DGX UPDATE Customer Presentation Deck May 8, 2017
NVIDIA DGX-1: The World s Fastest AI Supercomputer FASTEST PATH TO DEEP LEARNING EFFORTLESS PRODUCTIVITY REVOLUTIONARY AI PERFORMANCE Fully-integrated and pre-optimized Insights in hours instead of weeks Optimized frameworks and cloud managed for faster insights DGX software stack for fastest GPU performance in the industry 2
ONE YEAR LATER NVIDIA DGX-1 Barriers Toppled, the Unsolvable Solved a Sampling of DGX-1 Impact DGX-1 launch UC Berkley CSIRO MIT CMU Fidelity Skymind RIKEN April 2016 April 2017 OpenAI NYU Mass. General Hosp. DFK IDSIA Microsoft SAP NVIDIA SATURNV launch Nimbix Noodle.ai 3
INTRODUCING THE DGX FAMILY AI WORKSTATION AI DATA CENTER CLOUD-SCALE AI DGX Station DGX-1 NVIDIA GPU Cloud with with The Personal AI Supercomputer Tesla P100 Tesla V100 Cloud service with the highest deep learning efficiency The World s First AI Supercomputer in a Box The Essential Instrument for AI Research 4
REVOLUTIONARY AI PERFORMANCE 3X system performance over prior generation Software stack delivers additional 30% faster training performance vs other GPU systems NVIDIA DGX unlocks the full potential of NVIDIA GPU s powered by software innovation 10X I/O performance with 2nd generation NVLink vs PCIe-connected GPU s New Tensor Core architecture inspired by the demands of deep learning 5
EFFORTLESS PRODUCTIVITY Save $x00,000 s on software engineering of DL frameworks Depend on NVIDIA-optimized frameworks instead of evolving open source software Save $100k+/yr in admin OpEx with cloud management, streamlined collaboration Monthly framework releases ensure maximized performance for DL ROI NVIDIA DGX software stack delivers immediate productivity that saves time and money 6
COMMON SOFTWARE STACK ACROSS DGX FAMILY Single, unified stack for deep learning frameworks Predictable execution across platforms Pervasive reach NVIDIA GPU Cloud DEEP LEARNING FRAMEWORKS DEEP LEARNING USER SOFTWARE NVIDIA DIGITS THIRD PARTY ACCELERATED SOLUTIONS CONTAINERIZATION TOOL NVIDIA Docker GPU DRIVER NVIDIA Driver DGX Station DGX-1 NVIDIA Cloud Service SYSTEM Host OS 7
ENTERPRISE BENEFITS OF DGX SOFTWARE NVIDIA Investments in Deep Learning Performance and Manageability Popular deep-learning frameworks - GPU-tuned by NVIDIA Engineering Practitioner productivity with minimal setup Driver and library independence for each framework Clean, minimal O/S base image Optimized drivers and libraries for maximized multi-gpu performance Non-disruptive updates for software and security 8
DGX CLOUD SERVICES Start Faster, Stay Productive Benefits for Deep Learning Workflow Features Container Registry Single software stack Develop once, deploy anywhere Scale across teams of practitioners Web UI and CLI Job Scheduling and Management Host Telemetry User Management Python SDK and REST API 9
10 STEPS TO SETUP A DIY SYSTEM 380 PAGES OF DOCS TO READ Step 1. Install Ubuntu linux (10 pg) Step 2. Install CUDA (41 pg) Step 3. Install CUDNN (154 pg) Step 4. Install and Upgrade PIP (20 pg) Step 5. Install BAZEL (build TF source) (50 pg) Step 6. Install TensorFlow (15 pg) Step 7. Upgrade Protobuf (15 pg) Step 8. Install Docker (75 pg) Step 9. Test the installation Step 10. Debug and fix install NVIDIA CONFIDENTIAL. DO NOT DISTRIBUTE. 10
DL FROM DEVELOPMENT TO PRODUCTION Accelerated Deep Learning Value with DGX Solutions From Desk DGX Station DGX-1/SATURNV/Cloud To Data Center or To Cloud installed optimized scaled Fast Bring-up Productive Experimentation Training at Scale Procure DGX Station Install / Compile Experiment Tune/ Optimize Deploy Train Insights 11
OUR STRATEGY IN THE DATACENTER: NVIDIA DGX-1 Highest Performance, Fully Integrated HW System 8 TB SSD 8 x Tesla V100 16GB 960 TFLOPS 8x Tesla V100 16GB 300 GB/s NVLink Hybrid Cube Mesh 2x Xeon 8 TB RAID 0 Quad IB 100Gbps, Dual 10GbE 3U 3200W 12
DEEP LEARNING FRAMEWORKS DEEP LEARNING USER SOFTWARE NVIDIA DIGITS THIRD PARTY ACCELERATED SOLUTIONS CONTAINERIZATION TOOL NVIDIA Docker GPU DRIVER NVIDIA Driver NVIDIA DGX-1 SOFTWARE STACK Fully Integrated Software for Instant Productivity Advantages: Instant productivity with NVIDIA optimized deep learning frameworks Caffe, CNTK, MXNet, PyTorch, TensorFlow, Theano, and Torch Performance optimized across the entire stack SYSTEM Host OS DGX SOFTWARE STACK Faster Time-to-Insight with pre-built, tested, and ready to run framework containers Flexibility to use different versions of libraries like libc, cudnn in each framework container 13
