Intel® Xeon® Processor Platform Performance Hardware Plus Optimized Software INFERENCE THROUGHPUT TRAINING THROUGHPUT

HPC Goes Mainstream with AI & Intel®

SEETHA . NOOKALA DIRECTOR , IN TEL A P J H P C&AI Seetha Rama Krishna-(Seetha) A researcher turned to techno entrepreneur. • Director: DCG, E&G – HPC & AI at Intel for APJ (South)

• HPC & Enterprise Research and Solutions: • Over 30 years of HPC experience contributing to R&D, Product Dev, applying HPC to Enterprise ,Cloud, Research, E Governance ,Data centers. • Formed and lead Advanced computing solutions teams at C-DAC, TATA- CRL, TCS , Intel. • Started career as researcher and quickly moved on to systems and solutions. • Chief Architect of world’s 4th fastest “eKA” super computer , and Intel’s first in top10 computer at TATA-CRL • Member PARAM Series HPC Architecture teams at C-DAC. • Recognitions: • Global Innovation Award winner in 2009- From TATA Sons. Felicitated by Shri. Ratan Tata in the presence 100Plus CEO’ and CFO’s of the TATA group. • Recipient of PARAM technology Awards from C-DAC, IEEE,CSI recognitions for promotion of HPC in India.

2 Software FUELS Hardware

3 Review more success stories, Intel® Parallel Studio XE Case Studies deck Software Optimization Success Stories Science & research Finance Up to 35X Up to 2.7X improved performance** faster application performance compared to NVIDIA Tesla K80* NERSC (National Energy Research Scientific Computing Center) - see case study Monte Carlo European Options Benchmark* Life Science Visualization Simulations ran up to 7.6X faster Up to 5.17X performance with 9X energy efficiency** improvement** compared to NVIDIA Titan X* LAMMPS code - Sandia National Laboratories intel.com/content/www/us/en/high-performance-computing/hpc-xeon-phi-technology-brief.html Intel Embree v2.9.0

**Intel® Xeon Phi™ Processor (codenamed Knights Landing) Software Ecosystem Momentum Guide Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations & functions. Any change to any of those factors may cause the results to vary. You should consult other information & performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more information go to http://www.intel.com/performance. Optimization Notice: Intel's compilers may or may not optimize to the same degree for non-Intel microprocessors for optimizations that are not unique to Intel microprocessors. These optimizations include SSE2, SSE3, and SSSE3 instruction sets and other optimizations. Intel does not guarantee the availability, functionality, or effectiveness of any optimization on microprocessors not manufactured by Intel. Microprocessor-dependent optimizations in this product are intended for use with Intel microprocessors. Certain optimizations not specific to Intel microarchitecture are reserved for Intel microprocessors. Please refer to the applicable product User and Reference Guides for more information regarding the specific instruction sets covered by this notice. Notice revision #20110804 . Optimization Notice Copyright © 2017, Intel Corporation. All rights reserved. 4 *Other names and brands may be claimed as the property of others. Chief Technical Officer Software and Services Group

Open Source Developer Products Datacenter Software Intel Services Developer Relations Technology Center Division Division Division Division

SSG Operations SSG Group System Technologies Windows* OS Marketing and Optimization Division Division

Intel Confidential Software and Services Group ‹#› Affordable & Fast AI on Intel - Software tools + Hardware Intel Portfolio to accelerate time-to-solution – HPC to AI

© 2018, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Intel®Intel® Xeon® Xeon® processor processor scalablescalable family family Scalable performance for widest variety of AI & other datacenter workloads – including deep learning

Built-in ROI Begin your AI journey today using existing, familiar infrastructure The AI you need Potent performance On the chip you know DL training in days HOURS with up to 113X2 perf vs. prior gen (2.2x excluding optimized SW1) Production-ready Robust support for full range of AI deployments

1,2Configuration details on slide: 4, 5, 6 Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more complete information visit: http://www.intel.com/performance Source: Intel measured as of November 2016 Optimization Notice: Intel's compilers may or may not optimize to the same degree for non-Intel microprocessors for optimizations that are not unique to Intel microprocessors. These optimizations include SSE2, SSE3, and SSSE3 instruction sets and other optimizations. Intel does not guarantee the availability, functionality, or effectiveness of any optimization on microprocessors not manufactured by Intel. Microprocessor-dependent optimizations in this product are intended for use with Intel microprocessors. Certain optimizations not specific to Intel microarchitecture are reserved for Intel microprocessors. Please refer to the applicable product User and Reference Guides for more information regarding the specific instruction sets covered by this notice. Notice Revision #20110804 Intel® Xeon® processor Platform Performance Hardware plus optimized software INFERENCE THROUGHPUT TRAINING THROUGHPUT

