Fujitsu PRIMEHPC

Fujitsu PRIMEHPC
Name	PRIMEHPC
Developer	Fujitsu
Released	2017
Cpu	SPARC64, A64FX
Gpu	none
Os	Linux
Type	Supercomputer family

Fujitsu PRIMEHPC Fujitsu PRIMEHPC is a family of supercomputing systems designed by Fujitsu Limited for high-performance computing deployments in research, industry, and government, positioned alongside contemporaries such as Cray Inc., IBM, Hewlett Packard Enterprise, Intel Corporation, and NVIDIA Corporation. It targets workloads common to institutions like University of Tokyo, RIKEN, Los Alamos National Laboratory, Lawrence Livermore National Laboratory, and Argonne National Laboratory while integrating technologies influenced by projects involving ARM Holdings, Sony Corporation, Toshiba Corporation, Riken Center for Computational Science, and standards bodies including The Open Group.

Overview

PRIMEHPC series combines custom processors, interconnects, and software stacks to serve computational science use cases similar to systems such as K computer, Fugaku, Summit, Sierra, and Perlmutter. It is marketed to organizations like National Institute of Advanced Industrial Science and Technology, CERN, Max Planck Society, CEA, and DESY for simulation, modeling, and data analytics tasks that also feature in projects led by NASA, European Space Agency, US Department of Energy, and Chinese Academy of Sciences.

Architecture and Hardware

The architecture emphasizes processor designs from the lineage of SPARC, ARM architecture, and microarchitecture innovations associated with partners including Riken Center for Computational Science and manufacturers such as Toshiba Corporation and NEC Corporation. PRIMEHPC machines have used processors related to SPARC64, the ARM-based A64FX implemented with Fujitsu Semiconductor process technologies used by fabs like TSMC and design flows common to Cadence Design Systems, Synopsys, and ARM Holdings' ecosystem. Interconnect solutions draw from concepts in systems like InfiniBand, Dragonfly topology, Omni-Path, and custom networks comparable to designs deployed by Cray Research and Hitachi, Ltd..

Software and Ecosystem

The software stack integrates operating systems and tools such as Linux, compilers from vendors like GNU Project, Arm Ltd. toolchains, and commercial compilers from Fujitsu Limited and Intel Corporation for cross-platform optimization used in projects with OpenMP, MPI, OpenACC, and libraries similar to LAPACK, BLAS, ScaLAPACK, and HDF5. Development and job orchestration employ middleware comparable to Slurm Workload Manager, TORQUE Resource Manager, PBS Professional, and monitoring suites influenced by Prometheus (software), Ganglia (software), and Nagios. Ecosystem collaborations echo partnerships between Riken Center for Computational Science, software consortia such as The HDF Group, and standards groups like Khronos Group.

Models and Product Line

The product line evolved alongside machines comparable to K computer and Fugaku, with model variants offering different processor and interconnect configurations aimed at customers including MEXT, METI, National Institute of Informatics, and corporations like Mitsubishi Heavy Industries and NEC Corporation. Model names and configurations reflect scalability ranges similar to offerings from Cray Inc., HPE Cray, and rack-scale systems provided by Dell EMC. Optional modules support storage systems analogous to Lustre (file system), GPFS, and flash tiers provided by vendors like Western Digital, Seagate Technology, and Micron Technology.

Performance and Benchmarks

PRIMEHPC systems have been characterized through benchmarks in categories akin to High Performance Linpack, HPCG, SPEC CPU, and domain-specific tests used by institutions such as Sandia National Laboratories and Argonne National Laboratory. Reported performance metrics place certain configurations in contention with machines like Fugaku, Summit, and Sierra on metrics valued by organizations including Top500, Green500, and Graph500. Performance tuning often involves collaboration with compiler projects like LLVM and numerical libraries maintained by communities such as BLAS, Netlib, and research groups at University of California, Berkeley and Massachusetts Institute of Technology.

Deployment and Use Cases

Deployments have targeted weather and climate modeling centers such as Japan Meteorological Agency, computational chemistry groups at Riken, and life sciences teams in institutions like The Francis Crick Institute and Wellcome Sanger Institute. Use cases include large-scale simulations for projects associated with Large Hadron Collider, ITER, European Weather Centre, and national research programs in Japan. Industries adopting PRIMEHPC-like systems mirror customers of Siemens, BASF, Shell plc, and Toyota Motor Corporation for design optimization, computational fluid dynamics, and materials science workflows used by consortia such as CINECA and Jülich Research Centre.

Development History and Timeline

Development traces through milestones connected to earlier programs such as collaborations between Fujitsu Limited and Riken Center for Computational Science that produced systems succeeding the K computer and culminating in later joint efforts that influenced Fugaku. The timeline features procurement and installation events comparable to announcements from Top500 releases, procurement projects by MEXT, and procurement cycles similar to those documented by DOE National Laboratories. Ongoing evolution reflects shifts in processor and accelerator strategies observed across the industry among firms like Intel Corporation, NVIDIA Corporation, AMD, and consortiums such as E4 Computer Engineering.

Category:Supercomputers