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 | |
|---|---|
| Name | Fujitsu K |
| Type | Supercomputer |
| Manufacturer | Fujitsu |
| Introduced | 2011 |
| Architecture | Massively parallel scalar processors |
| Processors | SPARC64 VIIIfx |
| Peak | 10.51 PFLOPS (theoretical) |
| Memory | 1.6 PB aggregate |
| Storage | Lustre-based parallel file system |
| Operating system | Linux-based |
Fujitsu K Fujitsu K was a Japanese supercomputer developed by Fujitsu and installed at the Riken Advanced Institute for Computational Science. Announced in 2011 and operational in 2012, it achieved world-leading performance on the TOP500 list and served research projects spanning climate modeling, materials science, drug discovery, and seismology. The system combined bespoke SPARC64 processors, high-bandwidth interconnects, and a large-scale parallel file system to support petascale computation across academic and industrial partners.
K was designed to deliver petaflop-scale performance for large-scale simulations and data analysis at Riken's Kobe facility. It targeted workloads from national initiatives such as the High Performance Computing Infrastructure programs and collaborations with institutions including University of Tokyo, Kyoto University, and international centers like Oak Ridge National Laboratory and Lawrence Berkeley National Laboratory. K's installation coincided with contemporaneous machines like Titan (supercomputer), Sequoia (supercomputer), and Jaguar (supercomputer), influencing rankings on the Green500 and Top500 lists.
Fujitsu partnered with Riken to design a machine emphasizing energy efficiency and scalable performance, leveraging prior projects such as the K computer predecessor programs within Fujitsu and lessons from architectures like Earth Simulator. Design goals referenced national initiatives including Japan's Science and Technology Basic Plan and projects funded by the Ministry of Education, Culture, Sports, Science and Technology (Japan). The procurement and integration required coordination with vendors of interconnect technology, file systems, and cooling solutions from firms similar to NEC, Hitachi, Cray Inc., and vendors supplying InfiniBand-class technologies.
K's compute nodes were built around Fujitsu's custom-designed SPARC64 VIIIfx processors, each node containing multiple CPU cores, vector units, and large on-chip caches. The system used a high-radix network topology inspired by technologies employed in systems like K computer and designs from Cray families, tuned for low-latency message passing via MPI (Message Passing Interface) implementations analogous to those used on Blue Gene (supercomputer family) systems. Memory hierarchies and NUMA characteristics reflected practices from large machines such as IBM Roadrunner and Fujitsu PRIMEHPC FX10. Storage subsystems used parallel file system designs similar to Lustre (file system), with tape archive integration comparable to HPSS deployments at national laboratories.
K reached a theoretical peak in the petaflop range and obtained a sustained performance used to secure top positions on the TOP500 list during 2011–2012. Benchmarking included standardized suites such as HPL (High-Performance Linpack), with application-level benchmarks drawn from the SPEC benchmarks and domain-specific kernels similar to those used in NAS Parallel Benchmarks and codes ported from LAMMPS, GROMACS, and Quantum ESPRESSO. Comparisons were routinely made against contemporaries like Roadrunner, Tianhe-1A, and K Computer for sustained throughput and energy efficiency metrics tracked on the Green500 list.
The software stack consisted of a Linux-based operating environment augmented by system libraries and compilers for SPARC architecture, including vendor toolchains akin to the GNU Compiler Collection and high-performance math libraries comparable to Intel MKL equivalents. Parallel programming models supported included MPI (Message Passing Interface), OpenMP, and domain-specific frameworks used in climate and materials communities similar to WRF (Weather Research and Forecasting Model) and LAMMPS. Software distribution, job scheduling, and resource management used solutions analogous to SLURM, PBS (software), and middleware stacks adopted by major centers like NICS and NERSC.
K enabled simulations for national and international projects in areas such as long-term climate projection engaging groups like IPCC contributors, earthquake rupture modeling used by researchers linked with Japan Meteorological Agency datasets, and large-scale genomics analysis comparable to efforts at Broad Institute and Wellcome Sanger Institute. Industrial users in sectors like automotive, electronics, and pharmaceuticals collaborated through partnerships similar to those of Toyota, Sony, and Takeda Pharmaceutical Company to accelerate design simulations, materials computations, and molecular docking studies. Education and training programs at institutions such as Osaka University and Tohoku University leveraged K for graduate-level computational science curricula.
K's deployment reinforced Japan's position in international high-performance computing, influencing subsequent projects such as the Fugaku initiative and informing procurement strategies at national centers like Riken and corporate supercomputing efforts at Fujitsu and NEC. Lessons from K contributed to developments in exascale research programs, energy-efficient processor design, and interconnect technologies used in successor systems including Fugaku (supercomputer), collaborations with ARM Holdings-based efforts, and standards in scalable parallel file systems and scheduler designs used across European HPC and US DOE laboratories. K's operational record also fed into policy discussions at bodies like the Ministry of Economy, Trade and Industry (Japan) regarding science infrastructure investment.