A Blog about the World of Graphics Processing

There was a lot exciting news out of China last week as NVIDIA CEO Jen-Hsun Huang, Chief Scientist Bill Dally and General Manager of NVIDIA’s Tesla business Andy Keane visited Beijing. Highlights included:

 
• New CUDA Centers of Excellence- Tsinghua Universityand the Institute of Process Engineeringat the Chinese Academy of Sciences were recognized as the newest CUDA Centers of Excellence(CCOE). The organizations were selected for their commitment to GPU computing research and teaching of parallel programming courses based on the CUDA architecture. They join five other CCOEs: Harvard University in Cambridge, Mass.; University of Illinois at Urbana-Champaign; University of Utah in Salt Lake City; Cambridge University in the U.K.; and National Taiwan University in Taipei.
 
• CUDA “Future Star” Contest Awards- These awards recognize innovative programming work by university students in China. More than 70 CUDA-based entries were submitted, covering scientific computing, engineering and consumer applications. One of the contest judges, Professor Yangdong Deng from Tsinghua University, began the award ceremony by saying “CUDA is hot in China!”
 
Other events that the NVIDIA execs attended included a GPU computing seminar focused on trends in high-performance computing; recognition of the world’s first CUDA book, written by Tsinghua University professors; an interview of Jen-Hsun by popular online tech portal ZOL; and a heartwarming visit to the Haidian Foreign Language Experimental School. The NVIDIA Beijing team has forged a strong, ongoing relationship with students at this school following the tragic Sichuan earthquake last year.

-Reported by Calisa Cole and Ming Wei

When our distribution partner EVGA came to us with an idea to launch a new graphics card that had two GPUs on it (one for graphics, one for PhysX), and wanted to generate some buzz around the end of October, our thoughts naturally gravitated towards Halloween. And what’s Halloween without the cool costumes, miniature candy, and a rocking fun party?

So, instead of doing the usual, we decided to hold a party for 300 of our closest gamer friends. Invites were sent out, our location was secured on site and we made the critically important decisions -- what kind of food and how much to order.

Of course, this being NVIDIA, what’s a party without some cool PCs to game on? Since Batman: Arkham Asylum is probably the hottest title on the PC right now, we decided to build out about 15 full systems, each complete with EVGA’s new GTX 275 CO-OP PhysX GPUs. And we set up our NVIDIA 3D Vision technology so guests could compete in the Batman: Arkham Asylum challenge mode in 3D to win some awesome prizes.

Although the party started officially at 7 p.m., our first guests started arriving at the NVIDIA campus at 4:30 p.m.! That should have been an early indicator of what was ahead. By the end of the night, fans came out in droves, some from as far away as Ohio and even Australia! For the rest, this was a close to home “pre-Halloween” night out.   GPUs, Tee-shirts, and hats were among the many prizes given out.  Highlights of the night included a Batman-themed costume contest, a 3D theater showing new PhysX demos and new 3D movies, and of course, the aforementioned Batman gaming contest.  The group chomped though 180 pizzas and 500 cookies – we got the food just right -- while playing 200 individual games of Batman vying for the grand  prizes of the evening – two GTX 275 CO-OP PhysX cards from EVGA.   The final games were shown on the big screen at NVIDIA cafeteria in front of the cheering crowd.

Check out the enclosed pics, or mosey on over to YouTube and do a search for NVIDIA Halloween and check out some of the action.

We really had a great time hosting the event and enjoyed talking to everyone who came out.  We look forward to the next one… Left for Dead 2 anyone?

MirriAd, the embedded advertising startup that made such an impression at NVIDIA’s Emerging Companies Summit, is one of five promising companies that recently were named “Ones to Watch.”

The impressive technology can put a brand on just about anything within digital video, with placements ranging from discrete to in-your-face. Forrest Gump sits on a park bench with a subtle Coca-Cola ad placed onto it. Bill Cosby makes breakfast with a box of Special K cereal added to the family breakfast table. Seamlessly and convincingly, brand names appear on pizza boxes, wall space, billboards. It’s like having the ability to “Photoshop” video.

