A feasibility study of ray tracing on mobile GPUs

Ray tracing is considered to be a promising technology for enhancing visual experience of future graphics applications. This work investigates the feasibility of ray tracing on mobile GPUs. A ray tracer was developed by integrating state-of-the-art construction and traversal algorithms and implemented in both CUDA and OpenCL. We then performed a detailed characterization of the ray tracing workload in terms of runtime, memory usage, and power consumption on both NVIDIA Tegra K1 and PowerVR SGX 544-MP3 GPUs. The results are compared against mobile CPU and desktop GPU implementations. It is proved that the Tegra K1 GPU already allows constructing the acceleration structure of 1M-triangle scene in around 120ms and performing traversal at a throughput of 15 to 70 million rays per second.