NVIDIA GPU Applications Catalog
33552-print-layout-r2.indd 14/5/21 10:18 AM
Test Drive the World’sFastest Accelerator – Free!Take the GPU Test Drive, a free and easy way toexperience accelerated computing on GPUs. You canrun your own application or try one of the preloadedones, all running on a remote cluster. Try it log-1633552-print-layout-r2.indd 24/5/21 10:18 AM
GPU‑ACCELERATEDAPPLICATIONSAccelerated computing has revolutionized abroad range of industries with over six hundredapplications optimized for GPUs to help youaccelerate your work.CONTENTS1 Computational Finance58 Research: Higher Education and Supercomputing2 Climate, Weather and Ocean ModelingNUMERICAL ANALYTICS2 Data Science and AnalyticsSCIENTIFIC VISUALIZATION5 Artificial IntelligenceDEEP LEARNING AND MACHINE LEARNING12 Public Sector and National Government14 Design for Manufacturing/Construction:CAD/CAE/CAMPHYSICS63 Smart Spaces66 Tools and Management71 Agriculture71 Business Process OptimizationCFD (MFG)CFD (RESEARCH DEVELOPMENTS)COMPUTATIONAL STRUCTURAL MECHANICSDESIGN AND VISUALIZATIONELECTRONIC DESIGN AUTOMATIONINDUSTRIAL INSPECTION27 Media and EntertainmentANIMATION, MODELING AND RENDERINGCOLOR CORRECTION AND GRAIN MANAGEMENTCOMPOSITING, FINISHING AND EFFECTS(VIDEO) EDITING(IMAGE & PHOTO) EDITINGENCODING AND DIGITAL DISTRIBUTIONON-AIR GRAPHICSON-SET, REVIEW AND STEREO TOOLSWEATHER GRAPHICS42 Medical Imaging45 Oil and Gas46 Life SciencesBIOINFORMATICSMICROSCOPYMOLECULAR DYNAMICSQUANTUM CHEMISTRY(MOLECULAR) VISUALIZATION AND 2.indd 34/5/21 10:18 AM
hpc-gpu-apps-catalog-1633552-print-layout-r2.indd 44/5/21 10:18 AM
Computational FinanceCOMPANYNAMEPRODUCT DESCRIPTIONSUPPORTED FEATURESGPU SCALINGAcceleratedComputing EngineAPPLICATION NAMEElsenSecure, accessible, and accelerated backtesting, scenario analysis, risk analyticsand real-time trading designed for easyintegration and rapid development. W eb-like API with Native bindings forPython, R, Scala, C C ustom models and data streamsMulti-GPUSingle NodeAdaptiv AnalyticsSunGardA flexible and extensible engine for fastcalculations of a wide variety of pricing andrisk measures on a broad range of assetclasses and derivatives. C odes in C# supported transparently, withminimal code changes S upports multiple backends includingCUDA and OpenCL S witches transparently between multipleGPUs and CPUS depending on the dealsupport and load factors.Multi-GPUSingle NodeAlea.cuBase F#QuantAleasF# package enabling a growing set of F#capability to run on a GPU. F # for GPU acceleratorsMulti-GPUSingle NodeEstherGlobal ValuationIn-memory risk analytics system for OTCportfolios with a particular focus on XVAmetrics and balance sheet simulations. H igh quality models not admitting closedform solutions E fficient solvers based on full matrix linearalgebra powered by GPUs and Monte CarloalgorithmsMulti-GPUSingle NodeGlobal RiskMISYSRegulatory compliance and enterprise widerisk transparency package. R isk analyticsMulti-GPUSingle NodeHybridizer C#AltimeshMulti-target C# framework for data parallelcomputing. C # with translation to GPU M ulti-Core XeonMulti-GPUSingle NodeMACS AnalyticsLibraryMurexAnalytics library for modeling valuation andrisk for derivatives across multiple assetclasses. M arket standard models for all assetclasses paired with the most efficientresolution methods (Monte Carlosimulations and Partial DifferentialEquations)Multi-GPUSingle NodeNAGNumericalAlgorithmsGroupRandom number generators, Brownianbridges, and PDE solvers M onte Carlo and PDE solversSingle GPUSingle NodeOneviewNumerixNumerix introduced GPU support forForward Monte Carlo simulation for CapitalMarkets and Insurance. E quity/FX basket models withBlackScholes/Local Vol models forindividual equities and FX A lgorithms: AAD (Automatic AlgebraicDifferential) N ew approaches to AAD to reduce timeto market for fast Price Greeks and XVAGreeksMulti-GPUMulti-NodeO-Quant optionspricingO-QuantOffering for risk management and complexoptions and derivatives pricing using GPUs. C loud-based interface to price complexderivatives representing large baskets ofequitiesMulti-GPUMulti-NodePathwiseAon BenfieldSpecialized platform for real-time hedging,valuation, pricing and risk management. S preadsheet-like modeling interfaces P ython-based scripting environment G rid middlewareMulti-GPUSingle NodeSciFinanceSciComp, IncDerivative pricing (SciFinance) M onte Carlo and PDE pricing