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.
| Fujitsu K computer | |
|---|---|
| Name | K computer |
| Manufacturer | Fujitsu |
| Operator | RIKEN |
| Country | Japan |
| Released | 2011 |
| Peak flops | 10.51 PFLOPS |
| Architecture | SPARC64 VIIIfx |
| Nodes | 88,128 |
| Os | ITRON-derived microkernel and Linux |
Fujitsu K computer was a petascale supercomputer developed by Fujitsu and operated by RIKEN at the Kobe site in Japan. Announced in 2011 and achieving mainstream operation in 2012, the system reached sustained performance leadership in the TOP500 and influenced designs at institutions such as Oak Ridge National Laboratory, Lawrence Berkeley National Laboratory, and European Centre for Medium-Range Weather Forecasts. The project drew collaboration among corporate, academic, and governmental organizations including Fujitsu Limited, AICS, and the MEXT.
The system targeted exascale roadmap goals inspired by initiatives at DARPA, DOE, and national programs in China and South Korea. Built to support simulation campaigns for agencies such as the Japan Aerospace Exploration Agency and research consortia at University of Tokyo and Kyoto University, it prioritized power efficiency, scalability, and reliability. The deployment at the Kobe facility served national science priorities aligned with projects like Project Prometheus (Japan) and international collaborations at the ITER.
K used a massively parallel distributed-memory design based on the SPARC family: Fujitsu's proprietary SPARC64 VIIIfx processors, each chip integrating eight cores and vector units inspired by architectures seen in Cray XT5 and IBM Blue Gene/P. The machine comprised 88,128 compute nodes interconnected via a proprietary high-speed network, drawing on concepts from the Torus and Fat-tree topologies used in systems such as Sequoia (supercomputer) and Roadrunner (supercomputer). For cooling and power, the installation faced constraints similar to those encountered at Los Alamos National Laboratory deployments and used facility designs comparable to the Green500-aware halls at Argonne National Laboratory. Storage used parallel file system approaches akin to Lustre and management stacks influenced by embedded microkernel work from Toshiba research.
In the June 2011 TOP500 list, K achieved 10.51 petaflops peak and 8.162 petaflops LINPACK sustained, surpassing systems like Jaguar (supercomputer) and Roadrunner (supercomputer), and briefly topping IBM Summit predecessors. Its energy efficiency placed it prominently on the Green500 ranking, prompting comparisons with power-conscious designs at Fujitsu and benchmarks run by AIST. Benchmarks included LINPACK, HPC Challenge, and real-application workloads from Computational Fluid Dynamics teams at NASA Ames Research Center and materials science groups at Oak Ridge National Laboratory.
The software stack combined an ITRON-derived microkernel for system management alongside Linux for compute nodes, integrating MPI implementations similar to Open MPI and MPICH. Compilers and toolchains were based on Fujitsu's proprietary compilers influenced by GCC practices and optimizations used in Cray and IBM toolchains. Middleware supported parallel libraries such as ScaLAPACK and scientific packages used by users from KEK and RIKEN collaborators. Performance analysis used tools and methodologies from PAPI and profiling approaches seen at Lawrence Livermore National Laboratory.
Researchers used the system for large-scale simulations in climate science for groups at MRI, earthquake and tsunami modeling with teams from Tohoku University, cosmology simulations by researchers associated with IPMU (Kavli Institute for the Physics and Mathematics of the Universe), and materials modeling for projects linked to Japan Fine Ceramics Center. Workflows included multi-physics codes developed in collaboration with JAXA and computational chemistry packages comparable to those used at Max Planck Institute for Plasma Physics and Paul Scherrer Institute studies. The system supported national initiatives in disaster mitigation and industrial R&D by firms such as Nissan Motor Company and Mitsubishi Heavy Industries.
Development began as a multi-year program led by Fujitsu with funding and strategic direction from MEXT and coordination with RIKEN. Prototyping phases referenced lessons from projects like Earth Simulator and international procurements at ECMWF. Deployment at the Kobe center required extensive facility upgrades similar to those at Jülich Research Centre and coordination with local authorities in Hyōgo Prefecture. Operational management followed governance models used at National Center for Supercomputing Applications and included user access policies for academia, government, and industry.
K influenced Fujitsu's subsequent initiatives and hardware roadmaps that informed successors such as projects leading to systems at Fujitsu PRIMERGY lines and collaborations in the Post-K development path. Its impact shaped Japan's national strategy for high-performance computing alongside programs at AIST and Academia Sinica partnerships. Lessons from K contributed to software ecosystems and influenced global designs evident in successors like Fugaku and architectures pursued by vendors including NEC and HPE.