Up to Up to 198x Optimized 127x Frameworks

Intel® Xeon® Platinum 8180 Processor Intel® Xeon® Platinum 8180 Processor higher Intel optimized Caffe GoogleNet v1 with Intel® MKL higher Intel Optimized Caffe AlexNet with Intel® MKL inference throughput compared to training throughput compared to Optimized Intel® Intel® Xeon® Processor E5-2699 v3 with BVLC-Caffe Intel® Xeon® Processor E5-2699 v3 with BVLC-Caffe MKL Libraries Inference and training throughput uses FP32 instructions

Deliver significant AI performance with hardware and software optimizations on Intel® Xeon® Scalable Family

Up to 191X Intel® Xeon® Platinum 8180 Processor higher Intel optimized Caffe Resnet50 with Intel® MKL inference throughput compared to Intel® Xeon® Processor E5-2699 v3 with BVLC-Caffe Up to 93X Intel® Xeon® Platinum 8180 Processor higher Intel optimized Caffe Resnet50 with Intel® MKL training throughput compared to Intel® Xeon® Processor E5-2699 v3 with BVLC-Caffe

Performance estimates were obtained prior to implementation of recent software patches and firmware updates intended to address exploits referred to as "Spectre" and "Meltdown." Implementation of these updates may make these results inapplicable to your device or system. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more complete information visit: http://www.intel.com/performance Source: Intel measured as of June 8 2017. Configurations: See the last slide in this presentation. *Other names and brands may be claimed as the property of others. End-to-end Intel software value

Things And Devices Connectivity And Network Cloud And Data Center

0 1 0 1 0 1 0 1 1 0 1 0 1 0 1 0 1 1 1 0 1 0 0 1 0 1 0 1 0 1 0 0 1 0 1 0 1 1 0 1 0 1 0 1 0 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 1 0 1 0 1 0 1 0 1 1 0 1

Intel® Computer Wind River® Simics and Vision SDK Neon BigDL on Spark Software Readiness Qualification (SRQ) X-Platform Intel® Deep Developer Intel® CoFluent™ Technology

Enabling Security

Learning SDK Zone Ecosystem Ecosystem

Android Things OS OS

…among others… Enabling

9 Tools for High Performance Implementation Intel® Parallel Studio

Intel® MPI Library Cluster N Intel® MPI Benchmarks Scalable Tune MPI ? Y Intel® Compiler

Memory Effective Y N Vectorize Bandwidth Intel® Math Kernel Library threading Sensitive ? Intel® IPP – Media & Data Library ? Intel® Data Analytics Library N Y Intel® OpenMP* Optimize Thread Bandwidth Intel® TBB – Threading Library

Optimization Notice Copyright © 2018, Intel Corporation. All rights reserved. 10 *Other names and brands may be claimed as the property of others. 11 AI & HPC Convergence AI Self-Learning and Completely Automated HPC Computerized Human Thought Simulation and Actions Enable HPC to Advanced Towards Autonomic Enterprise shift up the Analytics Cognitive Maturity curve. Prescriptive Analytics Analytics Data preparation Predictive Operational Analytics accounts for ~80% of the Analytics Simulation-Driven Analysis work of data scientists. and Decision-Making Diagnostic Foresight Mature Data Lake Analytics What Will Happen, Drive When, and Why Descriptive AI HPC Hindsight Analytics Insight What Happened and Why What Happened Convergence AI

organizations expected industrialize to deploy AI Solutions increase offerings by 2020

1/5th of organizations Between 2018-19, IT already deployed AI budget spend on new solution technologies

“Newton’s cradle” initiative to reskill employees in AI, ML

Source: IDC AI Survey Dec 2017

and more * * * * ON enabled via Intel® nGraph™ Library

Learn more at ai.intel.com/framework-optimizations/

SEE ALSO: Machine Learning Libraries for Python (Scikit-learn, Pandas, NumPy), R (Cart, randomForest, e1071), Distributed (MlLib on Spark, Mahout) *Limited availability today Other names and brands may be claimed as the property of others.

