GRAPHICS TESTBED
Today, GPUs are powerful computational devices which are able to provide more than an order of-magnitude acceleration on a number of problems. The goal of the Graphics Polygon at RCC MSU is to provide a testbed for applications which use GPUs. It allows to run applications and test GPU programming tools on GPUs from two major manufacturers, AMD and NVidia. Server-grade GPUs, Tesla and FireStream, are available at Graphics Polygon. Researchers from MSU and other organizations can access machines with GPU remotely to run their applications and to solve computationally intensive problems.
At the end of Q3 2010, Graphics Polygon will be upgraded with a new single-rack GPU-cluster. This cluster will have latest-generation NVidia Fermi and AMD Fire-Stream GPUs and will reach 30 TFlops peak performance (double precision). It will also feature high-speed Infini-Band QDR networking between nodes. GPU-cluster will be used for numerically intense computations and as a testbed for developing larger heterogeneous GPUbased systems.
GPU testbed website: http://gpu.parallel.ru
Performance of solving N-Body problem on diff erent GPU architectures
Modeling liquid crystalline phases of DNA. Courtesy of V. Golo, MM MSU
Modeling particle fl ow in a cyclotron. Courtesy of E. Perepelkin, JINR
User login