At ECS, NVIDIA VP of Business Development Jeff Herbst caught up with Mark Popkiewicz, MirriAd CEO, to talk about the company and its technology (scroll to the bottom of the hyperlinked web page for the video link).

The idea behind the two-year-old company is simple. There needs to be an effective, scalable, trackable way to monetize video content. As digital video recorders and on-demand video proliferate, advertisers face challenges from viewers who skip over their commercials or who ignore traditional online ads. Popkiewicz, who has a background in broadcast, knows the advertisers’ pain points very well. As he says, “the 30-second spot isn’t what it used to be.”

To look to the future, Popkiewicz and team went back to an advertising concept that’s 50 years old - product placement. But physically placing products in spots is complicated and time-consuming (an advertiser might need to commit a year in advance). It’s also a gamble – what if the spot ends up on the cutting room floor, or what if it ends up being not to an advertiser’s liking?

So MirriAd made the process easy, web-based, guaranteed and – thanks to GPUs – really fast.

MirriAd’s ZoneSense technology scans video and identifies areas for product placement in real time. Using MirriAd’s online sales platform, an advertiser sees available content inventory as well as where the brand could go and what the spot’s value is. ZoneSense video processing capabilities insert the object where desired, and the content owner and the advertiser preview and approve it (using MirriAd’s online workflow).

The advantages to advertisers are enormous. By placing brands digitally into an event after the fact, advertisers can see what exactly they’re getting and make changes based on the audience. They can know what the value of that spot is, and they can track their ROI. Content owners – especially those paying a fortune to host their video content online – finally have a convincing prospect for monetizing their inventory.

Behind the scenes, a dozen or so algorithms run on MirriAd’s hosted Tesla cluster. The algorithms enable a computer to “see” what’s in any given scene and place advertising there. A planar tracker recognizes the lighting characteristics of each zone in the video. The technology takes the lighting in the video and places it onto the embedded image, integrating it seamlessly. If the embedded object should cast a shadow given the lighting conditions in the video, MirriAd algorithms ensure that it will. (As part of its demo at ECS, Popkiewicz showed a video in which the viewer saw a brand on the side of a truck from inside a passing car – in that instance, the MirriAd technology calculated exactly how the placed image should appear given the distortion of the car’s windshield.)

For MirriAd’s future “the GPU is essential,” says Popkiewicz. With GPUs, MirriAd has been able to transform a cycle that would take months or weeks into one that now takes hours – or seconds. With Fermi, they expect to get even faster.

Five years out, Popkiewicz would like to see the technology at a point where anyone can place a brand into video in real time and have a personalized experience. So if an advertiser knows you like, say, Budweiser, it can target you by placing Budweiser in particular contexts. Potentially, MirriAd’s technology could be used to help monetize user-generated content. In that case, YouTube might even make money one day.

For now, MirriAd is working with a number of heavy-hitting broadcasters and media companies around the globe, including ABC and Discovery Networks in the U.S., and Hunan TV, Endemol, ITV, Mediaset, and many others internationally. We could tell you to look for MirriAd’s work on these networks – but since the placements don’t intrude on the viewing experience, chances are you might not even realize you’re seeing advertising.

You can learn more about cutting-edge companies like MirriAd that leverage GPU computing over in NVIDIA’s GPU Venture Zone.

NVIDIA now has the high-octane fuel ready at the pump that software developers need to transform their applications.  The NVIDIA OptiX ray tracing engine is now available for fueling, er, downloading.

As the world’s first interactive ray tracing engine, OptiX leverages the massively parallel power of NVIDIA GPUs for maximum performance and scalability. In providing a programmable ray tracing pipeline, the OptiX engine gives developers great flexibility to accelerate their ray tracing applications, bringing previously unseen levels of interactivity to a wide range of uses. These include auto styling, design visualization and visual effects. It’s also ideal for non-rendering disciplines, such as optical design, acoustical design and collision analysis.

Jeff Brown, NVIDIA’s GM for Professional Solutions, explains why OptiX is invaluable: “This opens the door to a new level of interactive realism. Ray tracing’s inherent parallelism makes it a perfect fit for GPU computing. The OptiX engine makes it easy for developers to exploit that power to create an exciting new class of applications. It enables critical design tasks -- such as examining reflections, refractions and shadow – to be performed now in real-time.”

