Fujitsu PrimeHPC FX supercomputer series

2021: Fujitsu PrimeHPC FX1000 and FX700 as part of Ansys LS-Dyna

On October 21, 2021, Ansys introduced the CAE system for solving Ansys LS-Dyna strength problems.

According to the company, Ansys LS-Dyna is a commercial CAE system for solving strength problems, supporting the 64-bit Arm architecture based on the FUJITSU PRIMEHPC FX1000 and FX700 supercomputers. More details here.

2011: Availability PrimeHPC FX10

In November 2011, Fujitsu announced the global availability of the PRIMEHPC FX10 supercomputer, whose theoretical data processing performance can reach 23.2 petaflops (* 1) (PFLOPS).

PRIMEHPC FX10, which combines high performance, scalability and reliability with excellent energy efficiency, improves Fujitsu supercomputer technologies implemented in the "K computer" system, which set a world performance record (* 2) in June 2011. All components supercomputer - from processors to intermediate ON - were developed by Fujitsu to achieve the highest level of reliability and best performance. The system is able to scale according to the needs of the customer up to a configuration combining 1024 racks, allowing to achieve an ultra-high calculation speed of 23.2 PFLOPS.

The new system will help in solving socially important problems - the creation of new medicines, the prevention of natural disasters, the mitigation of their consequences, ensuring the safety and protection of people - as well as in conducting advanced research, for example, for the development and production of new materials without prototyping. This will enable companies to expand their competitive advantage.

By offering a high-performance, high-quality system globally, Fujitsu aims to help customers worldwide accelerate research and product development.

Recently, computer modeling has been used in a variety of fields - from basic research conducted by universities and research institutes to the development of commercial products. In the future, such modeling can be expected to solve such problems facing society as medical research, analysis of earthquakes and other phenomena of nature, and the development of new sources of energy. Thus, supercomputers, which serve as a platform for computer modeling, should steadily increase their performance and scale. In particular, in the field of high-end supercomputers, there is a huge demand for petaflops performance.

To meet this demand, Fujitsu developed the PRIMEHPC FX10, a world-class supercomputer whose performance can reach 23.2 PFLOPS. The new system will be available worldwide.

It is fully compatible with the supercomputer "K computer," created by the RIKEN Institute and Fujitsu as part of the initiative of the Japanese Ministry of Education, Culture, Sports, Science and Technology (MEXT). The system is optimally suited for the development of software designed for the supercomputer "K computer."

Continuously developing supercomputers, Fujitsu takes care of the interests of customers and seeks to contribute to the creation of a prosperous "man-centric intellectual society."

Product Features

1. World-class high-speed computing for ultra-scale computing with up to 23.2 petaflops

The supercomputer, in its largest potential configuration, containing 98,304 nodes in 1024 racks and 6 petabytes of memory, will realize a high-speed ultra-scale computing environment with a theoretical performance of 23.2 petaflops.

2. SPARC64 IXfx processors with high performance and low power consumption

The supercomputer PRIMEHPC FX10 has new SPARC64 IXfx processors, which provide a very significant increase in performance compared to the SPARC64 Αfx processors that form the basis of the "K computer" system. Each processor contains 16 cores and, thanks to liquid cooling and a high-performance design, has excellent computing performance per watt of power consumed: autonomous performance is 236.5 gigaflops, and performance per watt exceeds 2 gigaflops.

3. High performance for mass concurrency applications while reducing development effort

The PRIMEHPC FX10 supports many improvements for high-performance computing: wide memory bandwidth, scalable "Tofu" interconnections with 10 connections, each of which provides a high level of performance of 5 GB/s x2 (in both directions) as well as Fujitsu's specialized Technical Computing Suite middleware for HPC (HPC), which contains a compiler and a library for high performance when running mass concurrency applications.

The VISIMPACT architecture simplifies the implementation of the hybrid parallel programming model, combining the methods of efficient parallelization of computing flows (* 3) with the Message Passing Interface (* 4), and thereby facilitates programmers to develop applications with mass parallelism.

4. Patented reliable technology and high performance of a large-scale system

With mainframe-specific processors with RAS (Reliability, Availability, Serviceability) and "Tofu" intercomponents that provide a highly flexible 6-dimensional toric network architecture, the PRIMEHPC FX10 supercomputer delivers high reliability and availability.

Moreover, the Technical Computing Suite HPC system and operations management features, as well as its own high-performance distributed FEFS file system, which provides data sharing from 100,000 nodes, provide the large-scale PRIMEHPC FX10 system with excellent performance.