modelsSingle GPUSingle NodeSynerscope DataVisualizationSynerscopeVisual big data exploration and insight tools G raphical exploration of large networkdatasets including geo-spatial andtemporal componentsSingle GPUSingle NodeVoleraHanweckAssociatesReal-time options analytical engine (Volera) R eal-time analyticsMulti-GPUSingle NodeXcelerit SDKXceleritSoftware Development Kit (SDK) to boostthe performance of Financial applications(e.g. Monte-Carlo, Finite-difference) withminimum changes to existing code. C programming language, crossplatform (back-end generates CUDA andoptimized CPU code) S upports Windows and Linux operatingsystemsMulti-GPUSingle NodePOPULAR GPU‑ACCELERATED APPLICATIONS CATALOG Apr21 1hpc-gpu-apps-catalog-1633552-print-layout-r2.indd 14/5/21 10:18 AM
Climate, Weather and Ocean ModelingCOMPANYNAMEPRODUCT DESCRIPTIONSUPPORTED FEATURESGPU SCALINGCOSMOAPPLICATION NAMECOSMOConsortiumRegional numerical weather prediction andclimate research model R adiation only in the trunk release A ll features in the MCH branch used foroperational weather forecastingMulti-GPUMulti-NodeE3SM-EAMUS DOEGlobal atmospheric model used ascomponent to E3SM global coupled climatemodel. D ynamics and most physicsMulti-GPUMulti-NodeGalesKNMI, TU DelftRegional numerical weather predictionmodel F ull ModelMulti-GPUMulti-NodeGRAFIBM/TWCNew GPU-based global weather modelbased on MPAS from NCAR F ull applicationMulti-GPUMulti-NodeWRF AceCASTWRFTempoQuest Inc.WRF model from NCAR nowcommercialized by TQI. Used for numericalweather prediction and regional climatestudies. All popular aspects of WRF modelare GPU developed. A RW dynamics 1 9 physics options including enough to runthe full WRF model on GPUsMulti-GPUMulti-NodeData Science and AnalyticsCOMPANYNAMEPRODUCT DESCRIPTIONSUPPORTED FEATURESGPU SCALINGAnacondaDistributionAPPLICATION NAMEAnacondaThe open-source Anaconda Distribution isthe easiest way to perform Python/R datascience and machine learning on Linux,Windows, and Mac OS X. Developed for solopractitioners, it is the toolkit that equipsyou to work with thousands of open-sourcepackages and libraries. B indings to CUDA libraries: cuBLAS,cuFFT, cuSPARSE, cuRAND S orts algorithms from the CUB and ModernGPU libraries I ncludes Numba (JIT Python compiler),Dask (Python scheduler), NumPy, SciPy, I ncludes single-line install of numerous DLframeworks such as AnswerRocket leverages AI and machinelearning techniques to automate the hardwork of business analysis, empoweringteams to generate business intelligence andadvanced analysis in seconds. P luggable machine learning models A sk Questions in Plain English C reate Interactive Visualizations &Dashboards P rovides Augmented Analytics S upports a wide variety of data sourcesMulti-GPUMulti-NodeArgusSearchPlanet AIDeep Learning driven document searchtool. F ast full text search engine S earches hand-written and textdocuments, including PDF A llows almost any arbitrary requests(Regular Expressions are supported) P rovides a list of matches sorted byconfidenceMulti-GPUSingle NodeAutomatic SpeechRecognitionCapioIn-house and Cloud-based speechrecognition technologies R eal-time and offline (batch) speechrecognition E xceptional accuracy for transcription ofconversational speech C ontinuous Learning (System becomesmore accurate as more data is pushed tothe platform)Multi-GPUSingle NodeBlazingSQLBlazingSQLGPU-accelerated SQL Engine for analyticsavailable on all major CSP and on-premisedeployment. D istributed SQL Query Engine S upports petabyte scale applications S upports traditional big data formats anddata mory database built on top ofPostgreSQL G PU-Accelerated joins, aggregations,scans, etc. on PostgreSQL V isualization platform bundled withdatabase is called Py (https://github.com/cupy/cupy) isa GPU-accelerated scientific computinglibrary for Python with a NumPy compatibleinterface. CUDA m ulti-GPU supportMulti-GPUSingle Node2 POPULAR GPU‑ACCELERATED APPLICATIONS CATALOG indd 24/5/21 10:18 AM
