Running high-performance neural networks on a "gamer" GPU

Tagged: Brain Science, Gamers, gpu, Super Computer, Computer Hardware, Gaming, Technology
Source: Ars Technica - Read the full article
Posted: 3 years 8 weeks ago

"A recent project here at the Laboratoire de Chimie de la Matière Condensée de Paris (LCMCP) wants to make high-performance scientific computing cheaper by finding new ways to squeeze performance from consumer-grade "gamer" hardware. The idea is nothing less than building the equivalent of a $400,000 custom high performance computing setup for only $40,000.

The cluster, known as HPU4Science, is up and running, and the team behind it is tackling difficult scientific problems by developing novel computational methods that make good use of HPUs—Hybrid Processing Units—like CPUs and GPUs. The current cluster is a group of six desktop-type computers powered by Intel i7 or Core 2 Quad processors, together with GPUs that range from the GTX 280 to the GTX 590..."