The Southern GPU-cluster (SGPU-C): Graphics High Performance Computation Platform for the Associativity and Accelerate Research in the Life Sciences. This project was awarded by the University of Concepción, EQM150134. Funded by FONDEQUIP, http://www.conicyt.cl/fondequip) supported by Conicyt, Chile.
Aims:
- Support excellence research by incorporating GPU-HPC technology into model development and implementing simulations and optimization.
- Accelerate research in Life Sciences.
- Strengthen the partnership between researchers from the University of Concepcion and other national researchers to perform interdisciplinary research.
- Conduct dissemination of GPU-HPC utilization in industry and educational environments.
GPU-based computing. Computer equipment is clusters of unified computers called high performance computers (HPCs) that are based on computation in their central processing units, CPU. However, the computation in graphics cards, denominated GPU, has exceeded the processing speed of the units cost / benefit CPU. The barrier of its particular architecture has been overcome and practically all the applications available for CPU, have versions compatible with GPU with superior performance. This allows us standards of the most modern computing centers in the world.
The equipment are ~281,000 cores 420 Theoretical Tflops
- 52 Intel® Xeon® CPU
- 48 Tb Storage
- 100 GTX980Ti GPU
- 1,664Gb DDR4 RAM
- Switch 10Gb Network
- 1 APC SMART-UPS VT 40KVA
The equipment is composed by a:
Head Node.
- 2X Intel® Xeon® processor E52650 v3, 10C, 2.3 GHz 25M, Support DDR42133
- 8X 8GB DDR4 2133 ECC/REG Memory Module
- 17 4TB 7200RPM 64MB CACHE 3.5IN SATA Enterprise Class HDD in Raid 6
25X Compute Node
- 2X Intel® Xeon® processor E52650 v3, 10C, 2.3 GHz 25M, Support DDR42133
- 4X Nvidia GeForce GTX 980 Ti 6GB GDDR5, 256-bit, PCI Express 3.0
- 8X 8GB DDR4 2133 ECC/REG Memory Module
- 1TB 7200RPM 64MB CACHE 3.5IN SATA Enterprise Class HDD
Convergent computing requirements. The various disciplines of the life sciences form multidisciplinary teams to investigate phenomena of high complexity and interrelation. This approach includes researchers from the fields of chemistry and natural sciences to disciplines of applications such as biotechnology pharmacy. A common aspect of these is the availability of large databases and the increase of equipment that acquire large volumes of information in a single experiment. Thus, as the interpretation, analysis, simulation and prediction of phenomena are multifactorial problems that can only be approached from the scientific perspective with the support of adequate computer equipment.
Administration.
- Brigth Computing v7.2.
Compilators.
- GNU C v4.4.7/v5.2.0
- CUDA v7.5
Parallelization
Libraries
- FFTW v3.3.4/v2.1.5
- BLAS
- LAPACK
Molecular Dynamics
- To applied a account:Read the “Políticas de Uso”.
- Read the Bright User Manual.
- Fill this Google form.