Optimization Notice Copyright © 2018, Intel Corporation. All rights reserved. *Other names and brands may be claimed as the property of others. Adoption of Python continues to grow among Python* Landscape domain experts & developers for its productivity benefits

Challenge#1 Domain experts are not Most Popular Coding professional software programmers Languages of 2018

Challenge#2 Python performance limits migration to production systems

Intel’s Python Tools ▪ Accelerate Python performance ▪ Enable easy access

▪ EmpowerOptimization Notice the community Copyright © 2018, Intel Corporation. All rights reserved. 14 *Other names and brands may be claimed as the property of others. Intel® Data Analytics Acceleration Library (Intel® DAAL) High Performance ML and Data Analytics library Building blocks for all data analytics stages, including data preparation, data mining & machine learning

Pre-processing Transformation Analysis Modeling Validation Decision Making

Open Source  Apache 2.0 License Common Python, Java and C++ APIs across all Intel hardware Optimized for large data sets including streaming and distributed processing Flexible interfaces to leading big data platforms including Spark and range of data formats (CSV, SQL, etc.)

Emergency Response Financial services Machine Vision Cities/transportation

Autonomous Vehicles Responsive Retail Manufacturing Public sector

Optimization Notice Copyright © 2018, Intel Corporation. All rights reserved. *Other names and brands may be claimed as the property of others. AI: Distributed Compute from Device to Cloud Edge drivers: Bandwidth, Storage, Latency, Security

of data will be stored, share of AI tasks taking growth in devices with edge analyzed, and acted on at place on edge devices AI capabilities by 20232 45% the edge by 2019¹ 43% (vs. cloud) in 20232 15X

© 2018, Amazon Web Services, Inc. or its Affiliates. All rights reserved.1. Source: IDC FutureScape: Worldwide Internet of Things 2017 Predictions 2. ABI Research Introducing Intel® Neural Compute Stick 2 A Plug-and-Play Deep Learning Development Kit

Intel® Movidius™ Myriad™ X VPU delivers industry leading performance UP TO 8X¹ delivering On deep neural networks compared to + Intel® Movidius™ Neural Compute Stick Intel® Distribution of Powered Powered by OpenVINO™ toolkit accelerates HIGHER solution development and streamlines deployment PERFORMANCE

Order now for $99 MSRP*: Where to buy © 2018, Amazon Web Services, Inc. or its Affiliates. All rights reserved. *MSRP is not a guarantee of final retail price. MSRP may be changed in the future based upon economic conditions. 18 Intel Distribution OpenVINO™ Toolkit VISUAL INFERENCING AND NEURAL NETWORK OPTIMIZATION

Write once + scale to Diverse Accelerators

Broad Framework support

DEPLOY COMPUTER VISION AND DEEP LEARNING CAPABILITIES High Performance, high Efficiency for the edge TO THE EDGE

Optimization Notice Other names and brands may be claimed as the property of others VPU = Vision CopyrightProcessing Unit © (Movidius) 2018, Intel Corporation. All rights reserved. *Other names and brands may be claimed as the property of others. Intel® ACADEMY Code Modernization

Outreach Program For Partner program focused on Developers, And Startups market enablement

software.intel.com www.intel.com/sofwtare

Developers, Students Universities Intel Student ISV and SI in & Teacher trained Engaged Ambassadors Builder Program

20 * Against a goal of 15,000, set in April 2017 Get The Most From Your Code Today with Intel Tech.Decoded

Watch Big Picture Videos Dig deeper with Essential Get Started with Quick Hits

Focused conversations where Webinars covering strategies, Put into practice — short videos tech. visionaries share key practices and tools that help and articles that deliver the how- concepts on front-line topics, you optimize applications and to’s of specific programming what you need to know and why solutions performance. tasks using Intel tools. it matters.

Visual Code Systems & IoT Data Science Data Center & Computing Modernization Cloud Computing

Visit TechDecoded.intel.io — a video series where developers learn to put into practice key optimization strategies with Intel Development tools.

