You are on page 1of 24

TOKYO | Oct.

5, 2016

NVIDIA DGX-1: INTEGRATING THE


POWER OF DEEP LEARNING AND
ACCELERATED ANALYTICS
Jim McHugh, 10/05/2016
NVIDIA THE AI COMPUTING COMPANY
Pioneered GPU Computing | Founded 1993 | $7B | 9,500 Employees

GPU COMPUTING COMPUTER GRAPHICS ARTIFICIAL INTELLIGENCE

2
AI FOR EVERYONE

AI will Revolutionize Transportation AI will Revolutionize Healthcare AI will Revolutionize Society


3
DEEP LEARNING
A NEW COMPUTING MODEL
Software that writes software

LEARNING
ALGORITHM

millions of trillions
of FLOPS

little girl is eating


piece of cake"
4
DATA DELUGE TO DATA HUNGRY

ZETTABYTES
AI Sensors Infotainment
Systems
Streaming
Video

EXABYTES
User IoT Data
DIGITAL Generated
Content Social
Network Natural
User Click
Language
Stream
Processing

PETABYTES
Mobile Web
Web
WEB Logs A/B Sentiment
Wearable
Testing Devices
Offer
History Business
Data Feeds Cyber
TERABYTES

Dynamic Security Logs


Offer
Pricing
Details HD Video
Segmentation Search
Marketing Speech To
BUSINESS Text
Connected
Purchase Purchase Vehicles
PROCESS Detail Record Behavioral
GIGABYTES

Targeting Product/
Service Logs
Support Payment Dynamic Machine
Contacts Record Funnels SMS/MMS Data
INCREASING DATA VARIETY 5
6
THE ADVANTAGES OF
GPU-ACCELERATED DATA CENTER
TFLOPS NVIDIA GPU x86 CPU

6.0

5.0 P100

4.0

K80
3.0

Fast GPU
2.0 K40 +
K20 Strong CPU
M2090
1.0
M1060

0.0
2008 2009 2010 2011 2012 2013 2014 2016

7
SCALE OUT STRONG SCALE
Lots of Nodes Interconnected with Few Lightning-Fast Nodes with
Vast Network Overhead Performance of Hundreds of Weak Nodes

8
DATA & ANALYTICS USE CASES
$
AUTOMOTIVE COMMUNICATIONS CONSUMER PACKAGED GOODS FINANCIAL SERVICES EDUCATION & RESEARCH
Auto sensors reporting Location-based advertising Sentiment analysis of Risk & portfolio analysis Experiment sensor analysis
location, problems whats hot, problems New products

HIGH TECHNOLOGY / LIFE SCIENCES MEDIA/ENTERTAINMENT ON-LINE SERVICES / HEALTH CARE


INDUSTRIAL MFG. Clinical trials Viewers / advertising SOCIAL MEDIA Patient sensors,
Mfg. quality effectiveness People & career matching monitoring, EHRs
Warranty analysis

OIL & GAS RETAIL TRAVEL & UTILITIES LAW ENFORCEMENT


Drilling exploration sensor Consumer sentiment TRANSPORTATION Smart Meter analysis & DEFENSE
analysis for network capacity,
Sensor analysis for Threat analysis - social media
optimal traffic flows monitoring, photo analysis

9
GPU ACCELERATION OVERCOMES
THE CHALLENGES OF SLOW COMPUTE ON ANALYTICS

Long response time Issuing iterative queries Analyst creativity


constrains questions asked becomes wearisome is impaired

ASK QUESTIONS YOU DONT EXPLORE GO BEYOND


KNOW THE ANSWERS TO FURTHER WHATS BEING ASKED

10
WORKAROUNDS ARE NOT THE ANSWERS

$
Sampling misses Pre-aggregation Scale out on CPU
the whole picture struggles at scale infrastructure has
tremendous hidden costs

EXPLORE THE OUTLIERS RELY ON SCALE WITH A ROI


AND LONG-TAIL EVENTS ACCURATE DATA

11
NVIDIA ACCELERATED ANALYTICS
GPUs in the Data Center

ANALYZE VISUALIZE AI-ACCELERATE

12
DGX-1 FOR ANALYTICS SOLUTIONS
+ ARCHITECTURES
DEEP
LEARNING

ACCELERATED
VISUALIZATION
VISUALIZATION

ACCELERATED
DATABASES DATABASES

CORE CORE
TECHNOLOGIES TECHNOLOGIES
Spark Scheduler Mesos

TRADITIONAL GPU-ACCELERATED
DATA CENTER DATA CENTER
NVIDIA Tesla GPUs NVIDIA DGX Products Cloud

13
ACCELERATED DATABASE SOLUTION
Overview
Built from the ground up to scale linearly, Kinetica's distributed, in-memory database simultaneously ingests,
explores, and visualizes streaming data for truly real-time actionable intelligence. .

