1 / 56

GTC Europe 2017 Keynote

NVIDIA founder and CEO Jensen Huang took the stage in Munich — one of the hubs of the global auto industry — to introduce a powerful new AI computer for fully autonomous vehicles and a new VR application for those who design them.

nvidia
Download Presentation

GTC Europe 2017 Keynote

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. A NEW COMPUTING ERA J ensen Huang, Founder & CEO, GTC Europe 2017

  2. TWO FORCES DRIVING THE FUTURE OF COMPUTING Transistors (thousands) 107 1.1X per year 105 103 1.5X per year Single-threaded perf 1980 1990 40 Years of CPU Trend Data 2000 2010 2020 Original data up to the year 2010 collected and plotted by M. Horowitz, F. Labonte, O. Shacham, K. Olukotun, L. Hammond, and C. Batten New plot and data collected for 2010-2015 by K. Rupp 2

  3. TWO FORCES DRIVING THE FUTURE OF COMPUTING Transistors (thousands) 107 1.1X per year 105 103 1.5X per year Single-threaded perf 1980 1990 40 Years of CPU Trend Data 2000 2010 2020 The Big Bang of Deep Learning Original data up to the year 2010 collected and plotted by M. Horowitz, F. Labonte, O. Shacham, K. Olukotun, L. Hammond, and C. Batten New plot and data collected for 2010-2015 by K. Rupp 3

  4. RISE OF NVIDIA GPU COMPUTING GPU-Computing perf 1.5X per year 107 1.1X per year 105 103 1.5X per year Single-threaded perf 1980 1990 40 Years of CPU Trend Data 2000 2010 2020 NVIDIA Volta GPU | 21B xtors | 120 TFLOPS Original data up to the year 2010 collected and plotted by M. Horowitz, F. Labonte, O. Shacham, K. Olukotun, L. Hammond, and C. Batten New plot and data collected for 2010-2015 by K. Rupp 4

  5. RISE OF NVIDIA GPU COMPUTING 22,000 645,000 1.8M 2012 2012 2017 2012 2017 2017 Global GTC Attendees 10X in 5 Years GPU Developers 15X in 5 Years CUDA Downloads 5X in 5 Years 5

  6. NVIDIA GPU-ACCELERATED APPLICATIONS Computer Graphics Scientific Computing Deep Learning Data Science SIMULIA Abaqus Finite-Element Analysis AMBER Molecular Dynamics K-Means Clustering Gradient Boosting COSMO Climate Weather ChaNGa Astrophysics Support Vector Machine Generalized Linear Model ANALYTICS Gaussian Quantum Chemistry Schlumberger WG Seismic Processing ANSYS Fluent Computational Fluid Dynamics DATABASES PowerGrid Medical Imaging NVIDIA CUDA 6

  7. NVIDIA GPU-ACCELERATED INDUSTRIES Weather Photo Album Ride Sharing Radiology AV Truck Molecular Dynamics Volumetric Rendering Route Management Recommendations Self-Driving Cars Design & Simulation Warehouse Automation Materials Science Social Media 3D CT HPC INTERNET SERVICES TRANSPORTATION MEDICAL IMAGING LOGISTICS 7

  8. NVIDIA GPU ACCELERATES 2017 NOBEL PRIZES IN CHEMISTRY AND PHYSICS Resolution before 2013 Resolution at present Cryogenic Electron Microscopy J acques Dubochet, J oachim Frank, Richard Henderson Detection of Gravitational Waves Rainer Weiss, Barry Barish, Kip Thorne 8

  9. ANNOUNCING NVIDIA HOLODECK THE DESIGN LAB OF THE FUTURE Photorealistic Models Physically Simulated Interaction Virtual Team Collaboration GPU-accelerated AI 9

  10. CREATE AND COLLABORATE IN NVIDIA HOLODECK CATIA / Siemens NX Creo / Alias Maya / 3dsMAX 10

  11. 11

  12. ANNOUNCING NVIDIA HOLODECK THE DESIGN LAB OF THE FUTURE Photorealistic Models Physically Simulated Interaction Virtual Team Collaboration GPU-accelerated AI Early Access NOW nvidia.com/holodeck 12

  13. THE ERA OF AI 2017 6000 4500 $6.6B 3,000 2016 3000 1500 2002 2012 2017 2014 2017 -50 -25 0 25 AI Startup Funding 12X in 5 Years Deep Learning Papers Published 13X in 3 Years NIPS Registration 13

