Penguin Computing has started peddling a couple of fully integrated Nvidia GPU-based clusters targeted at technical workgroup computing. Their CPUs are AMD Shanghais.

The Linux-based Altus 1702 cluster has four twin 1U compute nodes, four Tesla S1070 GPUs, Gigabit Ethernet and Penguin's own Scyld management software for $44,985. It provides over 16 teraflops of compute power in a 9U rack configuration.

A cluster option expands the widgetry to eight twin compute nodes and eight GPUs for over 32 teraflops of performance for roughly $89,000.

All systems have Opteron 2376 CPUs and 8GB RAM per node, a 24-port GigE switch, and a Scyld annual subscription included for each node.

Each Tesla system has four T10 GPUs, each delivering almost one teraflop of single-precision and 80 gigaflops of double-precision performance.

The company says IT managers at companies and researchers at universities can order these pre-configured clusters and work with Penguin to customize them to their application set.

They are supposed to offer a quick and easy entry point to customers who have been considering adopting GPU technology. And the Scyld ClusterWare is supposed to ensure that customers can take advantage of GPU performance essentially out-of-the-box.

Penguin claims GPU technology offers superior performance for applications in areas such as life sciences, manufacturing, finance, signal processing and image processing.