A Blog about the World of Graphics Processing

One of the things that always impresses me when I attend a LAN Party like GeForce LAN or Quakecon is the huge amount of high quality case mods I always see in the BYOC section. Some of these systems can take hundreds of hours to put together and end up costing many thousands of dollars to create – but for die hard case modders that make them, the time and cost is well worth the satisfaction of creating a one of a kind gaming system.

At about the same time I went to Quakecon this past summer, I started to get involved with The Leukemia & Lymphoma Society - the world’s top non-profit blood cancer research organization. I soon realized that combining case mods and charity would be a fun way to raise some money for a great cause – and the charitycasemod.com project was born.

I consulted the help of a top case modder named Richard Surroz and started asking companies in the PC ecosystem if they’d be interested in donating parts for our charity case mod project. The response was astounding. 11 Companies in all donated their parts or services to the project – helping us create a one of a kind NVIDIA themed gaming system that is worth more than $10,000!

The system is complete with a NVIDIA 3D Vision glasses and Samsung monitor bundle, two NVIDIA GeForce GTX 295’s in SLI, an Intel Core i7 975 Extreme Edition CPU, 12GB of DDR3 Crucial Ballistix Tracer memory, two 256GB Crucial SSD hard drives, two 1TB Western Digital Caviar hard drives, a custom water cooling kit from Danger Den, and a one of a kind NVIDIA themed paint job from Smooth Creations.

Click here to check out the auction. The auction for the Ultimate NVIDIA Gaming System ends on Tuesday, November 24th at 8PM PST. 100% of the final auction price will go directly to The Leukemia & Lymphoma Society and the auction is 100% tax deductible as well.

Andy Keane, general manager of NVIDIA’s Tesla business wrapped up SC09.

“SC09 was a huge show for NVIDIA GPUs. Walking around the show floor, Tesla is everywhere, in clusters, in OEM booths and in demos being shown in dozens of research booths,” said Keane.

“Our booth theater was a big hit with attendees.  Our presenters did a fabulous job of talking through their work and highlighting where GPUs have helped them get major boosts in performance. In fact we had so many people for the session on our new Fermi GPU architecture that we took over the aisles and had many attendees sitting on the floor – thanks to everyone that came and listened, we hope you enjoyed the sessions and we promise to bring more chairs next year.”

NVIDIA also enjoyed a tremendous turn out for its all day tutorial on CUDA with more than 200 developers coming along to hear from NVIDIA and domain experts on how to get the most out of GPU Computing.

SC09 also saw NVIDIA highlight the very latest in 3D Internet technology with its RealityServer platform, which along with the iRay software, allows ray traced 3D images to be transmitted in near real-time to any web capable device, even a smartphone.

With nearly half of the 200 or so exhibitors showcasing NVIDIA Tesla products and talking about their use of GPUs, and with a constant flow of traffic that never let up across all 4 days of the exhibition, there can be no doubt that GPU computing was a central theme of SC09.

In his presentation at NVIDIA’s SC09 booth, Dr. Jack Dongarra, described his latest research and how he views the evolving HPC architecture.  At the University of Tennessee, Dongarra has expanded his research from traditional HPC computers with multi-core CPUs to hybrid computers that include GPU accelerators— where the GPU and CPU cooperate in a coprocessing model to greatly increase the performance of parallel applications. 

Dongarra indicated that he views scientific, HPC computing as a continuum, with researchers doing computing on PCs equipped with GPUs all the way up to multi-core systems capable of performing exascale calculations. “Small PC-based systems with GPUs are important in research because they are cost effective, accelerate processing speed and also save researchers time,” states Dongarra.

Dongarra believes that the scientific HPC community must take on the challenge of determining how to incorporate hybrid architectures into the HPC environment. “I am encouraged by the performance of GPUs and look forward to GPU accelerators becoming part of the process. But our scientific community must develop an understanding of the impact of GPUs and create standards relating to GPU processing.  I look forward to working on these standards,” states Dongarra.   

Researchers at NASA want to get a better understanding of what happens when galaxies collide.

At SC09, NASA demonstrated the computation and visualization of dust grain equilibrium temperatures using NVIDIA GPUs.  When calculating the infrared spectral energy distributions (SEDs) of galaxies in radiation-transfer models, the computation of dust grain temperatures is generally the most time-consuming part of the calculation. Because of its highly parallel nature, this calculation is perfectly suited for GPUs.

