by Anton Shilov
09/03/2010 | 01:05 PM
In spite of the fact that there are a lot of technologies that can potentially overcome the issues of traditional rasterization-based rendering, such as ray tracing, rasterization will live for at least another decade since other techniques of image construction have a lot of peculiarities that cannot be efficiently solved using today's and tomorrow's hardware.
Ray tracing is a very mathematically complex method of image construction that can greatly improve quality of complex visual effects, including ambient occlusion, fresnel reflection, depth of field, area light sources as well as optical effects, including, but not limited to scattering, reflection and refraction. In fact, ray tracing algorithm is much closer the behavior of light in nature than traditional rasterization technique that has been used for years in real-time 3D graphics, primarily video games.
Since ray-tracing greatly solves complex rendering problems, it makes sense to implement certain fixed-function hardware for ray tracing into the future graphics processing units (GPUs). In fact, a lot of ray-tracing engines based on general purpose computing on GPUs emerged in the recent years, but all of them have a number of limitations, which can be, at least partly, overcome by implementing special-function ray-tracing hardware into graphics chips.
"Our group has definitely done research in [fixed-function ray tracing hardware] area to explore the 'speed of light', but my sense at this time is that we would rather spend those transistors on improvements that benefit other irregular algorithms as well," said David Luebke, director of graphics research at Nvidia, during a public conversation earlier this week.
Graphics processors need to greatly improve programmability and computing performance in order to make techniques like real-time ray tracing feasible.
"Looking forward, increased GPU performance and programmability both combine to make simpler approaches - like ray tracing - practical. On the flip side, graphics is certainly not solved and there are many effects that we cannot do at all in real-time today, so you will continue to see games push forward on graphics innovation, new algorithms, new ways to use the hardware," said Mr. Luebke.
Nvidia is working hard to ensure improvements of GPGPU performance of its chips and the company claims that steps like Fermi's large unified L2 cache and unified memory space are examples of generic hardware improvements that benefit GPU ray tracing as well as many other workloads. Unfortunately, even in the next ten years ray tracing will hardly become feasible for high-end games and applications. As a result, rasterization will live on.
"Forward rasterization is a very energy-efficient way to solve single-center-of-projection problems (like pinhole cameras and point light shadow maps) which continue to be important problems and sub-problems in rendering. So I think these will stick around for at least another ten years," stressed the director of graphics research at Nvidia.