Fujitsu has launcheded the PRIMERGY CX600 M1 system, purpose-built to meet the exacting demand for massive parallel computing power from today’s high performance computing (HPC) applications in fields such as scientific research, product development and business intelligence. The new system comprises the PRIMERGY CX600 M1 chassis and PRIMERGY CX1640 M1 server nodes, which deliver ultra-dense clusters and a highly competitive price/performance ratio1.
Best-in-class components are a hallmark of the new modular server including Intel® Xeon Phi™ processors with up to 72 cores, making the FUJITSU Server PRIMERGY CX600 M1 ideal for HPC workloads that require high thread parallelism, large vectors, and significant memory bandwidth.
The PRIMERGY CX600 M1 delivers extremely high density processing, with eight independent servers in a 2U chassis, and nine times the compute performance of a standard 1U dual-socket rack server. A standard rack will integrate up to 168 compute nodes, for an unprecedent total peak performance of more than 500 TFLOPS/DP.
PRIMERGY CX600 M1 servers will also form the core of the recently-ordered 25 Petaflop supercomputer system "Oakforest-PACS", for the universities of Tokyo and Tsukuba in Japan. Consisting of 8208 compute nodes, the system is planned to be operational in December this year and will be one of the fastest supercomputers worldwide.
Uwe Neumeier, Vice President and Head of Datacentre, EMEIA Product Business at Fujitsu, commented: “As enterprises embrace digitisation, they demand ever more of their high performance computing infrastructure – for simulation and scenario modeling, including big data analysis from the Internet of Things. As it is designed to handle highly parallel processing, this is one reason why we consider the new PRIMERGY CX600 M1 to be a game changer in terms of enabling enterprises to extract value from large amounts of data in a datacentre environment that is fit for the future.”
The new system extends Fujitsu’s HPC portfolio of computing hardware, system software and user-ready middle ware. It is part of the Fujitsu Integrated System PRIMEFLEX for HPC, where all components are pre-defined and validated, with system software and middleware factory-installed for immediate project readiness. Combined with Fujitsu’s intuitive and innoviative web based HPC workplace – the HPC Gateway – users benefit from an easy to use and even more powerful HPC environment.
Shared infrastructure reduces energy consumption
Server nodes in the PRIMERGY CX600 M1 use a common, shared infrastructure, with central cooling fans and hot-plug power supplies integrated into the chassis. This reduces energy consumption while also allowing individual nodes to be serviced without affecting others. Furthermore, the full redundancy of shared components ensures uniform higher system availability.
Customers can also choose Fujitsu’s innovative direct-to-chip hot water (45°C / 113°F) based Cool-Central® Liquid Cooling, which removes heat directly from the server’s CPU and voltage regulator module. This alone can halve datacentre cooling costs and allows for higher density, enabling implementation of even the most ambitious HPC projects without the need to increase power and cooling supplies in existing datacentres.
