Nvidia Cuda Apps Jun27 11
1 like829 views
The document outlines advancements in GPU performance for various applications, focusing on significant improvements in molecular dynamics, visualization, quantum chemistry, and structural mechanics. It lists various software and tools, their release statuses, and the corresponding performance enhancements when utilizing single and multi-GPU systems. Furthermore, it highlights benchmarks comparing GPU performance to multi-core x86 CPU sockets across diverse fields, including scientific computing and engineering simulations.
Nvidia Cuda Apps Jun27 11
- 1. –– Accelerating High Performance Applications
- 2. Strategic Focus on Applications Senior-level relationship and market managers Dedicated technical resources More than 150 people devoted to libraries, tools, application porting and market development Worldwide focus
- 3. Reaching a Broad Range of Markets Scientific computing Creative pro Education / research
- 4. Strategic Partners CAD/ CAM/ CAE/ EDA Computational Computational Defence & Digital Physical Seismic CAID chemistry Finance Intelligence Content Sciences processing creation and visualization Autodesk Ansys Amber MATLAB Ikena Adobe Quda (L-QCD) Schlumberger Dassault Dassault NAMD Mathematica Intergraph Autodesk M&E WRF Landmark Systemes: Systemes: CATIA Simulia Solidworks PTC Nastran Gromacs NAG ESRI Avid ACUSA Paradigm Siemens LSTC Lammps Murex Manifold MainConcept HOMME Synopsys GAMESS Sony HYCOM
- 5. Leading MD Applications Features Application GPU Perf Release Status Notes Supported PMEMD : Single and multi-GPUs. AMBER Explicit & Implicit 8X V11 Released Expect 2x more performance in Solvent V11 patch release (shortly) Implicit (5x), Explicit Single GPU released, Next release: 2H2011 GROMACS (2x) Solvent 2x-5x Version 4.5.4 Better Explicit, MPI Lennard-Jones, Gay- LAMMPS Berne 6x Released Single and multi-GPU. Non-bond force NAMD calculation 2x-7x Released, v2.8 Single and multi-GPU. GPU Perf compared against Multi-core x86 CPU socket. GPU Perf benchmarked on GPU supported features and may be a kernel to kernel perf comparison
- 6. Additional MD/MM Applications Ramping Features Application GPU Perf Release Status Notes Supported TBD, 4-29X Single GPU. Abalone “Simulations” (on 1060 GPU) Released Agile Molecule, Inc. Production bio-molecular “µ-sec long Written for use on dynamics (MD) software specially ACEMD GPUs trajectories on Released optimized to run on single and workstation” multi-GPUs Two-body Forces, Link- V 4.0 Source only Next release: 2H2011 DL_POLY cell Pairs, Ewald SPME 4x Results Published Multi-GPU, multi-node supported forces, Shake VV HOOMD- Written for use on 2X Released, Version Single and multi-GPU. (32 CPU cores vs. GPUs 0.9.2 Blue 2 10XX GPUs) GPU Perf compared against Multi-core x86 CPU socket. GPU Perf benchmarked on GPU supported features and may be a kernel to kernel perf comparison
- 7. Viz and “Docking” Applications Related Features GPU Perf Release Status Notes Applications Supported Visualization from Visage 3D visualization of Imaging. Next release, 5.4, will Amira 5® volumetric data and N/A Released, Version 5.3.3 use GPU for general purpose surfaces processing in some functions Core GPU accelerated Up to Released, Suite 2011 Single and multi-GPUs. application 5000X Schrodinger, Inc. Hopping Real-time shape Single and multi-GPUs. FastROCS similarity 800-3000X Released Open Eyes Scientific Software searching/comparison High quality rendering, large structures (100 million atoms), GPU acceleration for 100-125X or Visualization from University of VMD computationally demanding analysis and visualization tasks, multiple GPU support for very fast display of greater Released, Version 1.9 Illinois at Urbana-Champaign molecular orbitals arising in quantum chemistry calculations GPU Perf compared against Multi-core x86 CPU socket. GPU Perf benchmarked on GPU supported features and may be a kernel to kernel perf comparison
- 8. Quantum Chemistry Features GPU Application Release Status Notes Supported Perf Libqc with Rys Single GPU supported in 10/1/10 Quadrature Algorithm, release. GAMESS-US integral evaluation, 2.5X Released Multi-GPU supported in closed shell Fock July 2011 release. matrix construction Triples part of Reg- Development GPGPU CCSD(T), CCSD & 3-8X Date TBA, NWChem EOMCCSD task projected in development benchmarks: www.nwchem- sw.org schedulers Date TBA, Various features 8-14x Q-CHEM including RI-MP2 projected In development Significant porting already 44-650X Single and Multi-GPU. “Full GPU-based vs. Completely redesigned to exploit TeraChem solution” GAMESS Version 1.45 released massive GPU parallelism CPU ver. GPU Perf compared against Multi-core x86 CPU socket. GPU Perf benchmarked on GPU supported features and may be a kernel to kernel perf comparison
