IBM Power 755 server

Highlights

• Optimised for running highly parallel computationally intensive workloads and algorithms such as weather and climate modelling, computational chemistry, physics and petroleum reservoir modelling.

The Power 755 server is a four-socket, 4U rack-mounted server that offers 32 POWER7 processor cores running at a frequency of 3.6 GHz. A maximum of 64 GB of memory is allowed on each processor card, making for a system maximum of 256 GB of memory in the four-processor card system. The IBM Power 755 compute node is designed for organisations that require a scalable system with extreme parallel processing performance and dense packaging.

Ideal workloads for Power 755 include high performance computing (HPC) applications such as weather and climate modelling, computational chemistry, physics and petroleum reservoir modelling that require highly intense computations where the workload is aligned with parallel processing methodologies. The Power 755 supports AltiVec instruction set and extended VSX Single Instruction Multiple Data (SIMD), acceleration, which can execute up to eight single-precision or double-precision floating point operations per clock cycle per core to improve fine-grained parallelism and accelerate data processing. Using 12X InfiniBand adapters up to 64¹ Power 755 nodes, each with 32 cores, can be clustered together providing up to 2,048 POWER7 cores. The IBM HPC software stack provides the necessary development tools, libraries and system management software necessary to manage a Power 755 server cluster running AIX 6.1.

The Power 755 is an ENERGY STAR-qualified server designed with features to help clients become more energy efficient. The superb performance of the IBM Power 755 translates into outstanding performance per watt and IBM Systems Director Active Energy Manager exploits EnergyScale technology enabling advanced energy management features to help clients achieve the most efficient computing environment for their systems.

¹ Capability to create a cluster of up to 64 nodes is planned to be available from March, 2010. Clusters larger than 64 nodes will be considered for support on a case-by-case basis.