A high-speed ray tracing engine built on a field-programmable system

Ray tracing is a method of rendering high-quality images and video by calculating what happens to virtual light rays in a 3-dimensional scene. It is capable of creating far more realism than traditional Z-buffering methods. This paper describes the design of a hardware ray tracing system implemented on a multi-FPGA Xilinx Virtex-E prototyping system. The result is a hardware ray tracer that is capable of out-performing a 2.4GHz Pentium 4, running a well-known high performance software ray tracing algorithm, by up to a factor of thirty. When these results are projected forward into a next generation FPGA system, consisting of a single large Virtex 2 Pro FPGA, it is found that the system should be able to out perform the same Pentium 4 by up to two orders of magnitude, and the fastest known hardware implementation, the AR350, by up to a factor of three.