The OptiX engine, which has been beta tested for the past six months, drastically shortens the development time required to create ray tracing apps by supplying state-of-the-art acceleration approaches that allow developers to concentrate on compelling features – not just performance.

OptiX apps will realize substantial performance gains as NVIDIA GPUs continue to advance. How? Well, NVIDIA’s current GPU architecture nearly doubled OptiX performance over its previous generation.  And tests on our upcoming Fermi GPU architecture show performance will greatly increase again.

Also, while today’s release requires NVIDIA’s professional solutions of Quadro FX and NVIDIA Tesla, the OptiX engine will soon expand its support to include NVIDIA GeForce GPUs with Fermi, as forthcoming performance will make ray tracing possible in consumer applications.

The best part?  The OptiX ray tracing engine is available free of charge and can be downloaded from the NVIDIA Developer Zone.

Want to see interactive ray tracing in action? Interactive ray tracing examples can also be downloaded and run using NVIDIA Quadro FX professional graphics solutions or NVIDIA Tesla computing solutions.

NVIDIA has several hundred engineers engaged with game developers, working to advance the state-of-the-art of gaming.  Tony Tamasi, a 10-year NVIDIA veteran, who serves as senior vice president of Content & Technology, talks in this brief video interview about his team’s and NVIDIA’s commitment to gaming.

I don’t know where this got started, but lately some people seem to think it’s okay to say that NVIDIA doesn’t care about gaming as much as it used to.  I’m baffled when I hear this, because I personally am out there racking up miles, meeting with developers and making sure they’ve got the tools, information and resources they need to take their games to the next level.

I work in NVIDIA’s Technical Evangelist group. I think NVIDIA’s technologies are critical to the development of PC Gaming (Consoles included).  And it’s my job to work directly with game programmers to make sure the PC gaming experience is a great one.  If you don't believe me, ask any game developer out there what tools do they use.  I'm fairly certain NVIDIA will listed.

A lot of folks probably don’t realize how much attention NVIDIA gives to this. I’m looking over my schedule for the next couple of weeks:  developer trips to Boston, Toronto, Montreal, Los Angeles, San Diego and Austin. I’m going to be busy, but when you have Windows 7 out, a new API and features that developers are excited about, there’s not enough time. 

I remember back in the day during the RIVA 128 launch in 1997 – there were only two or three folks focusing on game developers at NVIDIA.  Now, there are hundreds of engineers and a handful of guys like me who are responsible for evangelizing NVIDIA’s technology to developers.  If a developer has a question that I don’t know the answer to, there’s a guarantee that someone else will. So whether it’s support for 3D Vision, PhysX, DirectX 11, or any other technical matter, developers get what they need.

I’ve been at this for a long time. I’m thinking back over the more than 10 years I’ve been at NVIDIA and remembering developer trips like the Infogrames Developer Summit in the south of France;  Microsoft Meltdowns in Seattle; Developer Conventions in Sao Paulo; Tokyo Game Show in Tokyo; NVIDIA Developer events in Japan, the U.S., and London; and meetings with European Developers in Germany at the Game Convention.  Throughout those trips, not only have I managed to build long-lasting personal relationships with game developers – I’ve been able to have an effect on game development and features. And I can tell you, this commitment isn’t changing.

(Wait a sec. My nostalgia trip just got interrupted by a shipment of Razer Carcharias that came in. These are the new headphones from Razer.  Sweet!  There’s nothing like getting a chance to play with new toys!)

The next year is going to be an exciting one.  As the NVIDIA team heads out to visit with developers, we’re receiving great feedback on how PhysX can be applied to their games (some of what they come up with is stuff we’ve never even thought of!).  It’s a wonderful cycle.  NVIDIA creates technology – developers evaluate, integrate and provide feedback.  We then take the feedback and build it into future versions. And I’ll keep hitting the road, making sure we do everything we can to help developers create the best games possible.

NVIDIA just announced its “Ones to Watch” award, which recognizes five promising startups from among the 60 that took part in the Emerging Companies Summit (part of the GPU Technology Conference) in early October.

