) mission teams or logistics for Large Hadron Collider upgrades.
Hardware racks housed combinations of Intel Xeon CPUs and coprocessors inspired by Intel Xeon Phi; power delivery and redundancy considered practices from data centers operated by Google, Facebook, and Microsoft. Cooling systems used techniques related to those at Lawrence Livermore National Laboratory and Oak Ridge National Laboratory, drawing on chilled-water designs and airflow management strategies resembling deployments at Poughkeepsie IBM facilities and innovations by Schneider Electric and Siemens. Thermal management included monitoring methods similar to ASHRAE guidelines and energy-efficiency metrics compared to Green500 rankings, with power usage effectiveness (PUE) considerations akin to those at Yahoo! and Netflix content delivery infrastructures.
Software stack combined operating systems including Kylin (operating system), CentOS, and Ubuntu with middleware influenced by projects at NERSC, XSEDE, and PRACE. Application workloads targeted computational domains familiar to NASA Ames Research Center, NOAA, European Space Agency, and laboratories such as Los Alamos National Laboratory for codes in LAMMPS, GROMACS, NAMD, OpenFOAM, WRF (Weather Research and Forecasting Model), and cosmology codes used by teams at CERN and Max Planck Institute for Astrophysics. Development environments used compilers and libraries from Intel Corporation, GNU Project, OpenMPI, MPICH, and toolchains similar to those at Argonne National Laboratory and Oak Ridge National Laboratory.
The system influenced national technology strategy discussions involving Ministry of Industry and Information Technology (China), State Council of the People's Republic of China, and international trade dialogues with entities like the United States Department of Commerce, European Commission, and firms such as Intel Corporation and NVIDIA. Controversies included export-control and procurement debates paralleling issues seen with Huawei (company), ZTE Corporation, and sanctions involving Semiconductor manufacturing international corporation; policy responses drew attention from legislators in the United States Congress, agencies such as U.S. Department of Defense, and commentators at Brookings Institution and Council on Foreign Relations. The project impacted academic and industrial research collaborations involving Tsinghua University, Peking University, Chinese Academy of Sciences, and international partners including University of Cambridge, Massachusetts Institute of Technology, and Stanford University, prompting discussion in venues such as IEEE conferences and publications in Nature (journal) and Science (journal).