DatalogueDatalogueAI powered pipelines that automaticallyprepare any data from any source forimmediate & compliant use. D ata transformation O ntology mapping D ata standardization D ata augmentationMulti-GPUSingle NodeDeepGramDeepgramVoice processing solution for call centers,financials and other scenarios. S peech to text and phonetic search usingGPU deep learningMulti-GPUSingle NodeDriverless AIH2O.aiAutomated Machine Learning with FeatureExtraction. Essentially BI for MachineLearning and AI, with accuracy very similarto Kaggle Experts.H2O Driverless AI is an artificial intelligence(AI) platform for automatic machinelearning. Driverless AI automates someof the most difficult data science andmachine learning workflows such asfeature engineering, model validation,model tuning, model selection and modeldeployment. It aims to achieve highestpredictive accuracy, comparable to expertdata scientists, but in much shorter timethanks to end-to-end automation. DriverlessAI also offers automatic visualizations andmachine learning interpretability (MLI).Especially in regulated industries, modeltransparency and explanation are justas important as predictive performance.Modeling pipelines (feature engineering andmodels) are exported (in full fidelity, withoutapproximations) both as Python modules andas pure Java standalone scoring artifacts. A utomated machine learning and featureextraction A utomated statistical visualization I nterpretability toolkit for machine learningmodelsMulti-GPUSingle NodeGPUdbKineticaMulti-GPU, Multi-Machine distributedobject store providing SQL style querycapability, advanced geospatial querycapability,heatmap generation, anddistributed rasterization services. Q uery against big data in real time N o pre-indexing allows for complex, ad-hocquery chains I nteractively explore large, streamingdata setsMulti-GPUSingle NodeH2O4GPUH2O.aiH2O is a popular machine learningplatform which offers GPU-acceleratedmachine learning. In addition, they offerdeep learning by integrating popular deeplearning frameworks. A vailable algorithms include GradientBoosting Machines (GBM’s) G eneralized Linear Models (GLM’s) K -Means Clustering SVD PCA K-means X GBoost. I t can be used as a drop-in replacementfor scikit-learn with support for GPUs onselected (and ever-growing) algorithms. A new R API brings the benefits of GPUaccelerated machine learning to the R usercommunity. The R package is a wrapperaround the H2O4GPU Python package,and the interface follows standard Rconventions for modeling.Multi-GPUSingle NodeIntelligentVoiceINTELLIGENTVOICEFar more than a transcription tool, thisspeech recognition software learnswhat is important in a telephone call,extracts information and stores a visualrepresentation of phone calls to becombined with text/instant messaging andE-mail. Intelligent Voice’s search and alertmakes it possible to tackle issues beforethey arise, address data security concernsand monitor physical access to data. A dvanced Speech Recognition across largedata sets J umpTo Technology, for data visualisation E-Discovery E xtraction from phone calls I M & Email defining key phrases andemotional analysis C ompliance, defining key conversationsand interactionsMulti-GPUSingle NodePOPULAR GPU‑ACCELERATED APPLICATIONS CATALOG Apr21 3hpc-gpu-apps-catalog-1633552-print-layout-r2.indd 34/5/21 10:18 AM
JedoxJedoxHelps with portfolio analysis, managementconsolidation, liquidity controlling, cashflow statements, profit center accounting,treasury management, customer valueanalysis and many more applications. Allaccessible in a powerful web and mobileapplication or Excel environment. T his database holds all relevant data inGPU memory T esla K40 &12 GB on-board RAM S cales up with multiple GPUs K eeps close to 100 GB of compressed datain GPU memory on a single server system F ast analysis, reporting, and planningMulti-GPUSingle NodeLabellioKYOCERACommunicationSystems CoThe world’s easiest deep learning webservice for computer vision, allowingeveryone to build own image classifier withonly web browser. N eural net fine-tuning for image data D ata crawling and data browsing D rag-and-drop style data cleansing backedby AI supportMulti-GPUSingle NodeNumbaAnacondaNumba is an open source JIT compiler thattranslates a subset of Python and NumPycode into fast machine code.Think of it as a compiler for Python arrayand numerical functions that gives you thepower to speed up your applications withhigh performance functions written directlyin Python.Numba translates Python functions tooptimized machine code at runtime usingthe industry-standard LLVM compiler library.Numba-compiled numerical algorithmsin Python can approach the speeds of C orFORTRAN.You don’t need to replace the Pythoninterpreter, run a separate compilation step,or even have a C/C compiler installed. Justapply one of the