Highlights GTC 2016
2
NVIDIA is the pioneer of the GPU — a massively parallel processor that is driving two of today’s most important technologies: artificial intelligence and virtual reality.
Originally designed for 3D graphics and gaming, the GPU has evolved into the computer brain of AI and VR, allowing computers to understand our human world, and humans to create and immerse themselves in computer-simulated worlds.
NVIDIA technologies serve the most demanding computer users in the world — scientists, engineers, and designers who do work at huge scale and with high stakes. Ordinary computers cannot support their work. The breakthroughs they seek — self-driving cars, AI-assisted medical diagnosis, speech and image recognition, intelligent virtual assistants — all require a supercharged form of computing called GPU computing.
NVIDIA is inspired by these great challenges. So we dedicate ourselves to the singular endeavor of advancing GPU computing. We strive to make a lasting impact by doing the best work of our lives and achieving the highest level of our craft.
3
Today, there are more than 300,000 GPU developers, a 400% increase in four years.
The NVIDIA GPU Technology Conference (GTC) brings together the scientists, engineers and designers doing amazing work across a wide array of fields, from internet services and self-driving cars, to drones and robotics, smart cities, industrial IOT, life sciences and medical diagnostics, energy discovery, defense and supercomputing.
GTC 2016 was our biggest yet:
5,500 attendees, 600 sessions, 200 exhibitors
500M social impressions, 55K mentions
750 press articles worldwide
“Distributed Deep Learning at Scale”
“Hybrid Reality at NASA Powered by NVIDIA”
— NASA
“Drive Me: Volvo's Autonomous Car Program”
— Volvo
2X GTC Attendees
600 GTC Sessions
4
We announced five big things at GTC to advance the fields of VR, AI and self-driving cars.
IRAY VRNVIDIA SDK
TESLA P100
NVIDIA DGX-1
HD MAPPING, AI DRIVING
5
At GTC, two stunning demos brought the magic of VR to life.
Everest: With 100 billion pixels and physics simulation, NVIDIA and Solfar Studios recreated Mt. Everest, down to the snow swirling around its peak.
Mars 2030: Engineers at NVIDIA and NASA and developers at Fusion VR reconstructed 8 sq km of Mars based on satellite imagery and data gathered over years of missions. The virtual Red Planet included 1 million rocks, lava tubes, and a rover and habitat based on NASA’s designs.
6
NVIDIA Iray brings VR to real-world industries.
Iray, which simulates the physical behavior of light and materials so designers can work with photorealistic models, is now coming to VR.
With Iray VR, we can achieve a level of photorealism that will bring virtual reality to more and more real-world applications, from how products, cars and buildings are designed, built and bought, to how people communicate with each other across vast distances.
7
Leading companies in every industry are racing to tap into the power of this new computing model.
The combination of big data, deep neural networks and powerful GPU platforms has dramatically accelerated the advance of AI.
8
“‘Miracle’ Computer Chip Gives Big Boost to Artificial Intelligence”
— Newsweek
NVIDIA Tesla P100 was built from the ground up for AI. The result of five “miracle” technologies including the newest NVIDIA architecture, Pascal, the Tesla P100 delivers giant leaps in performance and efficiency. Already in production, the P100 GPU is expected to be adopted first by cloud service providers with hyperscale datacenters and will power Europe’s fastest supercomputer, CSCS’ Piz Daint. Cray, Dell, HP and IBM servers based on P100 will be available in Q1’17.
9
“ This is a new era of computing. New
approaches to the underlying technologies
will be required for AI and cognitive. The
combination of NVIDIA Pascal GPUs and IBM
POWER accelerates Watson’s learning of new
skills. Together, IBM and NVIDIA will advance
the artificial intelligence industry.”
Dr. John Kelly III, SVP,
Cognitive Solutions & IBM Research
“ NVIDIA GPU is accelerating progress in AI.
As neural nets become larger and larger,
we not only need faster GPUs with larger
and faster memory, but also much faster
GPU-to-GPU communication, as well as
hardware that can take advantage of
reduced-precision arithmetic. This is
precisely what Pascal delivers.”
Yann LeCun, Director of AI Research, Facebook
“ Microsoft is developing super deep neural
networks that are more than 1000 layers. NVIDIA
Tesla P100’s impressive horsepower will enable
Microsoft’s CNTK to accelerate AI breakthroughs.”
Xuedong Huang, Chief Speech Scientist,
Microsoft Research
“ AI computers are like space rockets: The bigger
the better. Pascal’s throughput and interconnect
will make the biggest rocket we’ve seen yet.”
