GTC 2015 Highlights

2
4,000 guests • 550 talks • 175 posters
“At the NVIDIA GPU Developer’s conference this week I’ll be thinking
about the future and wondering if I’m not already in it.” —TechZone

3
GTC 2015 focused on the promising field of deep learning.
And we made four major announcements that will fuel its advance.
TITAN X
The World’s Fastest GPU
DIGITS DevBox
GPU Deep Learning Platform
Pascal — 10x Maxwell
For Deep Learning
NVIDIA DRIVE PX
Deep Learning Platform
for Self-Driving Cars

4
“Let’s skip the foreplay. NVIDIA’s TITAN X
is the best single-GPU graphics card on
the market, and a remarkable feat of
engineering. This is an inarguable
conclusion.”
— Forbes
Our first announcement, TITAN X.
The world’s fastest GPU, TITAN X boasts
8 billion transistors, 3,072 CUDA cores,
and 12GB of memory. It can reach 7
teraflops of single-precision
performance.
“NVIDIA has now introduced four
unanswered graphics cards into the
market since AMD’s Radeon 285 in
August 2014.”
— Forbes

5
To illustrate the performance of TITAN
X, as well as the state of the art in real-
time graphics, we showed Epic’s latest
Unreal Engine 4 demo, Kite. But TITAN X
is also a breakthrough for deep learning
research, enabling data scientists to
train their networks in a fraction of
the time it used to take.

6
NVIDIA GPUs have been broadly adopted
in deep learning, a branch of artificial
intelligence.
Deep learning has been ignited by the
convergence of three trends: the flood
of data brought by web services
companies, recent algorithm
breakthroughs, and the ability to compute
massive amounts of data with GPUs.
Today, machines are being trained to
recognize images, text and speech.
But this is just the tip of the iceberg.
The world’s largest and most innovative
companies are deploying deep learning
across a variety of applications.
In 2012, GPUs enabled a breakthrough in
the ImageNet Challenge, the World Cup of
deep learning and computer vision. GPUs
have recently enabled machines to
outperform humans at this task.

7
We showcased leading-edge research in
deep learning from Andrej Karpathy of
Stanford. His work combines two neural
networks — one trained for image
recognition, one for language
processing. Connected “like LEGOs,”
the neural networks can not only
classify the objects in a photo, i.e.,
“bird” or “branch,” but also describe
them in the context of the scene.

8
Our second announcement, DIGITS
DevBox. To fuel the advance of deep
learning research, we created a very
powerful box.
“The DIGITS DevBox is comprised of both
DIGITS software and a quartet of TITAN
X GPUs — not to mention several
popular deep learning frameworks —
altogether of which promises up to four
times faster development.”
— ZDNet

9
Our third announcement, our latest GPU roadmap.
“NVIDIA also gave details of a future GPU technology, dubbed Pascal...the technology will be
particularly suited for humanlike computer chores known by the phrase ‘deep learning,’
offering a tenfold speed up in such calculations.” — The Wall Street Journal

10
Every major automaker in the world is
working toward self-driving cars.
Perhaps the biggest challenge facing
them today is the ability for cars to
navigate complex, urban situations
where human drivers make decisions
based on nuances and clues.
What may appear to be “free space”
for a car to drive through can change in
a heartbeat. For example, if a school
bus stops on the other side of the road,
or if the door of a parked car opens
suddenly.
For humans, the right response
becomes second nature with life
experience. But there are too many
possibilities to hard code into
machines. Deep learning offers a way
to augment traditional techniques to
pave the way toward self-driving cars.

11
Our fourth announcement, DRIVE PX. A
self-driving car computer, DRIVE PX can
augment traditional computer vision
techniques by powering a deep neural
network onboard the car. The work
builds on Project DAVE:
research by Urs Muller,
chief technologist of
autonomous driving at
NVIDIA, and Yann LeCun,
director of AI Research at Facebook,
when they collaborated at DARPA.
“The notion is that with powerful enough
hardware, self-driving vehicles will be
better able to recognize what they’re
seeing, learn from the environment and
make the right decisions.”
— re/code

12
“The days of humans driving their
own cars are numbered, according
to Elon Musk… NVIDIA's work will be
a ‘big enabler’ for Tesla's efforts.”
— Mashable
“Tesla and NVIDIA are among the small
set of Silicon Valley companies leading
the transformation of 21st century car
technology.”
— Fortune
“NVIDIA Steps on the Gas”
— The Wall Street Journal

13
“We love GPU cards. We just use a lot
of them.”
— Jeff Dean, Google
The theme of deep learning carried
through our guest keynotes. Jeff Dean,
senior fellow at Google, described how
the company is using GPU-powered
deep neural networks to bring greater
levels of intelligence to image, text,
and speech recognition. He also
highlighted work done by the recently
acquired Deep Mind. Using Atari video
games, the researchers trained a
network to not just classify, but take
actions in an environment.
Ultimately, the network
beat a series of games
and the work earned the
cover of Nature magazine.

14
Andrew Ng, widely recognized as a
leading thinker in deep learning and
currently chief scientist at Baidu,
China’s largest search engine, rounded
out the conference with his keynote.
Ng highlighted recent work on Baidu’s
Deep Speech engine, which uses deep
learning to recognize and process voice
commands even in noisy environments.
The GPU-powered neural network
trained on more than 100,000 hours of
speech samples to deliver the lowest
error rates ever seen in this field of
research.

15
“Yes, that’s right: VDI is as big at GTC as
it was at both Citrix Synergy and
VMworld last year.”
— Virtualization Practice
“One of the more fascinating talks here
at GTC 2015 is centered around deep
machine learning and its applications in
the medical field.”
— WCCFTech
More than 550 talks were presented on
the wide variety of fields and industries
that GPUs are disrupting, from cancer
research to the exploration of Mars. Our
exhibit hall showcased the latest
innovations from our partners. And our
Emerging Companies Summit once again
highlighted the work of startups.
Artomatix, this year’s winner of the
$100,000 Early Stage Challenge, is using
machine learning and big data analytics to
automate the creation of artwork for
video games.

16
Developers increasingly view GTC as
the place to come and learn about the
latest in GPU computing. This year,
more than 2,000 individual
programming labs — twice as many as
last year — were completed in areas
ranging from CUDA basics to computer
vision to deep learning.

17
We generated more than 1,300
articles from top business,
financial, tech, consumer tech, IT,
HPC, and vertical media.

18
“The #GTC15 keynote on deep learning
applications is blowing me away. Leaving me
w/ a totally different impression of @nvidia”
SOCIAL MEDIA HIGHLIGHTS
234,000
Total engagement on social media
(likes, clicks, shares)
95,000
Day 1 keynote live stream + replay views
90,000
Total views of blog posts
“I’ve struggled to explain DL to people before.
The #GTC15 explanation is awesome!”
“#GTC machine learning track room seats ~200
& standing room only in first session, feels
like academic conference #respect #nvidia”

19
“The ‘G’ (graphics) label for NVIDIA’s main product is becoming an anachronism. Instead, NVIDIA’s
hardware, software and engineering output are manifested in algorithms and APIs, not circuits
and interconnects. GPUs are a disruptive technology for databases, business analytics and
robotics that will allow unknown startups like those in the GTC Emerging Companies Summit
and giant corporations like IBM and Baidu to reshape markets.”
—Forbes

GTC 2015 Highlights

More Related Content

What's hot (20)

Similar to GTC 2015 Highlights (20)

More from NVIDIA (20)

Recently uploaded (20)

GTC 2015 Highlights

Editor's Notes