GTC India 2016

NVIDIA’s GTC India 2016 starts Tuesday in Mumbai but today is busy with preshow events including a packed set of NVIDIA DLI (Deep Learning Institute) sessions at IIT Bombay and this afternoon’s Emerging Companies Summit. Some of the hottest AI and Deep Learning startups will be presenting at ECS and I’ll update this post later with some of my favorites.

While I flew into Mumbai, I expect at least a few GTC attendees will arrive via Mumbai’s iconic Victoria Station, a Unesco World Heritage Site.

For more photos from GTC India, visit my SmugMug site.

Advertisements
Posted in Uncategorized

Energy Efficient Supercomputing with DGX SaturnV

Just how energy efficient is NVIDIA’s new DGX SaturnV? At about 9.5 gigaflops/watt, the Pascal P100 powered system is about 50% more energy efficient than the nearest non-NVIDIA powered system. But unless you follow the Top500 list of world’s fastest supercomputers, it probably is hard for you to visualize if 9.5 gigaflop/watt is a good or not. But many have taken note of NVIDIA’s 50% jump over competitors in efficiency. Think of buying a Tesla Model X that can go 300 miles on a single charge and then hearing a competitor had released a car capable of 450 miles. You would take note.

But back to DGX SaturnV’s intended use supporting NVIDIA’s AI software development. Does energy efficiency matter there? At a recent GTC, Baidu’s Andrew Ng stated that it cost about $100 in electricity alone to train their Deep Speech 2 Deep Neural Network (DNN) for English and Mandarin speech recognition. It is not uncommon for a data scientist to train a DNN 50 or 100 times before getting good results and Baidu has 100’s if not 1000’s of data scientists. The energy savings of a Pascal P100 powered system can add up quickly. All of us involved in the rapidly growing field of AI need to take energy efficiency into account when thinking about how to power our research.

DGX SaturnV is built out of 124 DGX-1 deep learning supercomputers each in turn with 8 Pascal P100 GPUs. But you don’t need to build the world’s 28th fastest supercomputer to take advantage of the Pascal P100’s energy efficiency. Last week, NVIDIA and Microsoft announced record breaking results for Microsoft’s Cognitive Toolkit running on DGX-1, an amazing 170 times faster than a standard two-socket x86 server. And if your budget can’t afford a DGX-1, you can get started with Microsoft’s Cognitive Toolkit using Azure’s new N-series GPU virtual machines for as little as 56 cents/hour. And of course in-between you have other options including the Titan X Pascal, the most powerful GPU available for PC developers.

Posted in Uncategorized

Tesla Shows Off The AI Car

The AI Car is here. Tesla’s latest video shows the capabilities coming soon to every new car now shipping from the Tesla factory.

Electrek has a more complete analysis of the video. This earlier article by Electrek detailed the NVIDIA GPU technology inside Tesla’s latest cars.

While all new Tesla cars rolling off the assembly line already include the NVIDIA hardware to power the capabilities demonstrated in the video, Tesla will continue to roll out new over the air software updates over the coming months, adding new autonomous capabilities with each release.

It will be interesting to see how fast other automobile manufacturers can introduce similar capabilities.

Posted in AI Car

Updates on NVIDIA’s Cloud-Mobile Products & Technology

@nvidia continues to push forward and deliver new capabilities in our cloud-mobile accelerated computing strategy.

On the gaming side, we recently announced GeForce NOW. Referred to by some of the press as “NetFlix for gaming”, GeForce NOW is a next generation cloud gaming service that streams PC games to NVIDIA SHIELD instantly, including the new SHIELD Android TV Box. Your game runs on powerful GPU servers running in the cloud and lets you enjoy gaming on your living room TV or anywhere you take your SHIELD portable or SHIELD tablet.

In public cloud, the big news has been Microsoft’s announcement of the new N-series GPU instances. Powered by NVIDIA’s latest Tesla M60 and Tesla K80 GPUs, N-series is the ideal cloud platform for running graphic intensive workloads with NVIDIA GRID 2.0 (M60) and HPC and Deep Learning workloads (K80).

Speaking of GRID 2.0, one of the key new features of GRID 2.0 is the ability to run native CUDA applications inside your GRID 2.0 virtual machine. Today’s enterprise applications don’t just need high quality graphics, but increasingly need to run accelerated algorithms inside graphical apps to crunch through huge quantities of big data. Now every CUDA and OpenACC app can run unmodified right in your VM.

Posted in Uncategorized

The GPU Accelerated Data Center

The rise of GPU accelerated data centers provides the foundation for a viable path forward in the post-Moore’s law era as called for by President Obama in his recent executive order, the National Strategic Computing Initiative.

NVIDIA is proud to be working on two of the largest pre-exascale systems announced to date, the Summit and Sierra systems which will achieve between 100 and 300 petaflops of peak performance when they are deployed starting in 2017. But already today there are many examples of organizations using GPU accelerated data centers. Some of the easiest examples to discuss include those running on GPU accelerated public clouds including NVIDIA’s latest K80 GPUs running on SoftLayer and Amazon Web Services G2 GPU instances

NVIDIA doesn’t just encourage customers to use GPU accelerated data centers, we do so ourselves. Many NVIDIA employees enjoy extended mobility, security, and super-fast data transfer speeds by running their graphics accelerated desktop as a virtual PC in our corporate data center using NVIDIA Grid. While you have to be an NVIDIA employee to have a virtual PC in our corporate data center, anyone can experience a similar experience with the VMware & NVIDIA Grid Test Drive which runs out of several public cloud data centers located around the world.

