OPTIMIZING NVIDIA VIRTUAL GPU FOR THE BEST VDI USER
OPTIMIZING NVIDIA VIRTUAL GPUFOR THE BEST VDI USER EXPERIENCE
NVIDIA VIRTUAL GPU PRODUCT POSITIONINGNVIDIA GRIDvPC/vAppsNVIDIA QUADRO VirtualData Center WorkstationKnowledge/BusinessWorkerEngineers/ Architects/DesignersTesla M10Tesla P4** Exception High End and Ultra High-End Use Cases
GRID vPC and Quadro vDWSUnderstanding the workflow to define scaleNVIDIA GRID vPC/vAppsScale determined byFramebuffer Size*NVIDIA QUADRO VirtualData Center WorkstationScale determined by3D Engine Performance andFramebuffer Size*All Maxwell and Pascal basedTesla boards provide sufficient3D Performance for typicalGRID vPC workloadsTesla M10 (8GB)8 UsersTesla P40 (24GB)24 UsersEnd-User Latency 200ms 200msFrames/User40004000SPEC ViewPerf12.1Tesla M10 (8GB)1 UserTesla P4 (8GB)1 User 25 80* Tested with Single Full HD Screen. Subject to change with non Pascal and Volta based GPUs3
QUADRO VirtualData Center Workstation9%CPUCPU
P4 provides 11% more Perf than each M60 GPUSPEC ViewPerf 12.1 - Single VM (FRL-Off)Tesla M60TESLA P4 BENEFITSTesla P4Performance1.251Price/Performance0.75Form Factor0.50.25Power Consumption0Tesla M60Tesla P43dsMax11.12Catia11.09Maya11.11Siemens NX11.18Solidworks11.06* Tested on Dell R740 (2x Intel Xeon Gold 6154 CPU @ 3.0 GHz, 18 Cores and is based on geometric mean across 3dsMax, Catia, Maya, Siemens NX and SolidworksPascal Benefits
New Intel CPU allows 6x Tesla P46x Users @ Comparable Guaranteed PerformanceGuaranteed Performance (SPEC ViewPerf 12.1)1x Tesla P46x Tesla P41.25New Intel CPU (3GHz 18c)allows the use of 6x P4s124 VMs4 VMs24 VMs4 VMs12 VMs2 VMs24 VMs4 VMs0.2512 VMs0.52 VMs0.7503dsMaxCatiaMayaSiemens NXSolidworks* Tested on a Dell R740 with 2x Intel Xeon Gold 6154 CPU @ 3.0 GHz, 18 CoresGuaranteed performance isclose to the performance ofsingle P4
P40 provides up to 2.3x more Perf than P4SPEC ViewPerf 12.1 - Single VM (FRL-Off)Tesla P4Tesla P4021.510.50Tesla P4Tesla dical11.8Showcase11.8Siemens NX11.6TESLA P4TESLA P40Many Low-Mid End UsersFew Mid-High End UsersPrice/PerformancePerformanceForm FactorHigh Framebuffer Profiles (12GB and 24GB)Power ConsumptionMultiple Profiles per Server (Many P4s)Solidworks11.2
NVIDIA vGPU Scheduling PoliciesEnterprise CustomersBest Effort SchedulerCloud ServiceProvidersFixed Share Schedulerdefault in Virtual GPU March 2018 Release (6.0)Reason:Maximum utilization of GPU cyclesConsider:Equal Share Scheduler forCompute WorkloadsDelivering Guaranteed QoSReasonReason:Guaranteed QoS – PerformanceGPU resources fenced off per profile
COMPARING THE SCHEDULING MODESA high level summary cheat sheetBEST EFFORTEQUAL SHAREFIXED SHAREMaxwell, sSupportedRecommendedRecommendedIdle cycle redistributionYesNoNoGuaranteed QoSNoYesYesNoisy neighbor protectionNoYesYesFRL requiredYesNoNoSupported HWPrimary Use casesvGPU awareNeeds mixed compute/graphics
Benchmarking Guaranteed PerformanceBenchmarkSynthetic workload (4x Speed)Human workflowHuman workflow (4x Speed)
Start with Guaranteed Performance explore individual scale for each customer during a POCDefining Scale by BenchmarkingDefining Scale with real End Users Same Methodology as Quadro Scale is individual to each customer Familiar Methodology to the customer Allows the Effect of time sharing Guaranteed Performance Can lead to higher scale Conservative Recommendation Performance at higher scale isn’t guaranteed Allows Mapping Quadro boards Leveraging the impact of time sharingrequires Best-Effort Scheduling PolicyP1000 Class Catia Users (SPEC ViewPerf 12.1)*4x Tesla P48 (4x 2)Customer Experience** 12-16 (4x 3-4)* Tested on Dell R740 (2x Intel Xeon Gold 6154 CPU @ 3.0 GHz, 18 Cores
9%GRID vPCCPUCPU
Defining User Experience (UX)Remoted FramesEnd-User LatencyImage QualityDescribes the number of framesthat are sent to the end user.Describes how remote the sessionfeels or how interactive/laggy thesession is.Describes how much the image wasimpacted & manipulated by theremote protocol.FunctionalityConsistencyDescribes if the remote desktopsupports the same range ofapplications (API Support).Describes how much the userexperience varies during the test run.NVIDIA CONFIDENTIAL. DO NOT DISTRIBUTE.
