Ai Forum at Computex 2017 - Keynote Slides by Jensen Huang
7 likes1,307 views
Nvidia CEO Jensen Huang discusses the transformative impact of GPU computing and deep learning at Computex 2017, highlighting trends in microprocessor performance and the rise of AI applications. With significant advancements in GPU technology, including the introduction of the Tesla V100, Nvidia aims to drive AI innovation across various fields such as autonomous vehicles and robotics. The presentation also emphasizes Nvidia's commitment to cloud services and optimized solutions for deep learning, fostering a future dominated by intelligent machines.
![NEW TENSOR CORE
New CUDA TensorOp instructions & data formats
4x4 matrix processing array
D[FP32] = A[FP16] * B[FP16] + C[FP32]
Optimized for deep learning
Activation Inputs Weights Inputs Output Results](https://image.slidesharecdn.com/jhhcomputex2017final-170607072927/85/Ai-Forum-at-Computex-2017-Keynote-Slides-by-Jensen-Huang-17-320.jpg)
Ai Forum at Computex 2017 - Keynote Slides by Jensen Huang
- 1. POWERING THE AI REVOLUTION JENSEN HUANG, FOUNDER & CEO, NVIDIA | Computex 2017
- 2. TWO FORCES DRIVING THE FUTURE OF COMPUTING The Big Bang of Deep Learning 1980 1990 2000 2010 2020 40 Years of Microprocessor Trend Data 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 Transistors (thousands) 102 103 104 105 106 107 Single-threaded perf 1.5X per year 1.1X per year
- 3. RISE OF GPU COMPUTING APPLICATIONS SYSTEMS ALGORITHMS CUDA 1980 1990 2000 2010 2020 102 103 104 105 106 107 Single-threaded perf 1.5X per year 1.1X per year GPU-Computing perf 1.5X per year 1000X by 2025 40 Years of Microprocessor Trend Data 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. RISE OF GPU COMPUTING 0 8 16 24 32 40 P100 2016 K80 2015 K40 2014 K20 2013 AMBER 12 CUDA 4 AMBER 14 CUDA 5 AMBER 14 CUDA 6 AMBER 16 CUDA 8 AMBER Performance (ns/day) 0 2400 4800 7200 9600 12000 cuDNN 2 CUDA 6 cuDNN 4 CUDA 7 cuDNN 6 CUDA 8 NCCL 1.6 cuDNN 7 CUDA 9 NCCL 2 8x K80 2014 8x Maxwell 2015 DGX-1 2016 DGX-1V 2017 Googlenet Performance (i/s)
- 5. RISE OF GPU COMPUTING GPU Developers 11X in 5 Years 2017 511,000 2012 2016 $5B 2012 AI Startups Investments 9X in 4 Years
- 6. NEW PROJECT HOLODECK Photorealistic models Interactive physics Collaboration Early access in September
- 7. GEFORCE GTX — THE WORLD’S LEADING GAMING PLATFORM New GeForce Cards New VR Backpacks External GeForce Consoles New G-SYNC Monitors New GeForce Laptops
- 8. 0 15 30 45 2014 2015 2016 0 2 4 6 2014 2015 2016 GREAT DESIGN AND PERFORMANCE Max-Q Max-Q Thinness Performance Avg.Height(mm)ofInstalledBase Avg.PerformanceofInstalledBase
- 11. MARVELS OF MAX-Q Powerful GPUs @ Peak Efficiency Optimal Game Settings Advanced Thermal Solutions Next-Gen Regulator Efficiency
- 13. THE ERA OF AI AlphaGo Beats the World’s Best“OK, Google”Self-Driving Cars Hit the Road
- 14. MODEL COMPLEXITY IS SKYROCKETING 2017 — Google NMT 105 ExaFLOPS 8.7 Billion Parameters 2016 — Baidu Deep Speech 2 20 ExaFLOPS 300 Million Parameters 2015 — Microsoft ResNet 7 ExaFLOPS 60 Million Parameters
- 15. POWERING THE AI REVOLUTION NVIDIA DEEP LEARNING SDK GPU AASGPU SYSTEMSFRAMEWORKS AUTONOMOUS VEHICLES ROBOTS NVIDIA DEEP LEARNING SDK AI in the Cloud HGX DRIVE PX JETSON TX AI at the Edge
- 16. NEW TESLA V100 GIANT LEAP FOR AI & HPC VOLTA WITH NEW TENSOR CORE 21B xtors | TSMC 12nm FFN | 815mm2 5,120 CUDA cores 120 TFLOPS 20MB SM RF | 16MB Cache | 16GB HBM2 @ 900 GB/s 300 GB/s NVLink
- 17. NEW TENSOR CORE New CUDA TensorOp instructions & data formats 4x4 matrix processing array D[FP32] = A[FP16] * B[FP16] + C[FP32] Optimized for deep learning Activation Inputs Weights Inputs Output Results
- 18. NEW HGX WITH TESLA V100 VERSATILE GPU CLOUD COMPUTING 8x Tesla V100 with NVLINK Hybrid Cube ~1 PETAFLOPS peak Tensor Core throughput Optimized for all NVIDIA Deep Learning, CUDA HPC, and Pro Viz stacks
- 19. ß ANNOUNCING NVIDIA AND WORLD’S TOP SERVER MANUFACTURERS ADVANCE AI CLOUD COMPUTING NVIDIA HGX PARTNER PROGRAM
- 20. NVIDIA DEEP LEARNING STACK DEEP LEARNING FRAMEWORKS DEEP LEARNING LIBRARIES NVIDIA cuDNN, NCCL, cuBLAS, TensorRT CUDA DRIVER OPERATING SYSTEM GPU SYSTEM NVDOCKER
- 21. NEW NVIDIA GPU CLOUD GPU-ACCELERATED CLOUD PLATFORM OPTIMIZED FOR DEEP LEARNING Containerized in NVDocker Optimization across the full stack Always up-to-date Fully tested and maintained by NVIDIA Beta in July Registry of Containers, Datasets, and Pre-trained models NVIDIA GPU CLOUD CSPs
- 24. TOYOTA SELECTS NVIDIA DRIVE PX FOR AUTONOMOUS VEHICLES
- 25. A WORLD OF INTELLIGENT MACHINES 10% of Manufacturing Tasks Are Automated 1M Pizzas Delivered Per Day by Domino’s 100M People 80+ Years Old Ag Tech: 70% Increase in Farm Yields by 2050 600K Bridges to Inspect in the U.S. 300M Operations per Year WW
- 26. ANNOUNCING THE ISAAC INITIATIVE — A ROBOTICS PLATFORM Jetson TX2 Astro AV Stack Isaac Lab AV Reference Platforms
- 27. JETSON TX2 EMBEDDED AI SUPERCOMPUTER CUDA Linux or ROS cuDNN, TensorRT 2 Core i7 PCs in <10W 256 CUDA cores >1 TFLOPS
- 30. ISAAC ROBOT SIMULATOR NVIDIA GPU Computer OpenAI GYM Robot & Environment Definition Isaac Robot Simulator Virtual Jetson
- 33. JETSON REFERENCE PLATFORMS Consumer Drone Industrial Drone Service Robot UGV USV RC Car
- 34. NVIDIA AI IN THE CLOUD AI AT THE EDGE TENSOR RT TRAINED NEURAL NETWORK NVIDIA JETSONNVIDIA HGX-1 NVIDIA GPU CLOUD
- 35. THE TIME FOR GPU COMPUTING HAS COME GeForce GTX Max-Q Design NVIDIA HGX Partners Isaac Initiative