The five winners– selected for their technology, market opportunity and potential to make an impact¬– use GPU technology to create new opportunities based on visual computing, video processing and computer vision.

• IQ Engines of (Berkeley, Calif.) has technology that recognizes and labels photos, making it possible for content owners to monetize images.
• Milabra (New York City) uses computer vision to recognize images and serve up ads alongside them, much in the way that Google AdSense does for text.
• MirriAd (London) has created an embedded advertising platform for video content. This is product placement after the fact – added in so seamlessly it looks like it’s always been there.
• Ubitus (Taipei) provides a video cloud computing platform, allowing, for instance, telecoms to offer premium services around live video sharing and consumers to access rich media on any device with a browser.
• Visuvi (Redwood City, Calif.) has created a visual search technology using computer vision and machine learning that lets you search for images without text. It has applications for medical imaging as well as consumer search.

What’s interesting about this lineup is what it says about the future of image processing and the way our online experience of images and video is going to change. IQ Engines, Milabra and MirriAd each take different approaches to solving one of the most nettlesome problems on the web today: how to monetize the ever-growing numbers of online photos and video. The intensive processing demand of their technologies means their offerings become cost-effective only with the GPU.

A common theme heard at ECS was that online video content is about to explode the way that photos did a few years ago. Ubitus anticipates the virtualization of video content – where it doesn’t matter where the video is physically stored, nor does it matter what device you choose to access it with. GPUs allow Ubitus to transcode video in real-time, streaming content across different networks, codecs, bandwidths and resolutions.
 
And in Visuvi’s (Chris Boone, CEO of Visuvi, receiving award) case, medical imaging is another area where GPU technology is poised to have a major impact. Visuvi’s algorithms allow for instance, doctors to take an image that might show cancer and feed it into the image search, returning images that are similar in order to make a more accurate diagnosis.

Back in March, NVIDIA launched its GPU Ventures program with the goal of identifying new companies that leverage GPU technology and supporting them with everything from direct investment to networking opportunities. With this award, GPU ecosystems just got a little healthier.

There was lots of news at today’s Windows 7 Launch by Microsoft in New York. Something that may have gotten lost in the larger storm was that NVIDIA received a partner award for how the NVIDIA ION chip is changing tiny PCs.

By combining an NVIDIA ION GPU with an Atom processor, we are able to deliver premium PC performance and features in a low-cost, small form factor.

Netbooks –  inexpensive, highly portable devices that act somewhat like notebooks but lack their oomph – have had a reputation for being weak, and incapable of running today’s more demanding applications.

Thanks to ION -- a small but high-powered GPU that runs together with a low-end CPU -- that’s no longer the case. I can’t put it much better than MobileTechReview.com:

“If we gave Editor's Choice awards to technologies rather than products, we'd give one to NVIDIA for their ION platform. It will shift the netbook from web surfing and email machine to a movie and gaming portable.”

Introduced earlier this year, ION-based devices have blurred the distinction between netbook and notebook. 

They provide new capabilities you’d never find in your mom’s netbook -- HD video, accelerated media conversion, faster overall performance under Windows 7 and gaming capabilities. They can also run Flash video. On ordinary netbooks, SD or HD flash videos are simply unwatchable. We’re working with Adobe to accelerate Flash across all NVIDIA GPUs to deliver a seamless, stutter-free experience on SD and HD videos on your favorite sites. But they’re priced far under a notebook.

Windows 7 launched today and it is sure to have a big impact on NVIDIA.
 
Microsoft’s new OS puts the GPU firmly in the spotlight because it marks a dramatic shift in the way the Windows operating system can take full advantage of the GPU, using it for both graphics and computing.

We consider Windows 7 DirectCompute a tremendous opportunity to broaden the reach of GPU computing.  DirectCompute will take GPU computing from a niche to the mainstream by making this technology available to the millions of users who will be adopting the Windows 7 operating system.

Windows 7 also brings with it a new PC graphics API – DirectX 11, which NVIDIA is very excited about.  It is the next great iteration of DirectX for PC gaming and will be the API of choice for some of the top gaming titles starting in 2010. NVIDIA’s announced DirectX 11 support in the Fermi architecture to bring these new gaming titles to life.
 