Andrew Ng, Chief Scientist, Baidu
10
NVIDIA DGX-1 brings the power of AI to every industry.
The world's first deep learning supercomputer, DGX-1 gives researchers and data scientists the ability to build AI-based services and autonomous machines that see and perceive the world like humans.
Packed with 8 Tesla P100 GPUs, DGX-1 delivers the equivalent throughput of 250 CPU-based servers and is fully integrated with deep learning software and development tools.
11
The first DGX-1s are headed to places where they can have the greatest impact.
Massachusetts General Hospital (MGH), which conducts the largest hospital-based research program in the U.S., recently opened its Clinical Data Science Center. As a founding technology partner, we’re working with MGH to apply the latest AI techniques to improve the detection, diagnosis, treatment and management of diseases. MGH will use DGX-1 and its database of 10B medical images to advance radiology, pathology and genomics.
We also donated 10 DGX-1s to the “pioneers of AI research” — ten universities that have been at the forefront of the field.
PIONEERS OF AI RESEARCH
12
GPU computing is the platform of platforms for modern AI.
The world’s largest cloud service providers, from Alibaba and Baidu to Facebook and Google, power their AI services with NVIDIA GPUs. These hyperscale datacenters use the powerful Tesla M40 to train their deep neural networks, while the energy-efficient Tesla M4 is used for production, or inferencing.
At GTC, we advanced our portfolio with the announcement of the GPU Inference Engine (GIE), which makes GPUs the most energy-efficient accelerator for deep neural networks
13
GPU computing gives IBM Watson a boost.
In his guest keynote, Rob High, CTO of IBM Watson, discussed how Watson has evolved from Jeopardy! champion into an intelligent assistant used by doctors, lawyers, financial analysts and marketers. He also demonstrated some of Watson’s latest applications in the fields of e-commerce, sentiment analysis and robotics.
Watson relies on GPUs for a variety of applications including deep learning training, where IBM has seen an 8x speed-up.
14
AI is coming to cars.
One of the most storied industries in the world is undergoing a major transformation.
Every major automaker in the world is working to bring more and more autonomous features into today’s vehicles, while building toward fully autonomous cars in the near future.
New entrants are working to reinvent transportation altogether.
Volvo Drive Me on Public Roads in 2017
Tesla Model 3: 300K pre-orders
Audi, BMW, Daimler Buy HERE Baidu and Uber Enter the Race
Tesla Model S Auto-pilot Toyota Invests $1B in AI Lab
15
NVIDIA DRIVE PX will accelerate the race to self-driving cars.
A one-of-a-kind platform, DRIVE PX is a scalable architecture that can cover every aspect of a self-driving car: deep neural network training, digital cluster and infotainment systems, ADAS, autonomous driving, and mapping.
Training on DGX-1
Driving with DriveWorks
KALDI
LOCALIZATION
MAPPING
DRIVENET
DAVENET
NVIDIA DGX-1 NVIDIA DRIVE PX
16
We expanded DRIVE PX with two breakthroughs.
First, a new, end-to-end HD mapping platform for automakers, map companies and startups. With the compute power of DRIVE PX in the car,and DGX-1 in the datacenter, they can rapidly create HD maps and update them on the go. We’re working with leaders HERE, TomTom and Zenrin to map the world.
We also revealed DaveNet — a breakthrough neural network that has learned to drive by mimicking human behavior. With just 3,000 miles of training, DaveNet has learned to navigate in some extremely challenging conditions, including in the rain, merging into highway traffic, and on dirt roads without lanes.
17
DRIVE PX is the engine in a new kind of race.
Roborace, the world’s first autonomous race, is set for the 2016-17 Formule E season. Ten teams will compete with identical cars. The “brains” of every car will be NVIDIA DRIVE PX.
“Now, computing powerhouse NVIDIA has announced that it will be providing AI supercomputers for the cars, making the whole Roborace series seem less like science fiction and more like just science.”
— Popular Science
18
Toyota: Simulation is key to self-driving cars.
In his guest keynote, Gill Pratt, CEO of Toyota Research Institute, set out Toyota’s vision of self-driving cars, including giving a sneak peek into its simulation facility.
Simulation is central to understanding how car and driver will work together as more autonomous features are introduced. Toyota’s NVIDIA-powered simulation facility will use GPUs — and deep learning — to train autonomous systems in real-world situations before they hit the road.
19
We rounded out GTC with a gift for developers.
The NVIDIA SDK collects our algorithms, libraries and tools into a single repository so developers across industries can take advantage of GPU computing to accelerate their work.