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.
| Taiwania supercomputer | |
|---|---|
| Name | Taiwania supercomputer |
| Developer | National Applied Research Laboratories (NARLabs) |
| Country | Taiwan |
| Released | 2018 |
| Architecture | Heterogeneous x86_64, GPU-accelerated |
| Memory | Up to several terabytes |
| Storage | Petabyte-scale parallel file systems |
| Speed | Multi-petaflop class |
| Os | Linux-based |
Taiwania supercomputer is a high-performance computing system developed to support scientific research, industrial innovation, and national-scale simulations. It provides computing resources for climatology, bioinformatics, artificial intelligence, and engineering across Taiwanese and international institutions. The system has been deployed in research centers and cloud-like resource pools to accelerate projects spanning from genomics to weather forecasting.
The project was initiated by National Applied Research Laboratories and coordinated with agencies such as Ministry of Science and Technology (Taiwan) and research institutes including Academia Sinica, Industrial Technology Research Institute, National Taiwan University, National Tsing Hua University, National Yang Ming Chiao Tung University, National Cheng Kung University, National Chung Cheng University, and regional partners. Hardware and software procurement involved vendors like NVIDIA, Intel Corporation, AMD, Supermicro, Dell Technologies, and integrators with experience in systems such as Fugaku, Summit, Sierra, Sunway TaihuLight, and Theta. The initiative aligns with international collaborations involving organizations such as European Centre for Medium-Range Weather Forecasts, World Health Organization, CERN, National Aeronautics and Space Administration, and academic consortia that include Purdue University, Massachusetts Institute of Technology, Stanford University, University of Cambridge, and University of Tokyo.
The system uses a heterogeneous architecture combining x86-64 CPUs from vendors like Intel Xeon and accelerator GPUs from NVIDIA Tesla and NVIDIA A100 families, with nodes interconnected using high-speed fabrics such as InfiniBand and Ethernet variants employed in previous machines like Mellanox Technologies installations. Storage subsystems integrate parallel file systems inspired by deployments at Oak Ridge National Laboratory and Lawrence Livermore National Laboratory, with SSD tiers and HDD capacities similar to configurations used in Google, Amazon Web Services, and Alibaba Cloud science clusters. Cooling and power distribution employ designs comparable to systems at National Supercomputing Centre (Singapore) and Swiss National Supercomputing Centre, with rack-level power delivery and redundant uninterruptible power supply units from manufacturers like Schneider Electric.
Taiwania-class systems have been measured using benchmarks such as LINPACK, HPL, and application-level suites resembling those used on TOP500 and Green500 lists. Reported peak performance reaches multi-petaflop scale, comparable to entries like Marconi and Tianhe-2A, while sustained performance on domain codes aligns with results published by NERSC, PRACE, and national labs including Argonne National Laboratory and Lawrence Berkeley National Laboratory. Benchmarking workflows draw on communities behind SPEC, NAS Parallel Benchmarks, and IOzone to evaluate I/O throughput.
The software stack is Linux-based with distributions and package ecosystems akin to those used at Ubuntu, Red Hat, and CentOS environments. Programming models supported include MPI, OpenMP, CUDA, OpenACC, and frameworks such as TensorFlow, PyTorch, Keras, Scikit-learn, HDF5, NetCDF, and domain-specific tools used in projects at European Organisation for Nuclear Research, Broad Institute, and Max Planck Society. Resource managers and schedulers mirror deployments using SLURM, PBS Professional, and Kubernetes for containerized workflows similar to practices at Docker, Singularity, and Helm-orchestrated clusters.
The system supports climate modeling groups linked to Taiwan Central Weather Administration, epidemiology teams interacting with Centers for Disease Control (Taiwan), genomics initiatives connected to Human Genome Project-style consortia and the Broad Institute, and engineering simulations in collaboration with automotive and semiconductor firms such as TSMC and Foxconn. Use cases include numerical weather prediction comparable to outputs from ECMWF models, molecular dynamics tasks akin to studies at Riken, machine learning projects inspired by research at DeepMind, and earthquake modeling paralleling work at USGS and Japan Meteorological Agency. Access has been granted to academic users from National Central University, Tamkang University, and international partners through science gateway programs like those used by XENON1T and LIGO Scientific Collaboration.
The initial procurement and deployment phases began in the late 2010s with announcements involving Ministry of Science and Technology (Taiwan), contracts with vendors resembling those used in acquisitions by European High Performance Computing Joint Undertaking, and milestones celebrated at institutions like Academia Sinica and National Taiwan University. Subsequent expansions mirrored upgrade cycles seen at Oak Ridge and Argonne facilities, adding GPU-accelerated nodes and increased storage capacity to support projects funded by entities similar to the National Science Foundation, European Research Council, and private-sector partners such as Microsoft Research and Google DeepMind.
Energy efficiency efforts align with metrics promoted by Green500 and practices used at centers like Barcelona Supercomputing Center and Swiss National Supercomputing Centre that employ liquid cooling, hot-aisle containment, and waste heat reuse strategies. Power usage effectiveness (PUE) targets were informed by guidelines from organizations like ASHRAE and equipment vendors such as Schneider Electric and Emerson Electric. Cooling implementations reference industrial examples from Fugaku and data centers run by Facebook and Microsoft that integrate chilled water loops and direct-to-chip cooling for accelerator arrays.
Funding and partnerships have involved Taiwanese institutions including National Applied Research Laboratories, Ministry of Science and Technology (Taiwan), and universities like National Taiwan University alongside industry partners such as NVIDIA, Intel Corporation, TSMC, and major system integrators. International collaborations reference working relationships with organizations such as European Centre for Medium-Range Weather Forecasts, CERN, World Health Organization, National Aeronautics and Space Administration, and research networks like GEANT and GLORIAD. Research outcomes have been disseminated through conferences and publishers including SC (conference), SC, International Supercomputing Conference, IEEE, ACM, Nature, and Science.