There’s been some noise implying that we aren’t 100% behind the industry standards that Windows 7 introduces – DirectCompute and DirectX 11. But that’s clearly untrue.  We are supporting all standard APIs and have demonstrated our support for DirectCompute by being the first to release support for DirectCompute in the GPUs we ship today.

There’s also been confusion about NVIDIA CUDA. Some are saying it competes with DirectCompute, and that’s also untrue. CUDA is the name of NVIDIA’s parallel computing hardware architecture.   Programs written using DirectCompute run on the NVIDIA CUDA architecture.   We support whatever language developers choose to enable these new innovations.

We not only support open standards like DirectCompute and DirectX 11, but we also support standards that allow NVIDIA to offer new innovations well in advance of industry standards, such as CUDA C and PhysX. We don’t believe it’s in our customer’s best interests for us to delay innovation. Our goal is clear: We are committed to leading the industry in amazing new directions and creating value for our customers.

NVIDIA has worked very closely with Microsoft on Windows 7 to streamline its own internal development processes. As a result, NVIDIA had stable, full performance, fully featured, WHQL-certified drivers five months before the launch of Windows 7. The transition to Windows 7 has been simplified for the entire industry because the Windows 7 driver model is an evolution of the Vista driver model. In contrast, when Vista first launched, the driver model was completely new.

We also hope Windows 7 gives the PC industry a much needed shot in the arm.  Industry analysts are hopeful that Windows 7 will increase PC spending. Research firm IDC predicts that around 177 million copies of the operating system will be in place by the end of 2010, and estimates that products and services surrounding Windows 7 will generate $320 billion.

We’re glad that it’s finally here and we hope you enjoy it as much as we will.

Windows 7 is a great operating system that is exactly what the PC industry needs. But two areas where Windows 7 will make a large impact really get our juices flowing - GPU computing and power gaming.
GPU computing uses the massively parallel architecture of the graphics processing unit (GPU) as a computational engine that can be programmed with high-level languages and APIs. Microsoft DirectCompute is a new API in Windows 7 that uses the GPU to accelerate applications..

DirectCompute gives software developers the ability to create both powerful new digital-media applications and next-generation PC games by harnessing the power of NVIDIA GeForce GPUs. DirectCompute is initially being introduced as part of the Microsoft DirectX 11 API, but works on both NVIDIA’s DirectX 10 and DirectX 11 GPUs.

We’ve already seen some applications come to market that take advantage of the GPU, and they have shown dramatic improvements in speed compared to CPU-only performance.  Elemental Technologies Badaboom media converter, Motion DSP’svReveal image stabilizer and CyberLink’s video applications are all prime examples.  DirectCompute will make it even easier for developers to write applications that turbo-charge digital media applications.  We believe DirectCompute will take GPU computing into the mainstream and make the benefits available to millions of users.

As for power gaming, DirectX 11 is going to make the PC gaming experience even better. At its core NVIDIA is about gaming, and we’re 100 percent behind DirectX 11. Here are some of the ways that DirectX 11 is going benefit power gaming.

• GPU tessellation - Tessellation allows the GPU to draw far-away models in low detail and close-up models in high detail while maintaining performance. The end result is an optimally tessellated mesh for each view distance.
• Multi-threading - By spreading the CPU load over multiple cores, developers can remove CPU bottlenecks and take advantage of GPU horsepower for graphics and effects. Developers can use multi-threading to immediately improve performance in their games.
• Improved texture compression - DirectX 11 adds two new texture-compression formats: one for high-dynamic range (HDR) textures and the other for better visual quality on low dynamic range textures.
• Shader Model 5.0 - This improves programmability for developers by adding features like double precision and increased resources and subroutines.

Looking ahead, our next-generation Fermi-based GeForce GPU will support DirectX 11, along with technologies like NVIDIA PhysX and NVIDIA 3D Vision. Fermi is the first GPU designed from the ground up for compute, and it will also be the fastest graphics processor in the world when it ships. Together, Windows 7 and Fermi-based GPUs are set to deliver a new level of immersion and interactivity that will transform the gaming industry.