SIMPLIFY PORTABILITY WITH NVIDIA DOCKER CONTAINERS Benefits of Containers: Simplify deployment of GPU-accelerated applications Isolate individual frameworks or applications Share, collaborate, and test applications across different environments 14
THE POWER TO RUN MULTIPLE FRAMEWORKS AT ONCE Container Images portable across new driver versions Containerized Applications NVIDIA Docker NVIDIA Docker NVIDIA Docker NVIDIA Docker NVIDIA Docker Microsoft Cognitive Toolkit... Other Frameworks and Apps TF Tuned SW CNTK Tuned SW Caffe2 Tuned SW Pytorch Tuned SW Tuned SW CUDA RT CUDA RT CUDA RT CUDA RT CUDA RT Linux Kernel + CUDA Driver NVIDIA DGX-1 15
DGX-1: 96X FASTER THAN CPU DGX-1 7.4 hours 8-way GPU Server 18 hours Dual Socket CPU 711 hours 1X 40X 96X Workload: ResNet50, 90 epochs to solution CPU Server: Dual Xeon E5-2699 v4, 2.6GHz NVIDIA CONFIDENTIAL. DO NOT DISTRIBUTE. 16
RIKEN SUCCESS STORY Fujitsu and NVIDIA Build AI Supercomputer With 24 DGX-1s CHALLENGE Enterprises and research organizations embracing AI/DL Needed to accelerated research in areas including medicine, manufacturing and healthcare Conventional HPC architectures too costly and inefficient SOLUTION Partnered with Fujitsu for scale-out AI architecture built on DGX-1 24 DGX-1 s deliver 4 petaflops powering the RIKEN supercomputer NVIDIA COSMOS streamlines AI researcher workflow, helping accelerate RIKEN productivity IMPACT Accelerated real-world implementation of scale-out AI Enables RIKEN team to take advantage of next-gen DL algorithms Helping create future in which AI finds solutions to societal issues 17
BENEVOLENTAI: TRAINING REDUCED TO DAYS Technology Review Article on DGX-1: The Pint-Sized Supercomputer That Companies Are Scrambling to Get https://www.technologyreview.com/s/603075/the-pint-sized-supercomputer-that-companies-are-scrambling-to-get/ SYSTEM INSTALLATION TRAINING MODELS NVIDIA DGX-1 Days Other GPU System Weeks of Training The cost of renting enough servers on Amazon Web Services would surpass the system s $129,000 price tag within a year. -Jackie Hunter, CEO, BenevolentAI DGX-1 3x-4x FASTER TRAINING 18
INTRODUCING NVIDIA DGX STATION Groundbreaking AI at your desk The Personal AI Supercomputer for Researchers and Data Scientists Revolutionary form factor - designed for the desk, whisper-quiet Start experimenting in hours, not weeks, powered by DGX Stack Productivity that goes from desk to data center to cloud Breakthrough performance and precision powered by Volta 19
3X FASTER THAN THE FASTEST WORKSTATIONS Supercomputing performance at your desk 480 TFLOPS Water-cooled performance the only workstation built on 4 Tesla V100 s 3X 30% 5X 3X the performance of today s fastest GPU workstations with 30% faster training over non-dgx stack solutions 5X increase in I/O performance with 4-way next generation NVLink vs. PCIe-connected GPU s 20
ANNOUNCING 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 NVIDIA GPU CLOUD Registry of Containers, Datasets, and Pre-trained models CSPs 21
DGX CLOUD SERVICES Detailed Feature Walkthrough 22
DGX CLOUD SERVICES Management Workflows Plug-in, Power up Authenticate appliances View appliances Node events (cloud connect, change of master, IP, etc.) Node state (connect, disconnect, ready, faulty) Recovery ISO Image Factory PXE Boot Image Container updates On-premises setup Cloud portal connectivity Cluster / Node Management Software Management 23
DGX CLOUD SERVICES Management Workflows Container Management Job Execution Job Status Hardware Health CPU, CPU, RAM Util. Alerts System Image System Software Docker Image Application / Job Scheduling Metrics / Notifications System Updates 24
DL FRAMEWORK WORKFLOW - NGC DGX Management in 7 Easy Steps BUILD MANAGE SCALE 1 2 3 4 5 6 7 User Mgmt Create Accounts User Mapping Assign Users to Projects Container Repo Pull/Push Frameworks to Node(s) Config Job Resources Assign GPU/CPU/RAM Review / Submit Job Ready / Warnings Job Mgmt Status / Detail / Clone Schedule FIFO Scheduler 25
NVIDIA DGX SYSTEMS Faster AI Innovation and Insight The World s First Portfolio of Purpose-Built AI Supercomputers Powered by NVIDIA GPU Cloud Get Started in AI Faster Effortless Productivity Performance Without Compromise For More Information: nvidia.com/dgx-systems 26