View Full Version : DNeg Accelerating Visual Effects with NVIDIA Quadro and CUDA

01-20-11, 07:40 AM
Double Negative (http://www.dneg.com/) (DNeg) is the largest visual effects facility in London, England. Their award-winning visual work can be seen on films ranging from ‚??Inception‚?? to ‚??Sherlock Holmes‚?? to ‚??2012.‚?? Recently, Double Negative used an NVIDIA GPU-based render farm, to accelerate components of their visual effects pipeline resulting in speed increases of up to 20X.

Central to DNeg‚??s VFX workflow is their proprietary fluid simulation system, known as ‚??Squirt.‚?? ¬*With Squirt, video professionals can simulate any dust, smoke, fire, liquid and volumetric-based effects. ¬*With the help of a 6 month R&D effort, DNeg optimized Squirt to leverage NVIDIA CUDA (http://www.nvidia.com/object/what_is_cuda_new.html), parallel computing architecture.¬* The change enabled dramatic increases in computing performance, and harnessed the power of hundreds of processor cores inside the NVIDIA Quadro (http://www.nvidia.com/page/workstation.html) and NVIDIA Tesla (http://www.nvidia.com/object/tesla_computing_solutions.html) GPUs.

‚??Moving our fluid solver onto the GPU allows our artists to get the results of their simulations back much faster, without any impact to their workflow,‚?? explained Dan Bailey, lead GPU developer, Double Negative. ‚??By default, fluid simulations are now sent to a specialized GPU farm, affording the artists more time to iterate and ramp up the complexity of a shot to achieve a more believable result for the big screen.‚??

With a key component of their fluid simulation system now ported to the GPU, DNeg is seeing a 20X performance increase versus the CPU, and is upgrading to the new Quadro 4000 to further boost performance with it and other apps.‚??

‚??When we embarked on developing in CUDA, there were a lot of resources available to make the task much easier,‚?? continued Bailey. ‚??The NVIDIA CUDA forums (http://www.nvidia.com/object/cuda_home_new.html) are very useful, and we also benefited greatly from information available from NVIDIA‚??s GTC (http://www.nvidia.com/object/gpu_technology_conference.html)presentations (http://www.nvidia.com/object/gpu_technology_conference.html) last fall. CUDA is great to work with. In the future, we‚??re looking at driving as much of our computation as we can onto the GPU, starting with image manipulation tools, deformers, and of course, fluid simulation.‚??

DNeg is continuing to advance their adoption of GPU-optimized workflows and will be using their GPU-accelerated on upcoming productions including ‚??John Carter of Mars,‚?? ‚??Captain America: The First Avenger,‚?? ‚??Paul,‚?? and ‚??Attack the Block.‚??

http://feeds.feedburner.com/~ff/nvidiablog?d=yIl2AUoC8zA (http://feeds.feedburner.com/~ff/nvidiablog?a=CvwZ4kB0Sgo:SWNED24nRiQ:yIl2AUoC8zA) http://feeds.feedburner.com/~ff/nvidiablog?i=CvwZ4kB0Sgo:SWNED24nRiQ:V_sGLiPBpWU (http://feeds.feedburner.com/~ff/nvidiablog?a=CvwZ4kB0Sgo:SWNED24nRiQ:V_sGLiPBpWU)

More... (http://feedproxy.google.com/~r/nvidiablog/~3/CvwZ4kB0Sgo/)