NVIDIA CLOUDXR ™ David Weinstein, 19 December 2019
Gaming Sports PERSONAL ENTERTAINMENT VR Movies Concerts Travel Retail 2
QUADRO PROFESSIONAL VR/AR MEDICAL ARCHITECTURE DIGITAL CONTENT CREATION DESIGN MANUFACTURING COLLABORATION NVIDIA CONFIDENTIAL. DO NOT DISTRIBUTE. 3
4
5
6
FUNDAMENTAL VR GRAPHICS CHALLENGES Pixel Throughput and Perceived Latency Very High Pixel Rendering Throughput Must Maintain Very Low Latency TRADITIONAL = 60 MP/S ( 1920 X 1080 @ 30 FPS) HMD updates pose 1080 CPU triggers redraw 1920 VIRTUAL REALITY = 450 MP/S ( 3024 X 1680* @ 90 FPS) GPU computes new image 1680 HMD flashes display Motion to Photon: ≤ 20 ms 1512 1512 7
WANTED: THE BEST OF BOTH WORLDS Highest Quality Graphics + Mobile Freedom 8
CLOUDXR VISION XR Streaming to Any Device Network Content RTX Servers Clients 9
MOBILE WORLD CONGRESS (LA), OCT 2019 CloudXR Announcement 10
ARCHITECTURE OVERVIEW Server-Side Driver, Client-Side Application and SDK audio, video, haptics pose, control inputs, video VR App HMD OpenVR Runtime (Valve) Late & Lens Warp VR Runtime Virtual HMD Driver Virtual Audio Driver Audio Decoder Video Encoder Audio Encoder Video Decoder Cloud Cloud Internet CLIENT app (Nvidia) SERVER driver (Nvidia) 11
ARCHITECTURE CLIENT WIN SERVER VR App Controller Display HMD & inputs (e.g. ROI, optional) RGB Frame Runtime API App Video Params VR Runtime Lens Warp VR Runtime ( e.g. , OpenVR) Time Warp Driver API OpenVR Interfaces Sensor, RGB Frame INSTALLER Pose & Controller Data RGB Frame Sensor & input data Encode Decode Server driver Client app HEVC Frame Sensor & input data HEVC Frame Audio Capture / Encode Audio Playback Audio Data (sensors, inputs, QoS) Audio/Video 12
QOS Reliability & Resilience Relevant Internet problems: Jitter • • Missed packets Bandwidth variation • 13
QOS Reliability & Resilience Challenges Solutions Control parameters Latency Custom HEVC profile NVENC knobs Jitter De-jitter buffer Buffer size Missed packets Forward error correction Correction strength Bandwidth variation Video compression Compression rate Server has no display Frame pacing to match Server render rate client display rate 14
ARCHITECTURE CLIENT WIN SERVER VR App Controller Display HMD & inputs (e.g. ROI, optional) Runtime API App Video Params VR Runtime Lens Warp VR Runtime ( e.g. , OpenVR) Time Warp Driver API OpenVR Interfaces INSTALLER Sensor, Pose & Controller data Jitter FIFO RGB Frame Sensor & input data Encode Decode Server driver Client app Frame Pacing HEVC Frame Audio Playback Audio Capture / Encode Audio Audio / Video / Data (sensors, inputs, QoS) 15
LATENCY Button to Photons Using: • HelloVR sample modified to change color on button press Latency meter (instrumented switch + photodiode) • • Local Cloud with Client on WiFi (average ping time 4 ms) 16
LATENCY Button to Photons Results: Latency [ms] Stdev Local VR 38.3 3.4 CloudXR 81.6 3.3 Diff 43.3 -0.1 Jitter FIFO + Ping 31.2 Diff – Above 12.1 Server Driver + Client App latency = 12.1 ms. The Jitter FIFO can be downsized at the cost of occasional missed frames. 17 *Average of 40 clicks
LATENCY CLIENT WIN SERVER VR App Controller Display HMD & inputs (e.g. ROI, optional) Runtime API App Video Params VR Runtime Lens Warp VR Runtime ( e.g. , OpenVR) Time Warp Driver API 1-3 ms OpenVR Interfaces < 0.5 ms 10-50 ms INSTALLER (depending on network) Jitter FIFO RGB Frame Sensor & input data RGB Frame 3-10 ms Encode Decode 3-10 ms Server driver Client app Frame Pacing HEVC Frame Audio Playback Audio Capture / Encode Audio Audio / Video / Data (sensors, inputs, QoS) 18
19
CLOUDXR CONFIGURATIONS Total Flexibility 5G Ethernet WiFi 20
EARLY ACCESS PROGRAM Test CLOUDXR for VR @ https://developer.nvidia.com/nvidia-cloudxr-sdk-early-access- program 21
NVIDIA 5G MEC XR DEVKIT 22
THE OPPORTUNITY: VR/AR/VGPU OVER 5G A Convergence of Visual Compute + 5G and a New Edge Topology Powerful GPUs Bandwidth and Low Latency Compute is at the Edge Incredible compute capabilities redefining Both are required for VR/AR Very few hops – Low Latency graphics 23
ENTERPRISE MARKETS Success Across Verticals ESI Enscape Training AEC Possibly the Largest Entire Design Process, Market Sales, Moving to IC.IDO Nurulize Construction M&E Virtual Production Autodesk Manufacturing Location-Based Entertainment 24 Engineering, Design, Marketing/Sales Enterprise/Consumer
ENTERPRISE AR Driving Value Propositions INSTRUCTION/COLLABORATION DIGITAL MODELS ANNOTATED 25
THE FUTURE OF MOBILE XR 5G + VR/AR + vGPU at the Edge Requirements: • Ultra Low Latency • High Bandwidth • Cutting-edge Compute • Cost Effective Delivery • Valuable Use Cases 26
NVIDIA RTX SERVER RTX Performance in a Server • Powered by Quadro RTX 8000/6000 • Ray Traced Global Illumination up to 96 GB Scenes • Remoting and Multi-GPU Virtualization with Quadro virtual Datacenter Work Station (vDWS) • All the technical advantages of Turing 27
XR/ V GPU 5G EDGE STACK Software/Hardware Stack Demands a Diverse Effort Containerized Edge Applications 5G transmission Enabling SDKs Low Latency Streaming Fundamental GPU Drivers Virtualization 5G Network Visual Computing and AI Cloud Orchestration 28
COME TRY THE DEMO & REGISTER ONLINE FOR EARLY ACCESS https://developer.nvidia.com/nvidia- Visit the VR Village on the Expo Floor cloudxr-sdk-early-access-program 29
Recommend
More recommend