Industry use cases


Retail: Customer 360/customer sentiment, supply chain optimization
Correlating data from point of sales (POS) systems, social media streams, weather forecasts, and even wearable devices. Better able
to track inventory in real time, enabling efficient replenishment and avoiding out-of-stock situations

Powering High Performance Analytics as a Service Solution: Delivering customer-focused services by leveraging all available
transactional data. Currently no ability for business user to do customized analytics; IT has to. Query response times taking 10s of
minutes, some over 2 hours, thus limiting ability to analyze and use data

Fin services: Large scale risk aggregations and billion+ row joins in sub-second time (5TB+ tables choke on RDBMS joins and Hadoop is
too slow). Also ideal for fraud and compliance use cases.

Ridesharing: View all passengers and drivers to monitor behavioral analytics. Watch for fast acceleration, sudden braking, too many
U-turns, etc. to avoid risk/lawsuits of faulty drivers

Manufacturing: Live streaming analytics on component functionality to ensure safety (avoid failures) and validate warranty claims
14
ACCELERATED ANALYTICS SOLUTION
Overview
MapD is a next-generation database and visual analytics layer that harnesses
the power of NVIDIA GPUs to explore multi-billion row datasets in
milliseconds.

Industry use cases


Telco: Correlates call records with server performance data to spot problems in real time,
plus build ad targeting profiles

Retail: Analyzed historical sales to assess geographic product demand for future inventory
and store locations

Finance: Hedge fund analysis of local and regional economic trends related to their portfolio
companies

AdTech: Assessing inventory availability by matching millions of audience members against


active ad units

15
DELIVERING UNPRECEDENTED DATA
CORRELATIONS TO CUSTOMERS

DEA theft of Silk Road bitcoins SIEM attack escalation Dropbox external sharing logs

Twitter botnet deconstruction Datacenter outages ML: Feature correlation, NLP


16
NVIDIA DGX-1
AI Supercomputer-in-a-Box

170 TFLOPS | 8x Tesla P100 16GB | NVLink Hybrid Cube Mesh


2x Xeon | 8 TB RAID 0 | Quad IB 100Gbps, Dual 10GbE | 3U 3200W

17
24
FIVE MIRACLES

Pascal Architecture 16nm FinFET CoWoS with HBM2 NVLink New AI Algorithms

18
DGX-1 A LEAGUE OF ITS OWN
16X

ResNet Inception v3 AlexNet vgg MSR

12X
Relative Training Performance

8X

4X

1X
0X
GeForce GTXTITAN
GeForce GTX TITANX X GeForce GTX1080
GeForce GTX 1080 Tesla P100
Tesla P100 DIGITS
DIGITSDevBox
DevBox(4X Quadro VCA (8X
Quadro Quadro DGX-1 (8XDGX-1
VCA Tesla P100)
(4X GeForceGTX
GeForce GTX Titan
TITANX)X) (8X Quadro
M6000)M6000) (8X Tesla P100)

NVIDIA CONFIDENTIAL. PRELIMINARY NUMBERS. NOT FOR DISTRIBUTION.


Caffe on DeepMark. GeForce TITAN X and GTX 1080 system: Intel Core i7-5930K @ 3.5 GHz, 64 GB System Memory | Tesla P100 (SXM2) system: Dual CPU server, Intel E5-2698 v4 @ 2.2 GHz, 256 GB System Memory 19
DGX THE ESSENTIAL TOOL
FOR DEEP LEARNING & ACCELERATED ANALYTICS

250 NODE AI & ANALYTICS TIME TO INSIGHT 100X MORE DATA IN


SUPERCOMPUTER-IN-A-BOX 10-100X FASTER MILLISECONDS

20
DGX STACK
Fully integrated Analytics and Deep Learning platform Instant productivity plug-and-
play, supports every AI framework
and accelerated analytics
software applications

Performance optimized across


the entire stack

Always up-to-date via the cloud

Mixed framework environments


baremetal and containerized

Direct access to NVIDIA experts

21
GPU-ACCELERATION HAS NO LIMITS
Kinetica
MapD Hardware costs that are 110 that of
MapD is 55x to 1,000x faster than standard in-memory databases
comparable CPU databases on billion+
row datasets

Graphistry
BlazeGraph See 100x more data at millisecond
200-300x speed-up speed

SQream
The supercomputing powers of the GPU combined with SQreams patented
technology, results in up to 100 times faster analytics performance on terabyte-
petabyte scale data sets
22
NVIDIA DGX-1 AI massive opportunity
The Essential Tool for
Data Scientists
Data Scientist productivity is vital
NVIDIA is the choice for Deep Learning and
AI-accelerated analytics
DGX-1 is fast, instantly productive

23
TOKYO | Oct.5, 2016

Thank you

You might also like