Tops Systems Corp of Japan, a venture involved in multicore technology, together with Toyota Motor Corp and Nihon Unisys Ltd, both of Japan, is developing a dedicated integrated circuit (IC) for ray tracing*, an image rendering method used in 3D computer graphics (3D CG) processing. A total of 73 heterogeneous cores designed specifically for ray tracing operations will be single-chipped, and nine of these chips interconnected (see Fig). With high-definition (HD) resolution at 1920 x 1080 pixels, the target processing speed is 800 tera floating point operations per second (TFLOPS).
* Ray tracing: A rendering method that traces light rays in reverse, from the point of view toward the pixels.
Researchers have figured out how to resolve the basic issues involved in the overall system architecture and application parallelism analysis, and will begin detail design and implementation as an application-specific IC (ASIC) shortly. They plan to fabricate the chips using 45nm manufacturing technology, and expect the chips to operate at 750MHz, integrating 130 million gates into a 17mm square footprint.
On Bezier SurfacesThe chip is intended for use in automobile design. Toyota Motor developed its own 3D CG rendering system in the early 1990s and used it in-house with the goal of evaluating design elements such as vehicle shape and coloration of painted surfaces through simulations to reduce the number of physical prototypes required. The new IC is being considered as the latest and greatest version of this system.
Most 3D CG realtime rendering used in games and other applications use a graphics processing unit (GPU) as an accelerator. Researchers this time, however, instead chose heterogeneous dedicated ICs because they can use direct rendering (ray tracing) on freely-curved surfaces (Bezier surfaces), without the polygons used in most 3D CG rendering systems. Compared to polygons, it is possible to generate much more precise imagery (Note 1).
Note 1: Most 3D CG CAD systems handle data for freely-shaped curve surfaces, but they generate images by approximation as polygons (tessellation).
GPUs have been evolving into general-purpose devices, but based on the assumption that polygon rendering is used.
The rendering method used is rasterizing*, used in most games, and not the ray tracing method that precisely simulates light reflection and other characteristics. As a result, researchers decided that freely-shaped surfaces were not a good choice for direct ray tracing. According to Yukoh Mtsumoto, president and CEO of Tops Systems, which was responsible for the design of the new chip, "There are reports of realtime ray tracing in the literature, but most are polygon-based."
* Rasterizing: A rendering process used in 3D CG where pixel values are calculated for each polygon individually.
The Toyota Motor rendering system, unlike conventional ray tracing systems, splits visible light into 35 wavebands, and calculates optical properties such as reflection and refraction for objects individually for each waveband. To execute this intense processing at HD resolution and in realtime demanded supercomputer-class performance: 800TFLOPS (Note 2).
Note 2: The performance figures cited for the newly developed system are for a specific application only, and therefore cannot be compared to general supercomputer performance benchmarks such as LINPACK.
Researchers hope to master this enormous number-crunching load by converting portions to fixed-point operations, improving the software, etc.
Ray tracing is highly parallel processing, making it possible to divide the image area into multiple regions, and assign multiple cores to each. Components with complex structures, such as automobile headlights, vastly increase the volume of calculation required, so dynamic load distribution is used to help equalize processing loads between cores.

Nikkei Electronics Asia magazine is available each month free of charge to engineers, managers and other qualified readers.