Numba decorators to yourPython function, and Numba does the rest.Numba generates optimized machine codefrom pure Python code using the LLVMcompiler infrastructure. With a few simpleannotations, array-oriented and math-heavyPython code can be just-in-time optimized toperformance similar as C, C and Fortran,without having to switch languages or Pythoninterpreters.Numba is designed to be used with NumPyarrays and functions. Numba generatesspecialized code for different array data typesand layouts to optimize performance. Specialdecorators can create universal functionsthat broadcast over NumPy arrays just likeNumPy functions do.Numba also works great with Jupyternotebooks for interactive computing, andwith distributed execution frameworks,like Dask and Spark. With support for GPUacceleration, Numba lets you write parallelGPU algorithms entirely from Python. O n-the-fly code generation (at import timeor runtime, at the user’s preference) N ative code generation for the CPU(default) and GPU hardware I ntegration with the Python scientificsoftware stack (enabled via Numpy) J IT compilation of Python functions forexecution on various targets (includingCUDA)Multi-GPUSingle NodeOmniSciOmniSciOmniSci is GPU-powered big data analyticsand visualization platform that is hundredsof times faster than CPU in-memorysystems. OmniSci uses GPUs to executeSQL queries on multi-billion row datasetsand optionally render the results, all inmilliseconds. U ses LLVM’s nvptx backend to generateCUDA kernels O penGL- (EGL) based rendering C an run in a docker container usingNVIDIA-dockerMulti-GPUSingle NodePolymaticaPolymaticaAnalytical OLAP and Data Mining Platform V isualization, Reporting, OLAP in-memorywith GPU acceleration D ata Mining M achine Learning P redictive AnalyticsMulti-GPUMulti-Node4 POPULAR GPU‑ACCELERATED APPLICATIONS CATALOG indd 44/5/21 10:18 AM
Sqream DBSQreamTechnologiesGPU accelerated SQL database engine forbig data analytics. Sqream speeds SQLanalytics by 100X by translating SQL queriesinto highly parallel algorithms run on theGPU. U p to 100TB of raw data can be stored andqueried in a standard 2U server I nserts and analyzes hundreds of billions ofrecords in seconds N o indexes required N o changes to SQL code or data scienceparadigms requiredMulti-GPUSingle NodeSynerScopeSynerscopeBig data visualization and data discovery, forcombining Analytics on Analytics with IoTcompute-at-the-edge smart sensors. R eal-time Interaction with dataSingle GPUSingle NodeFinancial analytics and data mining library M onte Carlo simulations P ricing of vanilla and exotic options F ixed income analytics D ata miningMulti-GPUSingle NodeZX Lib (Fuzzy Logic) TanayArtificial IntelligenceDEEP LEARNING AND MACHINE LEARNINGCOMPANYNAMEPRODUCT DESCRIPTIONSUPPORTED FEATURESGPU SCALINGAICAPPLICATION NAMETracxpointAIC (Artificial Intelligence Cart)revolutionizes the supermarket shoppingexperience with sensor fusion and machinelearning technology. T he smart IoT cart recognizes the shopper,loads their shopping list and buyingpatterns, suggests compatible productsand provides the most valuable offer R ecognizes the items placed in the cart andbill the customer at the end of the shoppingexperience with no checkout lanes F eature JetpackSingle GPUSingle NodeAiFi NanoAiFi Inc.Cashier-free (like Amazon grab and gosolution) and stock out retail software cuDNN TensorRT DeepStreamMulti-GPUSingle NodeAI Image LabelingFrenzyBuilds robust self-labeling training datasetsfor classifying exact objects and products invisual scenes at a fraction of the time andcost G PU in the cloudMulti-GPUSingle NodeAI LifescycleClarifaiClarifai brings a new level of understandingto visual content through deep learningtechnologies. Uses GPUs to train largeneural networks to solve practicalproblems in advertising, media, and searchacross a wide variety of industries suchas automated tagging, visual search,and recommendation engine, predictivemaintenance, demographic analysis andmore. G PU-based training and inference R ecognizes and indexes images withpredefined classifiers or custom classifiersMulti-GPUSingle NodeAllganize NLU APIsfor EnterprisesAllganize, Inc.Natural Language Understanding APIs forenterprise: Answer-bot based on documentswith unstructured data (text table), e.g.,manuals, instructions, FAQ documents;Review analysis; sentiment analysis,summarizing etc. Provided as APIs. T raining and inferencing