  14. SOLVING THE UNSOLVABLE NVIDIA Interactive Ray Tracing 14

  15. SOLVING THE UNSOLVABLE NVIDIA NVIDIA / Remedy Audio-driven Facial Animation Interactive Ray Tracing 15

  16. SOLVING THE UNSOLVABLE NVIDIA NVIDIA / Remedy Audio-driven Facial Animation WRNCH Interactive Ray Tracing Pose Estimation 16

  17. SOLVING THE UNSOLVABLE NVIDIA NVIDIA / Remedy Audio-driven Facial Animation WRNCH Interactive Ray Tracing Pose Estimation University of Edinburgh Character Animation 17

  18. SOLVING THE UNSOLVABLE NVIDIA NVIDIA / Remedy Audio-driven Facial Animation WRNCH Interactive Ray Tracing Pose Estimation University of Edinburgh Character Animation UC Berkeley / OpenAI One-shot Imitation Learning 18

  19. THE WORLD’S AI PLATFORM Every Framework NVIDIA Inception: 1,900 DL Startups Every Cloud and Data Center Other IT Services Fin. Services Drones Manufacturing Automotive Smart City Healthcare NVIDIA AI PLATFORM 19

  20. AI INFERENCE IS THE NEXT GREAT CHALLENGE Training DNN Model Inferencing 20

  21. EXPLOSION OF INTELLIGENT MACHINES 20M Inference Servers 100s of Millions of Autonomous Machines Trillions of IoT Devices 21

  22. EXPLOSION OF NETWORK DESIGN Convolution Networks Recurrent Networks Generative Adversarial Networks Reinforcement Learning 3D-GAN PRelu ReLu BatchNorm LSTM GRU Highway Rank GAN Conditional GAN DQN Dueling DQN Coupled GAN Concat Dropout Pooling Projection Embedding BiDirectional Speech Enhancement GAN Latent space GAN A3C 22

  23. EXPLOSION OF NETWORK COMPLEXITY 30X 10X 350X MoE DeepSpeech 3 Inception-v4 GNMT ResNet-50 DeepSpeech 2 OpenNMT DeepSpeech AlexNet GoogLeNet Inception-v2 2014 2015 2016 2017 2018 2011 2013 2015 2017 2013 2014 2015 2016 2017 2018 2012 2014 2016 Image Network Complexity GOPS * Bandwidth Speech Network Complexity GOPS * Bandwidth Translation Network Complexity GOPS * Bandwidth 23

  24. NEW NVIDIA TENSORRT 3 Programmable Inference Accelerator TESLA P4 J ETSON TX2 TensorRT DRIVE PX 2 NVIDIA DLA TESLA V100 Compile and Optimize Neural Networks | Support for Every Framework Optimize for Each Target Platform 24

  25. NEW NVIDIA TENSORRT 3 Programmable Inference Accelerator next input concat next input relu relu relu relu batch nm batch nm batch nm batch nm copy 3x3 CR 5x5 CR 1x1 CR 1x1 conv 3x3 conv 5x5 conv 1x1 conv 1x1 CR max pool relu relu max pool batch nm batch nm input 1x1 conv 1x1 conv input Weight & Activation Precision Calibration | Layer & Tensor Fusion Kernel Auto-Tuning | Multi-Stream Execution 25

  26. NEW NVIDIA TENSORRT 3 Programmable Inference Accelerator Images/Sec (ResNet-50) Sentences/Sec (OpenNMT) 5,700 600 550 6,000 450 4,500 300 3,000 150 1,500 300 25 140 4 - CPU + TensorFlow CPU + TensorFlow 0 V100 + TensorFlow V100 + TensorFlow V100 + TensorRT V100 + TensorRT CPU + Torch V100 + Torch V100 + TensorRT 40x Speed-up on ResNet-50 | 140x Speed-up on OpenNMT 26

  27. NEW NVIDIA TENSORRT 3 Programmable Inference Accelerator Images/Sec (ResNet-50) Sentences/Sec (OpenNMT) 600 6,000 450 4,500 14ms 280ms 300 3,000 153ms 7ms 7ms 150 1,500 117ms - CPU + TensorFlow CPU + TensorFlow 0 V100 + TensorFlow V100 + TensorFlow V100 + TensorRT V100 + TensorRT CPU + Torch V100 + Torch V100 + TensorRT 40x Speed-up on ResNet-50 | 140x Speed-up on OpenNMT | Half the Latency 27

