ACADEMIA
Exxact launches the NVIDIA Tesla GPU SimCluster Test Drive program and co-hosts an HMPP workshop at SC11
Exxact is collaborating with NVIDIA to launch the NVIDIA Tesla GPU SimCluster Test Drive program, and is co-hosting an HMPP workshop with NVIDIA and CAPS at the SC11 Conference.
The Tesla GPU SimCluster Test Drive program offers a special opportunity for research application users to evaluate GPU computing performance by running GPU-accelerated simulation models on a 4 Node M2090 Cluster or C2075 workstation remotely via dedicated broadband connections. These GPU SimCluster systems will deliver from 10 to 42 Teraflops, making more complex and realistic simulations possible with the massively parallel processing power of GPUs.
NVIDIA Tesla GPUs transform standard workstations into “computational laboratories” capable of running complex computing in fields such as Molecular Dynamics (AMBER, NAMD) and Computational Electromagnetics. Complex simulations that once required supercomputing resources can now be scaled down to the individual GPU cluster or workstation. In addition, these systems can be optimized through customization for increased accuracy and reduced wait time.
Exxact will also hold a HMPP workshop with CAPS and NVIDIA at the Pan Pacific Hotel in Seattle on Monday, November 14th, prior to the opening of SC11. The workshop aims to provide a detailed introduction to HMPP, a source to source complier for porting software onto many core systems, and to demonstrate how HMPP can take advantage of hardware such as GPUs to increase computing performance.
At the workshop, Duncan Poole, senior manager, HPC, from NVIDIA, will discuss the use of directives in parallel programming and NVIDIA’s growing third-party developer ecosystem. To better understand HMPP porting methodology, Francois Bodin, CAPS CTO and Founder, will provide a technology overview and share his 10 years of experience in code porting based on the HMPP programming model and toolset. The workshop will also include the introduction of the newly released HMPP 3.0 Workbench that enables developers to safely move applications to a hybrid CPU/GPU model, with CUDA and OpenCL.