Researchers there developed their own Monte-Carlo radiation transfer code, called “Sunrise”, using CUDA.  With an NVIDIA GPU, NASA can perform this calculation 55 times faster than using two quad core CPUs.  

Bryan Green of NASA’s visualization team (who gave us a wonderful demo) stated that a big reason for his enthusiasm of GPU computing is the CUDA Linux development environment NVIDIA has provided.

It’s official! With U.S. Patent no. 7609272, NVIDIA hit a major milestone – 1,000 patents – and the company’s rich IP portfolio continues to grow in strength.

The patent, which was awarded October 27, is for an invention that makes the pixel processing pipeline faster and more efficient. In particular, it helps the shader process textures in a way that makes full use of any extra circuits, speeding up output.

Previously, when a large texture needed to be read, one instruction would be issued, and one shader circuit would need to make several passes while other circuits sat idle. But patent authors Emmett Kilgariff and Rui Bastos (pictured below) –both longtime NVIDIANS – figured out a way to allow for a partial texture load. By breaking the texture load into smaller pieces – able to be completed in one pass each – all circuits can keep firing.

As Bastos recalls, the idea came from asking themselves, “What can we do to reduce the number of cycles required to run a program and get applications/games running faster? The key for the idea was an application of the divide-and-conquer principle.”

Textures can be 32-bit, 64-bit, or 128-bit. But anything larger than 32-bit requires more than one pass. Before Bastos and Kilgariff’s invention, texture lookups were monolithic instructions that took multiple cycles to be executed, leaving other shader functional units to sit idle. “The idle units in the pipe presented the opportunity to try to fit other, non-texture instructions in those slots – i.e., run more than one instruction per cycle,” says Bastos. But to do that, the monolithic texture-load instructions had to be split into chunks. Break a 128-bit texture into four pieces – each of which can be completed in one pass – and that lets one cycle-hungry instruction be broken into four instructions. Doing this means that other circuits keep processing instructions – no more waiting.

In addition, Kilgariff and Bastos discovered they could reorder instructions for greater efficiency. For instance, if a texture for instruction 1 is not immediately available, the shader circuit could get to work on instruction 2. Instructions don’t back up in a queue.

By providing for partial texture loads and reordering instruction sequences for greater efficiency, Kilgariff and Bastos found they could reduce the number of required passes. Ultimately, textures render faster and game play is more seamless.

The invention made its way to the market in 2004 in the GeForce 6 family of products. It also featured in the RSX – or Reality Synthesizer – GPU that NVIDIA co-developed for the Sony PlayStation 3. Thanks to Kilgariff and Bastos’s discovery, gamers were able to enjoy a level of detail never seen before.

In the NVIDIA booth yesterday, we had a chance to hear from Klaus Schulten and John Stone from the University of Illinois. They demonstrated how they use NVIDIA GPUs with NAMD and VMD research software to simulate and visualize cell structures.

Their research focuses on a variety of health issues but one important aspect of their research is centered on cell disruption and viruses, including the H1N1 virus. Their H1N1 research focuses on the structure of the that virus, how it reacts to drug treatment, and what may be happening in relation to drug resistance.  

Schulten went on to say that the swine flu epidemic provided a chance for theoretical scientists to apply ‘emergency computing’ to help a real world problem. 

Schulten and Stone use supercomputers but also use PCs running NVIDIA GPUs in their research. Schulten indicates that a problem in traditional computing where processing is done on the CPU is that processing speeds are too slow to adequately see the living cell. 

“When we use systems with GPUs running NAMD and VMD software, this speed is accelerated and we can do simulations of cells. Important biomedical cellular research problems can be solved by the acceleration offered by GPU chips. With NVIDIA GPUs, our calculations can be done between 200 – 400 times faster on the GPU,” states Schulten.

In his presentation, Stone also described how desktop or laptops with NVIDIA GPUs are used in post-processing analysis of gas molecules.  “Calculations that would run for a month can now be done in a day.  In addition, our researchers can do calculations on a laptop in a few minutes.  This allows us to perform calculations and research that previously would have been too time consuming to do,” states Stone.  

In his talk in NVIDIA’s theater, Ross Walker described the research being done by the San Diego Supercomputer Center (SDSC) and how NVIDIA GPUs are helping with this research.  The SDSC team performs research on biomolecules, biofuels, and flu viruses.