using V100Multi-GPUMulti-NodeAlphaSenseAlphaSensePaaS for Financial analysis based on publiccorporate information. Geared at financialanalysts within financial services. Allowsvery fast searches of public corporateinformation, and allows questing answeringformat (“the Google for Analyst research”) P aaS for Financial analysis based on publiccorporate information G eared at financial analysts within financialservices. A llows very fast searches of publiccorporate information, and allows questinganswering format (“the Google for Analystresearch”)Multi-GPUSingle NodeAlwaysAIAlways AIEasy-to-use platform to build and deploycomputer vision applications for embeddeddevices at the edge. Apply for an earlyaccess on the product link J etson NanoSingle GPUSingle NodePOPULAR GPU‑ACCELERATED APPLICATIONS CATALOG Apr21 5hpc-gpu-apps-catalog-1633552-print-layout-r2.indd 54/5/21 10:18 AM
AnacondaEnterprise EditionAnacondaThe end-to-end data science platform.The Anaconda enterprise platform isa comprehensive foundation for anyorganization that wants to use data scienceand machine learning to make betterdecisions and build differentiating solutions. B indings to CUDA libraries: cuBLAS,cuFFT, cuSPARSE, cuRAND S orts algorithms from the CUB and ModernGPU libraries N umba (JIT Python compiler), Dask (Pythonscheduler), NumPy, SciPy, S ingle-line install of numerous DLframeworks such as PYTORCHMulti-GPUSingle NodeAntuit DemandPlanning andForecastingAntuitExtracts maximum predictability fromthe available data. Proprietary “DynamicAggregation” logic with attribute-baseddisaggregation generates forecasts forall products, including new, slow-moving,and end-of-life. Spark and GPU clusters,along with optimized AI algorithms
(video) editing (image & photo) editing encoding and digital distribution on-air graphics on-set, review and stereo tools weather graphics 46 medical imaging 49 oil and gas 50 life sciences bioinformatics microscopy molecular dynamics quantum chemistry (molecular) visualization and docking
NVIDIA virtual GPU products deliver a GPU Experience to every Virtual Desktop. Server. Hypervisor. Apps and VMs. NVIDIA Graphics Drivers. NVIDIA Virtual GPU. NVIDIA Tesla GPU. NVIDIA virtualization software. CPU Only VDI. With NVIDIA Virtu
NVIDIA vCS Virtual GPU Types NVIDIA vGPU software uses temporal partitioning and has full IOMMU protection for the virtual machines that are configured with vGPUs. Virtual GPU provides access to shared resources and the execution engines of the GPU: Graphics/Compute , Copy Engines. A GPU hardware scheduler is used when VMs share GPU resources.
www.nvidia.com GRID Virtual GPU DU-06920-001 _v4.1 (GRID) 1 Chapter 1. INTRODUCTION TO NVIDIA GRID VIRTUAL GPU NVIDIA GRID vGPU enables multiple virtual machines (VMs) to have simultaneous, direct access to a single physical GPU, using the same NVIDIA graphics drivers that are
NVIDIA PhysX technology—allows advanced physics effects to be simulated and rendered on the GPU. NVIDIA 3D Vision Ready— GeForce GPU support for NVIDIA 3D Vision, bringing a fully immersive stereoscopic 3D experience to the PC. NVIDIA 3D Vision Surround Ready—scale games across 3 panels by leveraging
NVIDIA GRID K2 1 Number of users depends on software solution, workload, and screen resolution NVIDIA GRID K1 GPU 4 Kepler GPUs 2 High End Kepler GPUs CUDA cores 768 (192 / GPU) 3072 (1536 / GPU) Memory Size 16GB DDR3 (4GB / GPU) 8GB GDDR5 Max Power 130 W 225 W Form Factor Dual Slot ATX, 10.5” Dual Slot ATX,
Virtual GPU Software Client Licensing DU-07757-001 _v13.0 3 NVIDIA vGPU Software Deployment Required NVIDIA vGPU Software License Enforcement C-series NVIDIA vGPU vCS or vWS Software See Note (2). Q-series NVIDIA vGPU vWS Software See Note (3). GPU pass through for workstation or professional 3D graphics vWS Software
RTX 3080 delivers the greatest generational leap of any GPU that has ever been made. Finally, the GeForce RTX 3070 GPU uses the new GA104 GPU and offers performance that rivals NVIDIA’s previous gener ation flagship GPU, the GeForce RTX 2080 Ti. Figure 1.
transplant a parallel approach from a single-GPU to a multi-GPU system. One major reason is the lacks of both program-ming models and well-established inter-GPU communication for a multi-GPU system. Although major GPU suppliers, such as NVIDIA and AMD, support multi-GPUs by establishing Scalable Link Interface (SLI) and Crossfire, respectively .