We have had thousands of developers sign up for our on-going Deep Learning Courses. The first two of five free online courses have already debuted, but you aren’t limited to watching the YouTube replays, you can take the free hands-on lab which runs on GPUs in the cloud using qwikLABS. We also offer a number of other GPU computing courses on qwikLABS.

For this year’s Imagenet competition, NVIDIA is teaming with SoftLayer and providing qualified competitors access to our latest K80 GPUs in the SoftLayer cloud.

Of course all work and no play isn’t fun, so we are even extending our use of the cloud to stream your favorite video games to SHIELD Android TV, SHIELD tablet, and SHIELD portable.

Posted in Uncategorized

The Remaking of NASA & Kennedy Space Center

Five years ago today, at Kennedy Space Center (KSC), President Obama laid out his vision for NASA and space exploration in the 21st century. Yesterday, for a few seconds at least, time seemed to stand still as I stood with a small invited team of NASASocial members on the banks of the newly opened ITL Causeway viewing area at KSC to watch the launch of the NASA SpaceX CRS-6 mission. Standing about two miles from launch pad 40, the moment I had been waiting three days for was marked by the launch control center announcer’s voice slowly counted down “3 … 2 … 1 … liftoff”. It is a magic moment that anyone who has ever seen a space launch will never forget.

Even with all the great 4K video footage of modern launchs, there is nothing like being there live. I watched the SpaceX Falcon 9 rocket with its Dragon spacecraft slowly lift off the pad towards the space station. Then less than 10 seconds later the sound waves hit. Over 1 million pounds of thrust at under two miles is not subtle, nor is it something any sound system in the world can replicate. You can literally feel the sound as it rolls up the causeway.

With all the excitement of a launch, it is sometimes easy to forget the purpose. Sure, as outlined in the NASA Press Kit the Dragon capsule was carrying water, food, and spare parts up to the astronauts in the international space station. But a big reason those astronauts are up there is to help conduct science experiments that can’t be done on earth. In conjunction with NASA, the Center for the Advancement of Science in Space (CASIS) manages the International Space Station U.S. National Laboratory. CRS-6 carried a number of CASIS academic and commercial science experiments including ones to study bone growth and specialized types of proteins made by immune cells.

Presidential visions and 20 year NASA plans can be inspiring during a speech, but how is KSC doing in executing on that plan five years into it. In the three days leading up to the launch, I had a chance to find out. After stopping to receive my NASASocial credentials at the NASA badging office just outside KSC Gate 3, I drove onto the sprawling KSC facility and turned left down a long road leading to the iconic Vehicle Assembly Building (VAB) and the launch complex 39 press site. For several minutes as you drive down the road the VAB looms ever larger in front of you and it is hard to believe you are driving right up to it.

As impressive as the VAB is from the outside, the immense size of the 37 story building is best experienced from the inside. Standing on a sixteenth floor work area one can look out and see the sign marking the height of the shuttle when it was erected vertically and attached to the shuttle solid rocket boosters and immense fuel tank. Don’t drop any wrenches, cell phones, or sunglasses from here.

Of course even more amazing is to look up and see the sign at the top of the building marking where one of its original occupants, the Apollo program’s Saturn 5 rocket once stood. Then just a bit below that is a new sign labeled SLS. More on SLS later.

With the retirement of the Space Shuttle, NASA has joined with commercial partners SpaceX and Boeing for the Commercial Crew System which in about two years will start ferrying astronauts to the space station and back. The SpaceX system builds on the current Falcon 9 rocket and Dragon capsule used for current unmanned commercial resupply missions while Boeing is developing a new vehicle dubbed CST-100 both of which will launch out of KSC. SpaceX is also joined by Orbital Sciences Corporation in space station commercial resupply missions, although Orbital is currently undergoing changes to their design after last year’s tragic crash.

Each NASA’s commercial crew program flight will be able to carry at least four crew members to the space station. This will allow the space station to be manned by a crew of 7 instead of the current 6 and nearly double the crew-hours available each week for experiments. While that math may sound funny, it is due to the many other non-scientific tasks the crew must undertake. Today, only about 40 hours a week of crew time are available for scientific experiments.

Last year, SpaceX signed a 20 year lease for KSC’s launch pad 39A, previously home to both shuttle and earlier Apollo launches. Touring the pad this week, the shell of the new SpaceX vehicle assembly hanger is nearing completion to the left of the pad. SpaceX will use this building to assemble both Falcon 9 and Falcon Heavy rockets with their payloads. A new transporter-erector vehicle will then move the assembled rocket, still horizontal, to the launch tower and then erect the rocket to its vertical launch position.

