Quantcast

It’s the first day of the International Broadcasting Conference 2011 (IBC), in Amsterdam. Over 48,000 film, video and live broadcast professionals, and 1,300+ exhibitors from more than 140 countries have converged on the show to see the latest and greatest in video production technologies. NVIDIA is here with PNY at booth #7.J38 to demonstrate the power of the GPU and our just-announced GPUDirect for Video API.

Starting today, we are demonstrating new advances in interactive on-air graphics from Perceptive Pixel, using their 82-inch multi-touch display and interactive graphics. We’re also showing off a GPU-accelerated particle simulator for Autodesk Maya. And, video editors will want to check out our GPU-accelerated video editing demo showcasing Adobe Premiere Pro CS5.5 and Sony Vegas Pro.

Earlier today, we announced NVIDIA GPUDirect for Video, which reduces the latency in moving video to and from a GPU when using video I/O cards.  NVIDIA GPUDirect for Video is a free API that will allow 3rd party hardware companies to interface more directly with NVIDIA Quadro and Tesla professional GPUs to deliver video with the lowest possible latency. This technology is aimed at those who want to import video and export the video as fast as possible. The classic use case is broadcast graphics, but the medical imaging, manufacturing and defense industries are also ripe to take advantage of faster video processing.

How GPUDirect for Video works (click to enlarge)

Digital Video Systems GmbH (DVS) will be showcasing this new technology using their Atomix LT two-channel 3G hardware board and NVIDIA Quadro GPUs at their booth (#7.E25).

For those of you at the show, you’ll find more than two dozen NVIDIA partners on the show floor (download our PDF show guide here) demonstrating broadcast graphics, editing, compositing and effects, stereo production, and encoding and decoding solutions powered by NVIDIA Quadro professional graphics solutions based on the NVIDIA Fermi GPU architecture.

If you are at IBC, come visit us in Hall 7, stand J38.

  • http://profiles.google.com/rtfss1 rtfss none

    Sorry for saying that but AMD DirectGMA on firepro SDI cards seems better as is equal to CUDA P2P GPU-GPU functionality avoiding host copies.. yours is going through host mem altough goodly using no copies on host.. also sorry don’t know what magical feature is  NVIDIA GPUDirect for Video, it since CUDA 4.0 has cudahostregister for pinning host mem allocated without CUDA host mem alloc calls.. hope you improve soon to a use P2P PCIe transfers avoiding host transferes similar to directgma from AMD

  • Anonymous

    Thanks for reading our blog – we appreciate your feedback!

    Let me address some of your points:
    In terms of P2P, we actually do P2P transfers in our own NVIDIA Digital Video Pipeline product today, and after several years of experience with DVP, what we found is that we could achieve similar low latencies with the P2H2P approach that we have taken with GPUDirect for Video.

    Among other things, this gives us the benefit of being able to provide long term support for third-party video I/O cards without having our partners make changes as our GPU architectures change over time.

    GPUDirect for Video also allows transfers of data directly to the relevant GPU (whether Quadro or Tesla). Moving the data using CUDA C functions would invoke an extra interop stage to transfer into the graphics API, which of course increases latency. Similarly, GPUDirect has extensive support for synchronization which allows independent threads to work as efficiently as possible, further minimizing the latency of transfers. So yes we understand the merits of P2P but our architectural choices were made with all these factors in mind and so far are we are very happy with the results.

  • http://profiles.google.com/rtfss1 rtfss none

    Now makes sense thanks..