Wednesday, 22 September 2010

nTersect

nTersect

GTC Day 2 Keynote: GPUs Enabling Molecular Biology Research

Posted: 22 Sep 2010 11:13 AM PDT

The impact of NVIDIA's GPU technology has reached into the molecular frontier, enabling scientific researchers to observe everything from protein folding to photosynthesis in more detail than ever before. Such was the message renowned computational biologist Dr. Klaus Schulten sent during his morning keynote presentation on day two of NVIDIA's GPU Technology Conference. Since 2006, Schulten and his team of researchers at University of Illinois, Urbana-Champaign, where he's a professor, have tapped the power of GPU-powered computing to get a better look at some of the molecular processes that have long vexed the scientific community. "We've benefitted a lot from this technology," Schulten said. "It's led to several scientific discoveries."

GPUs are aiding Schulten's research in three important ways: by increasing the speed of simulations, improving their accuracy and opening doors to new fields that were unfathomable using GPU technology. They're doing this by enabling far more powerful microscopy, and, by pairing GPUs with powerful molecular dynamics codes and molecular visualization programs, enabling heretofore time-consuming computational analysis of these detailed images.

Here are some of the ways Schulten's team has put GPUs to work:

  • They've discovered Tamiflu drug resistance in the H1-N1 Swine Flu virus. Also were able to observe that a two-step binding process is necessary, via microscopic views far greater in detail than anything possible prior to GPGPU. This information will allow them to reconfigure Tamiflu to make it more effective in battling H1-N1.
  • They've made important discoveries about how viruses attach themselves to cells and respond to atomic microscopy, during which the viruses are prodded to see how they respond to pressure.
  • They've shortened to 90 seconds from one hour the time required to analyze electrostatics, which provide electrical charges during the photosynthesis process, opening the door to new insights on harnessing solar energy.
  • They've observed ribonomes in action—as opposed to the low-resolution static images possible via CPU-only configurations—providing a snapshot of a protein being born. Schulten likened this to watching a football game rather than just seeing pre-game pictures of the teams.
  • They've detected subtle differences in DNA strands, providing more information on the genetics of diseases like cancer and depression, and have reduced the time required to compute radial distribution functions, which are measurements of atomic density, from 15 hours to 10 minutes.
  • They've rendered electron clouds in real time, speeding the computing of molecular dynamics by up to 400x. By comparison, CPUs require an entire working day to get a snapshot of an electron cloud.  "This turns out to be heaven for GPUs," Schulten said. "(Chemists and GPUs) go hand-in-hand—they love each other."
  • They've observed the process of protein folding in unprecedented details, an important advance in being able to combat diseases that result from the unfolding of proteins.

Not surprisingly, such widespread impact led Schulten to conduct a recent workshop on how GPU computing can be used to further molecular dynamics. Based on the impact GPUs have had on Schulten's work thus far, the sky's the limit.

GPU Technology Conference: What They're Saying (Day 1)

Posted: 21 Sep 2010 06:37 PM PDT

Big day today for us, as the GPU Technology Conference officially started with a bang. While we've done a lot of our own reporting here on the blog, Twitter and Facebook - we wanted you to get an idea of what others were saying about the show. Here's a short roundup of some of today's coverage:

GPU Technology Conference – Day 1 Video Recap

Posted: 21 Sep 2010 06:23 PM PDT

The epicenter of the GPU computing revolution has shifted this week to the San Jose Convention Center. It's serving as home of this year's GPU Technology Conference, where several thousand attendees are learning more about what researchers are doing on the GPU front from medical to video animation. In this video wrap-up from day one, Ujesh Desai, VP of product marketing, provides highlights from the show floor.

  

GPU Acceleration in Large-scale CFD on the Tsubame Supercomputer

Posted: 21 Sep 2010 06:11 PM PDT

Professor Takayuki Aoki from the Tokyo Institute of Technology just finished giving a fascinating look under the hood of the university's Tsubame supercomputer, which has been used (among other things) in collaboration with a number of agencies in Japan to provide complex computational fluid dynamics modeling.

The Tsubame is notable because it leverages GPU clusters, and its success is one of the milestones for GPU's in supercomputing. The Tsubame 1.0 uses Tesla S1070 in a 680 GPU cluster. With it, scientists have been able to experience speed ups of up to 80x in problems like weather modeling. Coming in December, the next-generation Tsubame 2.0 will use the Tesla 2050 with 4224 GPUs and provide performance of more than 3 PFLOPs.

