Fujitsu today announced the global availability of the PRIMEHPC FX10 supercomputer, which is capable of scaling to a top theoretical processing performance of 23.2 petaflops (*1) (PFLOPS).
Combining high performance, scalability, and reliability with superior energy efficiency, PRIMEHPC FX10 further improves on Fujitsu’s supercomputer technology employed in the “K computer,” which achieved the world’s top-ranked performance (*2) in June 2011. All of the supercomputer’s components—from processors to middleware—have been developed by Fujitsu, thereby delivering high levels of reliability and operability. The system can be scaled to meet customer needs, up to a 1,024 rack configuration achieving a super-high speed of 23.2 PFLOPS.
By leveraging the new system, it will be possible to address societal challenges—including new drug development, disaster prevention, disaster mitigation, and other measures, to bring about a safe and secure society—and to pursue cutting-edge research, such as enabling the development and manufacturing of new materials without the need to make prototypes. This has the potential to help companies enhance their competitive edge.
In offering a high-performance and high-quality system on a global scale, Fujitsu aims to help expedite the academic research and product development of customers throughout the world.
Background
Recently, computer simulations have been employed in a wide range of applications, from basic research at universities and research institutions to corporate product development. Going forward, it is expected that such simulations will be leveraged to address problems facing society, such as drug discovery and medical research, analysis of earthquakes and weather phenomena, and development of new sources of energy. As a result, supercomputers, which serve as a platform for enabling computer simulations, are steadily increasing in performance and scale. In particular, in the area of high-end supercomputers, petaflops-class performance is in great demand.
To meet such needs, Fujitsu has developed PRIMEHPC FX10, a supercomputer capable of world-class performance of up to 23.2 PFLOPS. The new system will be available worldwide.
The new system is highly compatible with the “K computer” supercomputer, which is being developed by RIKEN and Fujitsu as part of an initiative led by Japan’s Ministry of Education, Culture, Sports, Science and Technology (MEXT). As such, it is optimal for developing software to be used on the K computer.
Fujitsu is aiming to bring about a “Human Centric Intelligent Society” by contributing to a prosperous future for society and customers through the continued development of supercomputers.
Product Features
1. Delivers a world-class, high-speed, ultra-large-scale computing environment of up to 23.2 petaflops
In its largest potential configuration consisting of 98,304 nodes, 1,024 racks, the supercomputer delivers a high-speed, ultrascale computing environment with a theoretical computing performance of 23.2 petaflops and 6 petabytes of memory.
2. SPARC64™ IXfx processors deliver high performance with low power consumption
PRIMEHPC FX10 runs on the newly-developed SPARC64™ IXfx processors, which offer a very significant boost in performance over the SPARC64™ VIIIfx processor on which they are based and which power the K computer. Each processor has 16 cores, and by employing a water-cooling method and highly-efficient design that emphasizes computing performance per watt of electricity consumed, achieves world-class standalone performance levels of 236.5 gigaflops and performance per watt of over 2 gigaflops.
3. Delivers high execution performance with massively parallel applications while reducing development workloads
PRIMEHPC FX10 features a variety of enhancements for high-performance computing, such as high memory bandwidth, scalable “Tofu” interconnects with 10 links, each delivering high performance levels of 5 GB/s x 2 (both directions), and Fujitsu’s original HPC middleware, Technical Computing Suite, which has a compiler and library, resulting in high execution performance for massively parallel applications.
VISIMPACT makes it easy to implement a hybrid parallel programming model that combines threads (*3) with Message Passing Interface (*4), which is effective for massively parallel programming, thereby reducing the burden of handling massively parallel programming for application developers.
4. Proprietary reliable technology and high operability for a large-scale system
Due to its processors, which include the same built-in RAS functions found in mainframes, and “Tofu” interconnects, which employ highly-flexible 6-dimensional mesh/torus architecture, PRIMEHPC FX10 delivers high reliability and availability.
Moreover, thanks to the system management and job operations management functions in the Technical Computing Suite HPC middleware, as well as its high-performance distributed file system, FEFS, that enables sharing across 100,000 nodes, PRIMEHPC FX10 delivers high operability for a large-scale system.
Product Specifications
Hardware
Node |
Theoretical computational performance |
236.5GFLOPS |
Processor |
SPARC64™ IXfx (1.848GHz, 16 cores) x 1 |
|
Memory capacity |
32GB, 64GB |
|
Memory bandwidth |
85GB/s |
|
Inter-node transfer rate |
5GB/s × 2 (bidirectional) / link |
|
System |
No. racks |
4~1,024 |
Nodes |
384~98,304 |
|
Theoretical computational performance |
90.8~23,248TFLOPS |
|
Total memory |
12~6,291TB |
|
Interconnect |
“Tofu” Interconnect |
|
Cooling Method |
Direct water cooling + air cooling (Optional: Exhaust cooling unit) |
Shipping Date
Starting January 2012
Sales Target
50 systems over the next three years.
Glossary and Notes
1. Petaflops
Quadrillion floating-point operations per second.
2. Top-ranked performance
According to the 37th TOP500 list announced on June 20, 2011, with a performance of 8.162 petaflops.
3. Thread Parallelization
A parallelization processing method in which a process is broken down into multiple threads. The automated parallelization functions of the compiler can be used for processing within the same node.
4. Message Passing Interface Parallelization
A parallelization processing method that uses a library called MPI (Message Passing Interface) to divide a program into multiple processes. Enables parallelization to be used within nodes and between nodes.
Press Contacts
Fujitsu Limited Public and Investor Relations Division
Inquiries: https://www-s.fujitsu.com/global/news/contacts/inquiries/index.html
About FujitsuFujitsu is a leading provider of information and communication technology (ICT)-based business solutions for the global marketplace. With approximately 170,000 employees supporting customers in over 100 countries, Fujitsu combines a worldwide corps of systems and services experts with highly reliable computing and communications products and advanced microelectronics to deliver added value to customers. Headquartered in Tokyo, Fujitsu Limited (TSE:6702) reported consolidated revenues of 4.5 trillion yen (US$55 billion) for the fiscal year ended March 31, 2011. For more information, please see:www.fujitsu.com
All SPARC trademarks are used under license and are trademarks of SPARC international, Inc. in the US and other countries. SPARC64 is a trademark of SPARC International, Inc. in the United States and other countries and used under license.
All other company or product names mentioned herein are trademarks or registered trademarks of their respective owners. Information provided in this press release is accurate at time of publication and is subject to change without advance notice.
Date: 07 November, 2011
City: Tokyo
Copyright 1995 - 2021 FUJITSU