- 9. Material Science Features GPU Application Release Status Notes Supported Perf BigDFT - 50% of the http://inac.cea.fr/L_Sim/BigDFT Abinit program (short 6-30X Released June 2009 /news.html convolutions) Quantum- PWscf package: linear algebra (matrix Created by Irish Centre for High- Espresso/ multiply), explicit TBD Released May 5, 2011 End Computing computational kernels, PWscf 3D FFTs GPU Perf compared against Multi-core x86 CPU socket. GPU Perf benchmarked on GPU supported features and may be a kernel to kernel perf comparison
- 10. Bioinformatics CUDA-BLASTP HEX Protein Docking CUDA-EC Jacket (MATLAB Plugin) CUDA-MEME MUMmerGPU CUDASW++ (Smith-Waterman) MUMmerGPU++ DNADist SARUMAN GPU Blast SeqNFind GPU-HMMER UGENE Additional details can be found at Tesla Bio Workbench: http://www.nvidia.com/object/tesla_bio_workbench.html
- 11. Structural Mechanics Application GPU Features GPU Perf Release Status Notes ANSYS Mechanical Linear eqn solvers 2x Total Today, release 13 SP2 FE implicit, single-GPU Abaqus/Standard Linear eqn solver 2x Total Today, release 6.11 FE implicit, single-GPU IMPETUS Afea Explicit solver, SPH 10x SPH, 2x Total Today, release 1.0 FE explicit, multi-GPU LS-DYNA implicit Linear eqn solver 3x Total Planned for 2011 FE implicit, multi-GPU MD Nastran Linear eqn solvers 2x Solver Planned for 2011 FE implicit, multi-GPU Marc Linear eqn solver 1.5x Total Planned for 2011 FE implicit, single-GPU RADIOSS Implicit Linear eqn solver 1.5x Total Demonstration FE implicit, single-GPU PAM-CRASH implicit Linear eqn solver 1.5x Total Demonstration FE implicit, single-GPU NX Nastran Linear eqn solver 1.4x Total Demonstration FE implicit, single-GPU GPU Perf compared against Multi-core x86 CPU socket. GPU Perf benchmarked on GPU supported features and may be a kernel to kernel perf comparison
- 12. Fluid Dynamics Application GPU Features GPU Perf Release Status Notes Altair AcuSolve Linear eqn solver 2x Total Today, release 1.8 FE unstructured NS, multi-GPU Autodesk Moldflow Linear eqn solver 2x Total Today, release 2011 FE unstructured NS, single-GPU FluiDyna LBultra LBM, particle CFD 20x Total Today, release 1.0 Structured LBM, multi-GPU FluiDyna Culises- Linear eqn solvers 3x Solver Today, release 1.0 Unstructured NS, single-GPU OpenFOAM Solver Vratis SpeedIT- Linear eqn solvers 3x Solver Today, release 1.2 Unstructured NS, multi-GPU OpenFOAM Solver Prometech MPS, particle CFD 4x-9x Total Q3CY11 release 2.5 Particle based, multi-GPU Particleworks Sandia NL S3D Chemistry kernel 8x SP, 5x DP kernel Demonstration Structured grid DNS, multi-GPU Turbostream Explicit solver 19x Total Today, release 2.0 Structured grid NS, multi-GPU SD++ (Jameson) Explicit solver 16x Total Planned for 2011 FE unstructured NS, multi-GPU GPU Perf compared against Multi-core x86 CPU socket. FEFLO (Lohner) Explicit solver 2x Total Planned for 2011 FE unstructured NS, multi-GPU GPU Perf benchmarked on GPU supported features and may be a kernel to kernel perf comparison
- 13. Electromagnetics Features Application GPU Perf Release Status Notes Supported Single & multi-GPU; Agilent EMPro FDTD 6X 2011.07 Released EMPro 2011 PR Transient (FIT) 9X on 1 GPU CST Microwave Single & multi-GPU; solver; Combined MPI to 20X+ on 4 2011 Released www.cst.com/perf Studio & GPU computing GPUs Single and multi-GPU; Remcom XFdtd FDTD 30-300X XF7 Released XStream GPU acceleration FDTD; Single and multi-GPU; SPEAG SEMCAD X Acceleware 100X 14.4.3 Released www.speag.com/perf GPU Performance compared against quad-core x86 CPU socket; Remcom XFdtd GPU performance compared against single core CPU
- 14. Climate/ Weather/ Ocean Application GPU Features GPU Perf Production Status Notes WSM5, WSM3, Ice WRF Microphysics models 4x-6x Models Today, release 3.2 single-GPU ASUCA Most routines 12x Total In production at JMA multi-GPU NIM Most routines 7x Dynamics Limited production multi-GPU HIRLAM Dynamical core 3x Solver Planned for 2011 multi-GPU HOMME Models 3x Models Planned for 2011 single-GPU CAM Linear eqn solver 2x Solver Planned for 2011 single-GPU 10x Models, 3x GEOS-5 Most routines Dynamics Demonstration multi-GPU MITgcm Linear eqn solver 3x solver Demonstration single-GPU HYCOM Linear eqn solver 2x solver Demonstration single-GPU GPU Perf compared against Multi-core x86 CPU socket. GPU Perf benchmarked on GPU supported features and may be a kernel to kernel perf comparison