2 1 HPC – Past…Present & Future Fundamental Discovery to Business Big Data Analytics Enabling Gain Fundamental Insights Transformation Data Driven Science

Clinical Genomics Information

“My goal is simple. It is the ✓ Better Products complete understanding of the Rapid Transition From Diagnosis universe, why it is as it is and ✓ Faster Time to Market to Personalized Treatments what exists at all” ✓ Reduced R&D Stephen Hawking

Image Sources in backup

22 Copyright© 2014, Intel Corporation. All rights reserved. *Other brands and names are the property of their respective owners. Aerospace Biology Brain Modeling Chemistry/Chemical Engineering Climate Computer Aided Engineering Cosmology Cybersecurity Defense HPC is Foundational to Insight & AI

Scientific Business Artificial Discovery Innovation Intelligence

Advancing Science High ROI: Machine Learning And Our Understanding Joins Computational of the Universe $515 Learning Theory with HPC Average Return Per $1 of HPC Investment1

Pharmacology Particle Physics Metallurgy Manufacturing / Design Life Sciences Government Lab Geosciences / Oil & Gas Genomics Fluid Dynamics

1 1Source: IDC HPC and ROI Study Update (September 2015) Copyright© 2014, Intel Corporation. All rights reserved. *Other brands and names are the property of their respective owners. The growing challenge in hpc

System Bottlenecks Divergent Barriers to “The Walls” Workloads Extending Usage Machine learning Big Optimizing hpc Data For Cloud visualization

Intel® Intel Parallel Computing Centers Movidius OpenVino Intel® Dev CLoud Fathom platforms Intel® DL Studio ToolKit

Frameworks Mllib BigDL

Intel® Data Analytics Intel® Nervana™ Graph* Intel Python Acceleration Library libraries Distribution Intel® Math Kernel Library (DAAL) (MKL, MKL-DNN)

Compilers – C/C++/Fortran Analysis Tools – Vtune, Advisor, Inspector HPC Software Tools Libraries – Math Kernel Library, TBB, IPP, DAAL, MPI LibraryCluster Checker and Trace * analysis tools hardware Compute Memory & Storage Networking

*Future Other names and brands may be claimed as the property of others. Intel Technology Programs 2018 with Partners

Webinars On Site Workshops Tools HOLs/Hacks Partner Meet /Training Ind Dev Forums

Reach tools users via our Build affinity/expertise on Intel Build affinity with tools channel partners for Deep Dive engagements through Driving deep adoption for Intel for OEM/SI/ISV community - Objective customers through continuous connect in tools for Strategy/Channel Accounts Tools through case studies serve as Intel extended sales channel reach local time arm. Build strong network of Top 30 influence customers – Reach out to technical Reach Decision Maker + Participate in Intel Intel software Academia, Finance, developers– incentivize use of technical teams and build DevCons, AWS Events, GSI Strategy champions/usage – go CSP/Ecommerce, Manufacturing, Intel tools through competition awareness of new features + days, IOT Forums, vertical + use cases – GSI / Access to tools Additions for software tools Nasscom, IITs SMG Funded

Universities/Research Target Developer teams within customers Senior level – OEMs/SI/ISVs Extended community of SMG/Tools Users institutes/Customers/Enterprise Audience that develop code Software Developers – IISER + Pune Customers Pre sales – OEMs/SI/ISVs 1 day (senior meet) + 2 day 2 hours – half day – Duration 45mins – 90mins 1 day. 2 Day, 3 Day 1/2/3/5 Day sessions technical training (SA teams) technical workshops Deep dive technical training on Product Lines TEC, EMB, Media TEC,EMB,Media Intel architectures and Intel TEC, EMB, Media TEC, EMB ,Media 51 Programs , 23,000 Developers,products via AWS 5100 customers Across India , Singapore, Bangalore, Pune, Hyderabad, Delhi, Location Pan India Bangalore, Singapore India, Australia, Singapore Malaysia, ANZ Singapore, ANZ

Intel® Parallel Studio XE Intel Code Modernization Program ▪ Overview, features, support, code samples ▪ Overview ▪ Dev tools training materials ▪ Live training ▪ Reviews & Case Studies ▪ Remote Access ▪ More Intel® Software Development Products

Denver, CO – Nov. 11-12

software.seek.intel.com/fall-webinar-series intel.com/content/www/us/en/events/hpcdevcon/overview.html

➢ Convergence of HPC & Artificial intelligence (AI), the next big wave in computing, is an increasingly important source for competitive advantage that is already transforming industries

➢ Today is the ideal time to begin integrating these technologies into your products, services and business processes

➢ Partner with Intel/Boston to WIN in HPC/AI with complete software tools portfolio, world-class silicon – EVOLVE THE CONVERSATION

Get started with Team Intel – www.intel.com/software/products 29