Unlike a renderer with a prescribed look, or a language limited to rendering, the OptiX engine is eneral, enabling software developers to quickly accelerate whatever ray tracing task they wish and run it on standard hardware. Flexibility within OptiX begins by abstracting the execution to single rays, simplifying calculations to one ray at a time. The data each ray carries and gathers is fully customizable, extending ray calculations beyond basic image creation. The data fed to OptiX is also programmable, enabling programmable shading for new techniques, programmable intersection for procedurally accurate surface types, and programmable cameras for new composition potential.
The OptiX engine also provides applications with critical features such as: parallelism (both within the GPU and across GPUs); acceleration structures (BVH and KD trees); and traversal algorithms. Exploiting the GPU is also managed through built-in load balancing, the scheduling of shading & tracing, and recursion methods. A tight coupling with graphics APIs like OpenGL is also provided, allowing raster and ray tracing approaches to be combined for additional flexibility and performance possibilities.
NVIDIA's current "GT200" GPU architecture enabled OptiX to double its speed as compared to its previous "G80" architecture, and preliminary tests show that NVIDIA's next generation "Fermi" GPU architecture will deliver substantial OptiX performances gains yet again, according to Nvidia. Also, while today?s release requires NVIDIA?s professional solutions of Quadro FX and NVIDIA Tesla, the OptiX engine will soon expand its support to include NVIDIA GeForce GPUs with Fermi, as forthcoming performance will make ray tracing possible in consumer applications.
The OptiX SDK is free of charge and the engine is free to license in applications.
OptiX SDK Download
* Operating System: 32 or 64-bit versions of Windows XP, Windows Vista, Windows 7, Linux
* CPU: x86 compatible
* System Memory: matches graphics board recommendations
* GPU: NVIDIA Quadro FX or NVIDIA Tesla (GT200 class required for multi-GPU scaling and technical support)
* Frame buffer memory: varies with data complexity
* Driver: NVIDIA Unified Driver r190 or later, CUDA toolkit 2.3 or later
* C/C++ Compiler: Visual Studio 2005 or 2008, along with CMAKE