  28. NVIDIA TENSORRT 10X BETTER DATA CENTER TCO 160 CPU servers 45,000 images / second 65 KWatts 28

  29. NVIDIA TENSORRT 10X BETTER DATA CENTER TCO 1 NVIDIA HGX with 8 Tesla V100 GPUs 45,000 images / second 3 KWatts 1/6 the Cost | 1/20 the Power 4 Racks in a Box 29

  30. INFERENCE ON IMAGES 30

  31. 31

  32. INFERENCE ON SPEECH 32

  33. 33

  34. VINCENT AI 34

  35. 35

  36. THE AUTONOMOUS VEHICLE REVOLUTION 36

  37. NVIDIA DRIVE AV COMPUTING PLATFORM LIDAR Camera Deep Learning (Detection) Camera Deep Learning (Freespace) HD Map Camera Computer Vision (SLAM) RADAR Fusion Path Planning Point-Cloud Processing (Localizing to Map) DRIVE AV DRIVEWORKS SDK DRIVE OS DRIVE PX —AI CAR COMPUTER Sensor Fusion: RADAR, LIDAR, Camera | Deep Learning, CV, Parallel Computing Diversity of Algorithms | ASIL-D Functional Safety | Fully Integrated into NVIDIA BB8 37

  38. 38

  39. “ RIDE-HAILING INDUSTRY EXPECTED TO GROW EIGHTFOLD TO $285B BY 2030” —Goldman Sachs Uber Lyft Waymo GM / Cruise Ford / Argo.ai Baidu Zoox nuTonomy Yandex 39

  40. ROBOTAXI DEMANDS EXTREME COMPUTING + 10X camera resolution + Surround LIDAR point-cloud processing + Camera & LIDAR localization to HD map + Tracking all surrounding objects + New map generation + Sophisticated path planning & control + Algorithm diversity + Sensor & computing fail-operate + Excess computing capacity Level 2 Level 5 40

  41. STATE-OF-THE-ART DRIVERLESS VEHICLES 41

  42. ANNOUNCING “ PEGASUS” ROBOTAXI DRIVE PX 320 TOPS CUDA TensorCore | 16x GMSL | 4x 10G | 8x 1G | 16x 100M | Auto-grade | ASIL D 500W | Late Q1 Early Access Partners Supercomputing Data Center in your Trunk 42

  43. ANNOUNCING “ PEGASUS” ROBOTAXI DRIVE PX 320 TOPS CUDA TensorCore | 16x GMSL | 4x 10G | 8x 1G | 16x 100M | Auto-grade | ASIL D 500W | Late Q1 Early Access Partners Supercomputing Data Center in your Trunk 43

  44. STATE-OF-THE-ART DRIVERLESS VEHICLES 44

  45. ANNOUNCING “ PEGASUS” ROBOTAXI DRIVE PX 320 TOPS CUDA TensorCore | 16x GMSL | 4x 10G | 8x 1G | 16x 100M | Auto-grade | ASIL D 500W | Late Q1 Early Access Partners Supercomputing Data Center in your Trunk 45

  46. ANNOUNCING NVIDIA DRIVE IX SDK Intelligent Experience Toolkit Exterior Driver Recognition Automatic Personalization Inattentive Driver Alert Cyclist Alert Distracted Driver Alert Driver/Passenger Recognition Customer Application DRIVE AV DRIVE IX Object, Path, Wait Perception Gaze, Head Pose, Gestures, Recognize Face, Voice Recognition & Lip Reading DRIVE OS Sense Inside & Outside the Vehicle | Deep Learning Powered | Early Access Q4 Your Car is an AI 46

  47. NVIDIA DRIVE AI CAR PLATFORM SIL HIL Data Factory Training on DGX 3D-SIM on DGX DRIVE AV on DPX RE-SIM on DGX From Training to Driving | From Processor to AI to AV Stack | From AV to IX From 3D Simulation to Super-Real-Time Testing | Open Platform for Partners 47

  48. SUPER REAL-TIME SIMULATION WITH DGX AND NEW TENSORRT 3 With 8 NVIDIA DGX Re-Sim 300,000 miles in 5 hours Virtually Drive Every Paved Road in U.S. in 2 Days 48

  49. 49

  50. 145 AV STARTUPS ON NVIDIA DRIVE 50

More Related