Performance metrics are an important scorecard with supercomputers. But Professor Aoki explained how performance achieved really depends on the application as well as tuning and optimization – and gave some insights into how his group has been able to improve results. With full GPU implementation, Tsubame experienced acceleration of 10x to 100x over CPU-only performance, but notes that the numbers depended on the application.

With such intensive computation, any performance increase makes a huge difference – as do any bottlenecks. Communication issues between GPUs over cluster nodes can create some additional overhead, and Professor Aoki explained an overlapping technique they've developed at Tokyo Tech to deal with this.

While sessions was highly technical, audience members of all comfort levels could appreciate the incredible demos of CFD simulations, including water splashing into a container, a milk crown, and dendrite solidification in metal.

CUDA Boosts Video Editing with Adobe Premier Pro

Posted: 21 Sep 2010 05:37 PM PDT

Video editing is one of the disciplines making striking use of CUDA technology, as presenters from Adobe Systems Inc. made clear during a session at the GPU Technology Conference. Al Mooney, product manager for Adobe's Premier Pro video editing application, and computer scientist Steve Hoeg showed a room full of video professionals how the parallel processing capabilities of NVIDIA's GPUs are making it possible to enjoy peak performance while tapping the full capabilities of Premier Pro.

Video editors face numerous processing challenges, including huge data streams, hundreds of formats, and increasing pressure to deliver bigger results in shorter timeframes. Despite these challenges, editors across the board want to be able to play back any format or frame rate without conversion, mix multiple formats and frame rates within the same project, apply visual effects without slowing down the project, and deliver to multiple output formats quickly.

Steve Hoeg, left, and Al Mooney, fielding a technical question during their session--Adobe Premier Pro screen in background

This combination of factors is pushing the processing needs of video editors to new levels, and CUDA is helping Premier Pro, which represents the largest commercial CUDA deployment to date, deliver the goods.

As an example of the profound impact CUDA is having, Hoeg noted that GPUs are able to process color correction 75 times faster than the latest Intel Nehalem processors. "CPUs just cannot cope with this," Hoeg said.

Additionally, whereas CPUs have struggled to simultaneously apply video effects while decoding source files, Hoeg said that offloading the effects functions to GPUs enables CPUs to handle the decoding easily. What's more, CUDA makes it possible for users of Premier Pro to work straight from those source files rather than having to convert them to a pre-determined format.

"CUDA has allowed us to do a lot of things previously not possible," said Hoeg.

To nail that point home, Mooney demonstrated how Premier Pro handles playback of five simultaneous video streams. First, he shut off the GPU acceleration and showed how the software would freeze and blip during the playback of a superimposed figure standing before four videos running in separate quadrants in the background. Then, once he turned the GPU acceleration back on, playback proceeded seamlessly.

The implications were clear: CUDA is taking the world of video editing to places it's never been by allowing Adobe to build ever-more powerful editing tools into its software while also speeding up complex computing processes.

If that's not innovation, then I don't know what is.

WebGL Brings GPU Accelerated 3D to a Browser Near You - Presented by Mozilla

Posted: 21 Sep 2010 04:58 PM PDT

Imagine interacting with silky-smooth 3D applications all within your Web browser. Vladimir Vukicevic, principal engineer at Mozilla is working to make that a reality with the WebGL API. He's been involved with WebGL since its creation and during the GTC session titled, "WebGL: Bringing 3D to the Web", he spoke about the emerging standard and what's necessary to make it a viable platform in the next generation of Web applications.

Originally called Canvas 3D, WebGL essentially allows 3D graphics to be implemented into a Web Browser without the use of plug-ins by rendering content in a HTML5 Canvas. Sounds great, right? Vladimir revealed that future versions of Firefox, Chrome and potentially Safari will include and support WebGL. With HTML5 already taking advantage of the GPU, and more browsers becoming GPU-accelerated, the challenge of delivering compelling 3D content on the Web should quickly become a reality.

Medical CUDA Applications Abound Day 1 at the GPU Technology Conference

Posted: 21 Sep 2010 04:30 PM PDT

The use of CUDA for accelerating medical techniques is one of the major themes at the GPU Technology Conference today. The focus on medicine began in Jen-Hsun Huang's keynote this morning, as he showed amazing video of robotic, minimally-invasive surgery on a patient's beating heart.

