The VPP5000 vector supercomputer has up to 512 processors and a capacity of 4,9 Tflop/s and 8 Tbyte of memory
The hpcLine is a family of highly scalable high performance computers, which is based entirely upon standards and off the shelf components. Compute bound applications within areas such as flow simulation, structural analysis, computational chemistry, simulation of electromagnetic compatibility, seismology or also 3D image processing can efficiently be executed with the hpcLine.
hpcLine systems scale from starter kits with 8 compute nodes to configurations with hundreds of processors. The interconnect can be a traditional fast Ethernet or two high performance interconnect options: SCI or Myrinet.
The SCI solution is a scalable 2D torus topology with latency for message passing of 5µs and aggregate bandwidth for 16 nodes: 5.3 GB/s (4x4 torus). The SCI based system is packaged with Scali Universe Cluster Management Software an ScaMPI (Message Passing Library). The Myrinet2000 Interconnect (optional) has a latency for message passing of 8µs with a throughput 2 x 2 Gbit/s bi-directional. The Myrinet based system is packaged with a ParaStation communication subsystem with highly optimised PVM, MPI, Sockets and Java RMI and a ParaStation Single System View.
The system line PRIMEPOWER is extremely scalable and offers the highest RAS (Reliability, Availability, Serviceability) functionality within its six different models, which are all based on a Crossbar architecture. Model 200 can be built out with two processors, model 400 with four and model 600 can have up to eight processors. Each of these three models are available as floor-stand and as rack-mount systems.
In the high-end line, Fujitsu offers the 800 models with 16 SPARC64 processors, the 1000 model with 32 and the new top model, PRIMEPOWER 2000 with up to 128 processors with a single shared memory.
The following is a selection of typical HPC customers, mainly for the hpcLine.
- The Proteomics Centre at the Ruhr University in Bochum has implemented a Linux Cluster from Fujitsu Siemens Computers' hpcLine for DNA and protein analysis.
- 2. The car manufacturer Audi in Ingolstadt, Germany, uses several hpcLine systems with 52 Pentium III duo processor nodes and 24 Pentium 4 processor nodes with a memory of 57 Gigabyte for computer-aided engineering applications.
- The Leibniz Supercomputer Center has a VPP700/52 for scientific research.
- The Research Center Karlsruhe is using a VPP5000 and hpcLine clusters for scientific research.
- The Paderborn Center for Parallel Computing has the largest hpcLine configuration so far with 192 processors for computer science.
The following is a selection of typical HPC customers, mainly for the hpcLine.
- The Proteomics Centre at the Ruhr University in Bochum has implemented a Linux Cluster from Fujitsu Siemens Computers' hpcLine for DNA and protein analysis.
- 2. The car manufacturer Audi in Ingolstadt, Germany, uses several hpcLine systems with 52 Pentium III duo processor nodes and 24 Pentium 4 processor nodes with a memory of 57 Gigabyte for computer-aided engineering applications.
- The Leibniz Supercomputer Center has a VPP700/52 for scientific research.
- The Research Center Karlsruhe is using a VPP5000 and hpcLine clusters for scientific research.
- The Paderborn Center for Parallel Computing has the largest hpcLine configuration so far with 192 processors for computer science.
