Using Grid Computing Network to Access GPGPUs Computing Platforms

Monica Hernandez, Engineer of the High Performance and Scientific Computing Unit of UIS-Colombia does CUDA Instruction to SC-CAMP'12 Participants

"Today is the GPGPU Computing Day in SCCAMP'12. Using GridUIS-2, the distributed computing platform of the  Universidad Industrial de Santander in Colombia, the participants can access to GUANE-1, the most famous GPGPU platform in Colombia, because is the most powerful GPGPU cluster of the country in this moment. With 64 C2075 NVIDIA TESLA GPUs distributed in 8 nodes, each one with 104 GB of RAM, 3 high bandwidth networks and more special characteristics, GUANE-1 was performed to support mainly research activities in different domains related with energy, environment and bio in tropical conditions.

GPGPU computing allows to treat scientific problems exploiting GPUs, graphic cards designed to support multi-threading. Different platforms in the world are deployed to some projects, from science life to weather research. Nowadays, GPGPU computing is more than a trend in high performance and scientific computing. During the CUDA Day, the "SCCAMPers" learn basis of CUDA seeing specific examples seen with other paradigms (as MPI and OpenMP) to compare performance and other features. The CUDA Day is managed by Pedro Velho, researcher at UFRGS in Porto Alegre, Brazil and Monica Hernandez, Engineer of the High Performance and Scientific Computing Unit (SC3-UIS) in Bucaramanga, Colombia. "