NVIDIA's Presentation

NVIDIA’s Presentation

Just like at last year’s SIGGRAPH, NVIDIA Vice President and General Manager of Professional Visualization (the Quadro business unit), Bob Pette, was the main presenter supported by Greg Estes, Vice President of NVIDIA Developer Programs, and Sandeep Gupte, Director of Professional Visualization at NVIDIA.

This year the theme of the presentation was NVIDIA Brings AI to Graphics. Greg Estes pointed out that the AI revolution is the defining technology of this generation and NVIDIA is also bringing AI to graphics for digital content creators.NVIDIA’s presenter pointed out that everything starts with research.  And NVIDIA has been partnering with multiple universities and also with game developers to use AI in graphics.

Using Remedy’s vast store of animation data, NVIDIA GPUs, and deep learning, NVIDIA researchers trained a neural network to produce facial animations directly from actor videos.

Instead of using artists to perform labor-intensive data conversion and touch-up for hours of actor videos, NVIDIA’s solution requires only a few minutes of training data. The trained network then automatically generates all facial animation with the same or better quality than current methods for an entire game from a simple video stream.   AI can also be used for temporal anti-aliasing to eliminate that shimmering and crawling in game when the camera is in motion.  And AI can also be used for predicting which light rays are really important in a scene.

New research is ongoing in AI in the areas of facial animation, denoising, anti-aliasing and light transport.  Denoising can now be done almost instantly using AI which allows content creators to have more control over their work and they are able to see their results far more quickly than previously.
NVIDIA introduced Optix 5.0 SDK at SIGGRAPH which now uses AI-enhanced Ray Tracing.  The new ray tracing capabilities will speed up the process required for visualization. It features new AI denoising capability to accelerate the removal of graininess from images, and brings GPU-accelerated motion blur for realistic animation effects.Here’s a few seconds of video that we shot of AI denoising in action:

Running OptiX 5.0 on NVIDIA’s DGX Station AI workstation will give content-creatiors the rendering capability of 150 standard CPU-based servers. To achieve equivalent rendering performance of a DGX Station, NVIDIA claims that content creators would need access to a render farm with more than 150 servers that require some 200 kilowatts of power costing $4 million dollars over 3 years, compared with 1.5 kilowatts for a $75,000 DGX Station.
We got very close to the DGX station which was running 4 water-cooled Tesla GV100 cards, but we could not hear it running in a moderately noisy room.

NVIDIA also released a new performance driver for TITAN X to make it faster than ever before with applications such as Autodesk, Maya, and Adobe Premier Pro.NVIDIA also pointed out that most designers like a thin and light notebook but generally need much more power than CPU graphics can generally provide.

Professionals with underpowered graphics notebooks can now harness the power of discrete NVIDIA TITAN X or NVIDIA Quadro graphics cards through an external GPU (eGPU) chassis and significantly boost the performance of their applications.  Beginning in September, Quadro graphics will be available through NVIDIA qualified eGPU partners for those who use high-end content-creation applications as well as for CAD and simulation apps.

NVIDIA is continuing to support the VR ecosystem.  We found NVIDIA Research in the Emerging Tech area demonstrating their work in Optic and Haptics for AR and VR\ and they also announced the ZCAM integration of VR Works 360 Video SDK.

HP Mars is a NVIDIA project with Technicolor, Autodesk, and HP to bring designers, inventors, space enthusiasts and experts together to examine the possibilities for colonization of the red planet.

NVIDIA’s VRWorks SDK uses research to bring more immersion to VR.

We have seen how VR works with fix foviated in VR to focus the most graphics power on where the eye actually looks, and NVIDIA has already been able to bring Lens Match Shading and Multi-Resolution Shading to allow for higher details using the same amount of graphics power. New simulators for AI in VR via an Isaac-trained robot in Holodeck.  Here’s a video of Isaac:

The following slide in NVIDIA’s presentation is what they want us to take away from their hour-long presentation – that NVIDIA is the world leader in computer graphics and they already provide a complete solution for gamers, developers, scientists, and content creators.  We saw NVIDIA introduce new cameras, new SDKs, and new support at last year’s SIGGRAPH, and they are continuing to build on their strengths to deliver what they believe their customers want.

Of course, AMD has different strengths than NVIDIA as they can leverage both CPU and GPU, and they are just beginning to break into AI and deep learning.  Let’s take a look at their presentation which was held last Sunday evening attended by an enthusiastic crowd  of 2,300 at the nearby club, The NOVO.