With commercial partners now primed to take over transporting crew and supplies to low earth orbit and the space station, NASA is now freed to focus on greater challenges including deep space exploration and future manned missions to Mars. For that role, NASA is building a rocket even more powerful than the Saturn 5 from the Apollo program, SLS. SLS will launch from a new mobile launch platform under construction today outside of the VAB.

Inside the VAB, one of the giant 50 year old Saturn 5 and Shuttle transporters is being retrofitted to carry the heavier SLS vehicle which when mated to the mobile launch tower will weight more than ten million pounds.

The transporter will roll under the SLS mobile launch platform, carry it into the VAB for SLS rocket assembly, then move the entire platform out to launch pad 39B.

During this week’s trip to KSC I was honored to hear four time astronaut and KSC director Bob Cabana talk about the transformation KSC has undergone in the last five years. Standing out on the press viewing area by VAB it is hard to argue with the progress Bob, KSC, NASA, and their commercial partners have made in making KSC into a multi user spaceport. SpaceX is launching from Pad 40 and rebuilding Pad 39A under a 20 year lease. United Launch Alliance, another commercial partner, is launching out of two other pads for NASA and the Air Force. Boeing is getting ready to assemble the CST-100 at KSC. And NASA itself is well underway to returning the VAB to its earlier glory with the nation’s new SLS headed to Mars and beyond. Orchestrating such complicated projects, across commercial companies and federal agencies, to build and operate a diverse set of space vehicles, is no small task. But Bob and the rest of the NASA team, along with their commercial partners, are off to a good start.

Posted in Uncategorized

How NVIDIA GPUs Support NASA’s Space Programs

Most of the team attending this week’s NASA Social SpaceX CRS-6 launch either play computer games on NVIDA GeForce GPUs or have kids who do. But I also received a lot of questioning looks when I said my interest in the launch was in the application of GPUs to supporting the space program. So here are a few examples.

With yesterday’s launch being scrubbed at T-3 minutes due to anvil clouds, weather is top of mind approaching today’s 4:10.40 pm (ET) instantaneous launch window. While last minute weather aborts are based mainly on realtime weather radar of the launch site, we have all become used to routinely seeing 10 day weather forecasts from numerous commercial web sites. These forecasts typically start with global weather simulations generated by one of the national weather services. Billions of numerical calculations are required to model the interactions occurring on earth and in our atmosphere. Using these global weather models and simulations as the starting point, regional models are then often used to provide more localized forecasts. Even the world’s most powerful supercomputers are not yet powerful enough to model the weather days in advance at a small enough scale to 100% accurately forecast the weather in a localized area like the Kennedy Space Center launch site.

Just as NASA is modernizing many of the Kennedy Space Center facilities in preparation for the upcoming Space Launch System, leading weather researchers around the world are updating their weather forecasting software to take advantage of powerful new GPU technology. One good example is the work by MeteoSwiss and CSCS on the COSMO regional weather model. Just like any backpacker has their favorite Swiss army knife and any foodie has their favorite Swiss cheese, everyone in the supercomputer field knows that CSCS operates the largest GPU powered supercomputer in Europe and has very accurate weather forecasts. The weather impacts everyone, no matter if you are trying to launch a spacecraft, take a commercial aircraft flight, or walk across the street on a cloudy day.

Now back to today’s CRS-6 launch. Sitting atop the SpaceX Falcon 9 rocket is the Dragon spacecraft, an unmanned capsule carrying over 4000 pounds of supplies for the International Space Station. Last year, SpaceX unveiled their next generation Dragon v2 spacecraft designed to carry up to 7 astronauts to the space station and back.

The 15 minute video on YouTube is worth a watch for more details. About 13 minutes into the video, Elon Musk steps into the spacecraft and shows off the pilot control displays, which have been widely reported to use an NVIDIA SOC (system on chip). Today’s NVIDIA Tegra line of SOCs incorporate smaller versions of the same GPUs used in some of the world’s most powerful supercomputers. This greatly simplifies life for software developers by enabling a scalable developer platform ranging from $192 NVIDIA Jetson TK1 developer kits to multi-million dollar GPU supercomputers that can run the same programs (albeit at different speeds).

Of course another area where GPUs are broadly used is indirectly related to NASA in that processing the science data from experiments carried out in space often uses GPUs. CRS-6 carries several science experiments that will generate data similar to what is routinely processed by GPU powered university systems like Clemson’s Palmetto cluster.

Last but not least, GPUs are increasingly being used to simulate future rocket engines, allowing them to be more efficient, powerful, and safe. Many of the rockets flying today are based on decade-old designs and could only be tested by firing them off in the middle of the desert. Just as a modern car is safer today because manufacturers can use computers to simulate many thousands of crashes instead of just test-crashing a few physical cars, future rocket engines will be safer because GPUs will provide the computational power necessary to simulate flying the rocket which involves not only complex computational fluid dynamics simulations but also molecular level simulations of the combustion process.

At NVIDIA, we are passionate about all the uses of GPUs, but for now, I’m most excited about that moment, planned for 4:10.40 pm ET today, when we hopefully hear those 3 magic words “we have liftoff”.

Posted in Uncategorized