In addition to cardiac surgery, researchers from several prominent universities also presented technical sessions today on how they are using CUDA to accelerate medical imaging. Here are a few of today's sessions:

  • Algorithms for Automated Segmentation of Medical Imaging Studies Utilizing CUDA – Researchers at the University of Pennsylvania explained how GPU computing can help doctors make sense of modern imaging studies. They mainly focused on algorithmic approaches to segmentation as it pertains to computed tomography angiography studies.
  • Unveiling Cellular & Molecular Events of Cardiac Arrhythmias -- George Mason University talked about using CUDA technology to get a 20x speed-up in simulations of intracellular calcium dynamics, thought to play a major role in the generation of cardiac arrhythmias.
  • Nearly Instantaneous Reconstruction for MRIs – GE Global Research shared research on a computationally intensive, widely used algorithm in MRI Reconstruction using Parallel Imaging. They showed that an optimized CUDA implementation of their algorithm on a GPU can enable nearly instantaneous reconstruction and speedups of up to 10x over an optimized dual socket QuadCore CPU implementation.

GPU Technology Conference Session: A Workingman’s Guide to 3D Video Editing

Posted: 21 Sep 2010 04:12 PM PDT

With 3D content proliferating, more and more people are becoming 3D content creators. The audience of "A Workingman's Guide to 3D Video Editing" got a high level walkthrough of the ways and hows of creating and editing 3D video, along with specific tips and guidelines from NVIDIA's Ian Williams and Kevan O'Brien.

GTC 3D video editing session - Ian Williams

High-end 3D video still belongs to the movie studios, but prosumers are getting more options as 3D cameras come on the market and 3D TVs appear in more living rooms. GPUs are playing a role in the proliferation of 3D as well, since the accelerated processing power makes it dramatically easier to edit 3D content and perform some of the computations needed to reconform content after editing. And at the high-end, you definitely need GPUs, or you're facing unworkable processing times. As Kevan put it, "You can't do real-time 2K or 4K work without a GPU behind it."

Ian, the director of NVIDIA's applied engineering group, took us through the ways that 3D video can add value and can be distributed (for instance, YouTube is a boon to the novice 3D videographer). Some of the guidelines and best practices:

  • go for a camera that gives you want full manual control, avoid high-contrast because you'll get ghosting on LCD technology
  • group objects and people close together because the lens will naturally make objects appear farther apart
  • don't do a lot of close-ups, and pan slowly – otherwise you risk giving your audience nausea from the movement.

According to Ian, "It's all about subtlety in 3D."

The real highlight of the session was when Kevan, a filmmaker, showed exclusive previews of a new work in progress, "The Time Machine." The entire audience put on NVIDIA active shutter 3D Vision glasses and got to see the editing process in real time. Very cool.

Poster Hall at GPU Technology Conference Highlights New GPU-Powered Research

Posted: 21 Sep 2010 02:13 PM PDT

In addition to the plentiful lineup of speakers and bustling exhibition floor, the GPU Technology Conference is providing attendees the opportunity to learn more about what researchers are doing on the GPU front. Displayed in the San Jose Convention Center concourse are dozens of poster boards illustrating countless research projects stretching the GPU frontiers.

There are descriptions of the impact GPUs are having on everything from cardiac function simulations and vision-enhancement systems to black hole simulations and optimized speech recognition technologies. Poster owners will be available to discuss their projects Tuesday evening at 5:00pm PT.

Ade Olubummo studying Stanford 3D Vision Lab's motion detection research poster board

While I was perusing the posters, I ran into Ade Olubummo, founder of Bootpin Inc., a Chicago-based startup that's building an activity-recognition application. He found himself drawn to a poster board detailing work Stanford University's 3D Vision Lab has been doing in the area of motion-detection, an important component of Bootpin's software. Olubummo said he's been evaluating such technologies, and that what he saw convinced him that he needed to get in touch with the 3D Vision Lab to find out more.

Bootpin's application targets functions in settings that Olubummo described as being "wherever human eyes are deployed." That could mean anything from a warehouse security operation to a geriatric watch unit.

Malik Khan of University of Utah describing his research

Elsehere, Malik Khan, a researcher in the computer science department of University of Utah, was sizing up the competition. Khan authored a poster board depicting the work he's been doing on transferring the optimization of sequential code libraries for compatibility with GPUs. That, in turn, will allow designers of scientific applications to benefit from the performance gains of GPU technology without having to re-write their applications.

As Khan surveyed a poster board on automatic program generation produced by a team at Carnegie Mellon, he was intrigued with applicable details he could borrow from in his research. "This work seems to be focused on the same kinds of scientific applications," he said.

Despite the competitive nature of the work, Khan said there's a lot the two research efforts can learn from each other.  This is just one example of how attendees are learning about the best and the brightest in GPU Computing, while at the GPU Technology Conference.
Posted by at

No comments:

Post a Comment

Subscribe to: Post Comments (Atom)