NVIDIA vPC BenchmarkMany User, Many BehaviorsUSER #3Windows Media PlayerMS Word 2016Microsoft Edge (PDF)MS Excel 2016Google Chrome (Web)MS Word 2016Microsoft Edge (PDF)MS Excel 2016Google Chrome (Web)Google Chrome (Video)USER #4Google Chrome (Web)Google Chrome (Video)Windows Media PlayerMS Word 2016Microsoft Edge (PDF) User nGoogle Chrome (Video)Windows Media PlayerMS Word 2016Microsoft Edge (PDF)MS Excel 2016USER #2User 1User 2USER #1Different TimingTimeModern Apps
Horizon 7 Image Quality ImprovementsReferenceimageYUV 4:2:0YUV 4:4:415
Horizon 7 Image Quality ImprovementsReferenceimageYUV 4:2:0YUV 4:4:416
Horizon 7 Image Quality ImprovementsReferenceimageYUV 4:2:0YUV 4:4:417
End User Latency (Click-To-Photon)T2 Timer StopResponseObservedLatency T2 – T1MouseClickT1 Timer Start18
Best End-User Latency with NVIDIA vPCDecrease of 140-160ms for best remoted user experienceVMware Horizon 7.4 (YUV 4:4:4) End-User Latency decrease of 140ms with 1VM End-User Latency decrease of 160ms with 64 VMs19
40% More Remoted Frames with GRID vPCVMware Horizon 7.4 (YUV 4:4:4)20
Up to 25% CPU offload for Highest DensityVMware Horizon 7.4 (YUV 4:4:4)21
TESLA M10 MEETS THE NEEDS OFKNOWLEDGE WORKERSTesla M10 GPU and Encode Engine match the needs of Windows 10Tesla M10 GPU Utilization for32 VMs (8/GPU)Tesla M10 Encoder Utilizationfor 32 VMs (8/GPU)VM Framebuffer UtilizationM10-1BCirrus Knowledge Worker Workload (Excel, Word, PowerPoint, Chrome, Media Player, PDF) with VMware Horizon 7.4 YUV 4:4:422
NVIDIA GRID VGPU FOR HIGHEST DENSITY ANDBEST USER EXPERIENCEBest UserExperienceHighest DensityTesla M10 forWin1023
THANK YOU
3dsMax Catia Creo Energy Maya Medical Showcase Siemens NX Solidworks Tesla P4 1 1 1 1 1 1 1 1 1 Tesla P40 1.3 1.1 2.3 1.9 1.1 1.8 1.8 1.6 1.2 0 0.5 1 1.5 2 SPEC ViewPerf 12.1 - Single VM (FRL-Off) Tesla P4 Tesla P40 TESLA P4 TESLA P40 Many Low-Mid End U
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
GPU Virtualization with Virtual Compute Server NVIDIA Virtual Compute Server enables the benefits of VMware virtualization for GPU-accelerated PowerEdge servers. With Virtual Compute Server, data center admins are able to power compute-intensive workloads with GPUs in a virtual machine (VM). Virtual Compute Server software virtualizes NVIDIA .
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.
