In case you were not paying attention
I usually don't post things vendors send to ClusterMoney because we are not a big time news site (nor to I care for marketing piffle). I will at times summarize some important news and events with enough links to send you on your way to HPC enlightenment. Each Year NVidia sends me a year in review which is a good summary of Tesla HPC events -- complete with many URLs so readers can explore further. [A note to vendors: Company news (aka press releases) with URLs and good background and no jargon may get posted here.] The NVidia round-up begins below:
NVIDIA Tesla - 2009 Year in ReviewGPU Computing had a ground breaking year in 2009. In just two and a half years from its launch, the Tesla brand has truly established itself in the HPC community. This wouldn't have happened without the efforts of real GPU Computing pioneers such as Prof. Wen-mei Hwu at University of Illinois who taught the very first courses in parallel programming on the GPU and Prof. Satoshi Matsuoka at Tokyo Institute of Technology who put the first Tesla GPU-enabled supercomputer onto the Top 500 (Top 30 in fact), just one and half years after we launched the brand.
A "tipping point" is defined as a level at which momentum for change becomes unstoppable - we genuinely believe that we are witnessing the tipping point for GPUs in the high performance computing space and the SC09 conference in Portland, Ore. in November cemented that belief.....but we'll come to that in due course :)
Here are our Top 10 takeaways for the year:
- NVIDIA invested heavily into the CUDA Center of Excellence (CCOE) program in 2009. This program seeks out universities that are breaking boundaries in both their educational programs and their research projects and embracing GPU Computing and the CUDA programming model to achieve their goals. 2009 saw Harvard University, National Taiwan University, Chinese Academy of Sciences, Tsinghua University, University of Tennessee and University of Maryland join the program and we look forward to seeing their research advance as a result. A total of nine universities are now CCOE, but with more than 270 universities around the world teach the CUDA programming model today, I am sure we'll see this number increase this year.
- One of the big stories for 2009 was software - with so many people wanting to take advantage of the GPU, we spent (and continue to spend) a great deal of time developing a range of tools to ease developers' transition to parallel programming. In 2009 we launched the world's first debugger and profiler for GPU Computing in CUDA 2.2 and we announced our integrated development environment for Windows Visual Studio, codenamed Nexus. OpenCL was also released this year and NVIDIA has taken a front seat in this upcoming open standard for GPU Computing as well. We were the first to pass conformance, first to get drivers into developers' hands, in fact NVIDIA is still the only vendor supporting OpenCL features beyond the minimum conformance level, such as double precision and OpenGL interoperability. At SC09, we introduced the beta of the CUDA 3.0 toolkit which enables developers to start writing applications today for our upcoming GPUs based on the "Fermi" architecture.
- With all these new tools and CUDA enhancements now available for developers
to use, it's no surprise that 2009 saw the development community really start
cooking on gas! Our GPU Computing registered developer program nearly quadrupled in
size, going from 2000 to more than 7400, and these developers are posting new codes
and research papers nearly every day. Fortunately, this community is great at making
their work public, so just check out CUDA Zone or search
for CUDA on
to see hundreds of examples. Here are a few of my favorite CUDA stats of the year :)
- 2700+ CUDA-related citations on Google Scholar
- 800+ CUDA-related videos on YouTube
- 670+ submissions to CUDA Zone by CUDA community
- 350+ registrants for current CUDA Superhero Challenge
- The Tesla Personal Supercomputer continues to be a resounding success with the computational research community and this year the two largest OEMs, Dell and HP, announced Personal "Supers" of their own. We also partnered with SuperMicro in June to bring the world's first hybrid GPU/CPU 1U system to market. From single cards and multi-GPU personal supercomputers to hybrid and GPU-only 1U systems, the Tesla product family has grown into a complete top-to-bottom range of HPC solutions to suit any installation and budget.
