Tesla GPU SimCluster

Boston, a distribution partner for Super Micro Computer, has introduced the Nvidia Tesla GPU SimCluster, an off-the-shelf, ready-to-deploy cluster solution. Available in configurations that deliver 8 to 42 teraflops of HPC performance, it delivers the easiest way to buy, use and manage a GPU-based research cluster. With the Tesla GPU SimCluster, academic research departments can take GPU-accelerated apps from workstation to multi-node cluster, speeding research and discovery.

Boston offers three preconfigured GPU computing configurations to jumpstart the research of scientist in various scientific domains. The basic SimCluster configuration consists of eight Nvidia Tesla M2075 GPUs and four Intel CPUs, providing a compute performance of eight teraflops. The price of this configuration is £26,990. The performance configuration combines 16 Nvidia Tesla M2090 GPUs and eight Intel CPUs to a compute performance of 20 Teraflops, priced at £52,990. And finally, the maximum configuration hosts 32 Nvidia Tesla M2090 GPUs and 16 Intel CPUs in a 42U rack providing 42 teraflops of compute power. This price for this supercomputer will be £112,999.


Analysis and opinion

Robert Roe investigates some of the European projects focusing on preparing today’s supercomputers and HPC programmers for exascale HPC


The Jülich Supercomputing Centre (JSC) at Forschungszentrum Jülich in Germany has been operating supercomputers of the highest performance class since 1987. Tim Gillett talks to Norbert Attig and Thomas Eickermann


Gemma Church investigates how simulation and modelling de-risks the detection and extraction of geothermal energy resources


Robert Roe investigates the importance of upgrading legacy laboratory informatics systems and the benefits this provides to scientists