“My current research is directed towards developing new molecular dynamics techniques for studying protein reactivity and reaction pathways with the goal of furthering the understanding of such complex mechanistic processes in order to aid the development of new drugs and to improve current biofuel technologies,” states Walker. The research involves high speed simulations using AMBER, a widely used package of molecular simulation programs, used to study the dynamics of biomolecules.

SDSC is using computers running NVIDIA’s Tesla GPUs in their research. Walker reports that “The NVIDIA CUDA architecture accelerates AMBER simulations by up to 100  times when compared with a CPU.”

In addition, using GPUs has changed how research is done. Traditionally, researchers had to submit simulations to a queue and had to wait several hours or up to 2 or 3 days for the results. If there was a problem during the run, then the work had to start all over. With desktop computers using NVIDIA Tesla GPUs and CUDA software, researchers can run simulations and see the results quickly. “Using NVIDIA tools allows SDSC researchers to complete simulations in a reasonable time, we can work interactively with the data, see results quickly and make modifications in a reasonable timeframe,” states Walker.

In his talk to a packed crowd at SC09 today, Jeffrey Vetter (Oakridge National Labs researcher and professor at Georgia Tech) described the upcoming Keeneland system. The Keeneland Project is a grant awarded by the National Science Foundation (NSF) for the deployment of an experimental high performance system.

According to Vetter, the Georgia Institute of Technology (Georgia Tech) and its partners, the University of Tennessee at Knoxville and the Oak Ridge National Laboratory will acquire and deploy the Keeneland system. It will be an experimental, high-performance heterogeneous computing system consisting of HP servers integrated with NVIDIA Tesla GPUs.

(Slide courtesy of Jeffrey Vetter, ORNL)
“In evaluating solutions for the Keeneland, we looked at a variety of experimental architectures such as GPUs and FPGAs and determined that GPUs provide a great solution. We selected NVIDIA’s Tesla GPUs to run on the Keeneland system.”

Vetter indicates that the Keeneland project team will start by developing scientific libraries and programming tools to facilitate the development of science and engineering research applications. As the system is developed, the project team will provide consulting support to researchers who wish to develop applications for the system using OpenCL or to port applications to the system. Eventually, the system will be open to other scientific researchers and may even be available to educational research work.

NVIDIA and Mellanox introduced new software yesterday that will increase cluster application performance by as much as 30% by reducing the latency that occurs when communicating over Mellanox InfiniBand to servers equipped with NVIDIA Tesla GPUs.

The system architecture of a GPU-CPU server requires the CPU to initiate and manage memory transfers between the GPU and the InfiniBand network.  The new software solution will enable Tesla GPUs to transfer data to pinned system memory that a Mellanox InfiniBand solution is able to read and transmit it over the network. The result is increased overall system performance and efficiency.

Prof. Satoshi Matsuoka  from the Tokyo Institute of Technology spoke about the impact this technology will have on their next-generation supercomputer, Tsubame 2.0,

"NVIDIA Tesla GPUs deliver large increases in performance across each node in a cluster, but in our production runs on TSUBAME 1 we have found that network communication becomes a bottleneck when using multiple GPUs," said Prof. Satoshi Matsuoka from Tokyo Institute of Technology. "Reducing the dependency on the CPU by using InfiniBand will deliver a major boost in performance in high performance GPU clusters, thanks to the work of NVIDIA and Mellanox, and will further enhance the architectural advances we will make in TSUBAME2.0."

JJ Abrams, the force behind TV and movie blockbusters “Alias,” “Lost,” “Mission: Impossible III,” and “Star Trek” -- to name just a few-- recently did an exclusive interview with NVIDIA. Abrams chatted with John Gaudiosi about gaming with his kids, the new Star Trek Blu-ray disc and DVD and the challenges of 3D filmmaking. Here’s a quick peek:

NVIDIA: What videogames impressed you when you were growing up?

JJ Abrams: Asteroids was my first love. I played Dark Castle and Castle Wolfenstein. The first Tomb Raider game when it came out, got me really excited about the soundtrack, the score, the mood of it. I thought it was really fun. Recently Portal was kind of amazing, and I was never a Rock Band or Guitar Hero person until The Beatles: Rock Band came out and now my three kids and I are sort of Beatles: Rock Band obsessed.

For the full interview, head over to NZone.com, NVIDIA’s online community for gamers.