- There can be no doubt that THE event for the GPU Computing community in 2009
was our inaugural a
GPU Technology Conference or GTC. It was a phenomenal event for NVIDIA and GPU Computing and brought together more than 1500 GPU Computing professionals from the world of scientific research and industry - almost twice the number we predicted in fact. More than 130 hours of presentations took place during the 3 day event, and NVIDIA only took up around 30% of them - the rest featured real developers, real researchers, doing amazing work with the GPU. In nearly 9 years of being with NVIDIA, this was by far the most exciting atmosphere I have experienced at a developer conference, and that was an attitude shared by many of our visitors, like John Leidel from insideHPC
"I've been attending technical conferences from coast to coast for a number of years and never have I experienced the electricity that I felt at GTC. Never have I seen the (often fickle) members of the HPC community latch on to a technology so fast and with such ferocity. All in all, its pretty amazing."
- Obviously a major highlight for 2009 was the introduction of Fermi, our next generation CUDA GPU architecture. This is one of the most exciting GPU architectures we have ever built, with a feature set that pretty much checked every requirement box for the high performance computing industry, as was evidenced by Jeffrey Nichols from Oak Ridge National Labs standing on stage at GTC and talking about their upcoming Fermi-based project that when deployed, will be 10- times more powerful that the today's fastest supercomputer. It's going to be an awesome graphics processor as well, but I'm going to leave the details on that to the GeForce guys. Tesla GPUs based on Fermi are still on target for Q2 2010 and we're really excited to get them out in the market.
- Every day in the Tesla business we come across a new story, a new use case and more often than not, it's in a field that completely takes us by surprise. A great example of this was the work that Lowry Digital was doing to restore the Apollo 11 footage. The original footage of the first steps on the moon was accidentally deleted, so by using GPUs and their own special algorithms, Lowry Digital restored TV footage, even footage taken from cameras pointed at NASA monitors during the moon walk, and turned them into HD quality images to ensure that history always has a record of that momentous day. Pretty cool stuff. Other companies that spoke publically about their work with Tesla GPUs included BAE Systems, Beckman-Coulter, Bloomberg, BNP Paribas, Los Alamos National Lab, NASA, Petrobras, Sandia National Lab and many more.
- 2009 also saw an explosion in the GPU Computing developer and ISV ecosystem. Tool providers such as the Portland Group, TotalView, Allinea, CAPS and EMPhotonics released CUDA optimized versions of their solutions. New libraries critical for high performance computing applications such as BLAS, LAPACK, FFT, NPP and more are now also available and, perhaps most exciting, more than a dozen bioscience codes used by approximately 250,000 computational scientists are now available. These are the applications that are speeding up areas of research such as drug discovery and DNA sequencing, areas that can truly change the world we live in.
- The cloud remains a hotly debated topic in the media and we were excited to start talking about our own efforts in this area in 2009. RealityServer is the world's first GPU-based platform for cloud computing that enables the streaming of photorealistic 3D applications to any web connected PC, laptop, netbook or even a smartphone. As someone who is personally re-designing a kitchen, the ability to view interactive, photo-real images of my exact kitchen that I can change on the fly and solicit feedback from my mother-in-law just by showing her my iPhone, I think is pretty cool. Expect to hear a lot more about this platform this year.
- Finally in November, we attended the Supercomputing (SC09) conference in Portland, Oregon. This was by far the most exciting SC conference for us to date. We showed off early samples of the new Tesla 20-series family of GPUs based on the Fermi architecture, we packed the booth every day for our in booth theater where industry luminaries such as Jack Dongarra, Satoshi Matsuoka, Jeff Vetter joined NVIDIA engineers to talk about GPU Computing. On top of all that we won 5 industry awards from HPCWire including a Readers Choice for Fermi, which we always love to get, and Dr. Tsuyoshi Hamada of Nagasaki University, Japan won the prestigious Gordon Bell award for best price/performance in an HPC application by using a 256 GPU cluster to run astrophysics and fluid turbulence simulations. We blogged every day during the show, so if you missed any of the news, go ahead and check out the stories here.
The NVIDIA Tesla Team