NVIDIA Presentations Archive from SIGGRAPH 2008

SIGGRAPH 2008 again saw NVIDIA presenting leading-edge development work on topics ranging across character animation, rendering, CUDA High-Performance Computing, and more.

Slides and other media from many of these presetations are available from the links below.

NVIDIA Sponsored Talks

Next-Generation Hardware Rendering of Displaced Subdivision Surfaces
Ignacio Castaño, NVIDIA Corporation

In this talk we provide an overview of the next-generation tessellation pipeline and its motivation. Our focus is on one of the primary applications: rendering of displaced subdivision surfaces, which dramatically increases the realism of animated characters. We also show how to adapt your production pipelines in order to create compelling content that takes advantage of this innovative rendering model.

Real-Time Rendering of Realistic Hair
Sarah Tariq, NVIDIA Corporation

Simulating and rendering realistic hair with tens of thousands of strands is something that until recently was prohibitively expensive for real-time use. In this session, we discuss how to render realistic hair with high geometric complexity in real-time on the GPU. Amongst other things we cover efficient creation and rendering of high amounts of geometry for hair (essential for creating realistic hair especially when in motion), shading, self-shadowing, level of detail, and important performance optimizations. We also talk about how to use next-generation hardware tessellation to make creating and rendering hair much more intuitive and efficient.

Adaptive Terrain Tessellation on the GPU
Iain Cantlay, NVIDIA Corporation

Next-generation techniques implement highly-programmable tessellation entirely on the GPU. We explain how tessellation can be applied to terrain rendering with displacement mapping. Our tessellation scheme is adaptive, with the polygon LOD varying as a function of terrain roughness and also with view-dependent silhouette detection.

Getting Physical:
  Solutions and Case Studies for Creating Scalable PhysX Content
Monier Maher, NVIDIA Corporation

Using physical simulation in applications takes their level of immersion to new heights. NVIDIA’s PhysX enables developers to add an unprecedented number of physical objects into scenes while maintaining high performance. This talk features the latest PhysX features and tools as well as real case studies that highlight common challenges and solutions.

A New Generation of Performance Analysis and Shader Authoring Tools
Jeffrey Kiel and Christopher Maughan, NVIDIA Corporation

This talk covers the latest releases of NVIDIA's popular PerfKit and FX Composer software products, as well as the brand-new NVIDIA Shader Debugger. First, learn how to extract maximum GPU performance using PerfHUD 6.0 (for real-time debugging and profiling - with tons of powerful new features), GLExpert (for OpenGL debugging), and PerfSDK (an API for accessing GPU performance counters).

Next, we show how FX Composer 2.5 and the Shader Debugger can make shader authoring, profiling, and debugging a breeze for programmers, artists, and technical directors. This session showcases new features such as a source-level shader debugging for Cg and HLSL10 shaders, Direct3D 10 support (including geometry shaders, stream out, and texture arrays), visual models & styles, particle systems, a revamped user interface, and much more.

CUDA: The Democratization of Parallel Computing
Paulius Micikevicius, NVIDIA Corporation

Massively parallel computing, once the domain of supercomputers, is now widely accessible in the form of millions of CUDA-enabled GPUs. These GPUs are fully programmable, support tens of thousands of concurrent threads, and have accelerated computations in a variety of disciplines by up to 2 orders of magnitude.

We provide an overview of the newest GPU architecture, the CUDA programming model, and latest development tools. CUDA enables efficient implementation of parallel algorithms by providing a small set of readily understood extensions to the C/C++ languages, eliminating the need to learn a new language. Development is facilitated by insightful profiling and debugging tools. Since the GPU is the only widely available commodity "manycore" chip, we explore it as a research platform for parallel programming and architecture.

Interactive Ray Tracing with CUDA
David Luebke and Steven Parker, NVIDIA Corporation

Ray tracing has long been associated with high-quality graphics, but it has not been suitable for interactive use. With CUDA and an NVIDIA GPU, it is now possible to ray trace reflections from curved surfaces, refractions, and accurate shadows. By combining these effects with rasterization to efficiently compute viewing ray intersections, accurate inter-reflections and other effects can be achieved at high resolutions and frame rates.

NVIDIA Presence in Regular SIGGRAPH Sessions

Course: Real Time Physics
Lecture & Course Chair: Matthias Müller-Fischer, NVIDIA Corporation

Physical simulations have become an important component of computer games. In next-generation games, players expect to see fully dynamic and destructible worlds, and this requires fast and stable simulation methods. In this class, lecturers who have made significant contributions in simulation methods present a wide spectrum of state-of-the-art methods for real-time simulation of rigid and deformable solids, and smoke and liquid simulation. In addition to the underlying physical equations, they present practical simulation methods and algorithms that will help physical-simulation developers and game developers apply these techniques properly.

To Trace or Not to Trace:
Image-Space Horizon-Based Ambient Occlusion
Louis Bavoil, Miguel Sainz, and Rouslan Dimitrov, NVIDIA Corporation

Ambient occlusion is a lighting model that approximates the amount of light reaching a point on a diffuse surface based on its directly visible occluders. It provides a soft shadow appearance which enhances depth perception and spatial relationship between objects. In this talk, we present a new algorithm for rendering ambient occlusion as a post-processing pass by sampling a depth buffer and its associated normal buffer. We discuss how to integrate this approach in real-time engines as well as provide performance analysis.

Let's Get Physical:
Real Time Hair Simulation and Rendering on the GPU
Sarah Tariq and Louis Bavoil, NVIDIA Corporation

Simulating and rendering realistic hair with tens of thousands of strands is something that until recently was not possible in real time. In this talk we present a method for simulating and rendering realistic hair in real time using the power and programmability of modern GPUs (Graphics Processing Units). Our method utilizes new features of graphics hardware (including Stream Output, Geometry Shader and Texture Buffers) that make it possible for all simulation and rendering to be processed on the GPU in an intuitive manner, with no need for CPU intervention or read back. In addition, we propose fast new algorithms for inter-hair collision, and collision detection and resolution of interpolated hair.

Want to Learn More? NVIDIA Documentation Home Page

NVIDIA® Corporation (Nasdaq: NVDA) is the worldwide leader in graphics processors and media communications devices. For more information about NVIDIA please visit our website: www.nvidia.com.

© 2008 NVIDIA Corporation