This winter, I ported my code to the GPU (using DirectX 11), the main advantage of which is SPEED. On my old Core 2 Duo (2.4 GHz), it took about a day just to render a 3K x 3K image. But now, running on just a $140 Radeon GPU (the one that came with my PC), I can usually render an 8K x 8K image in 1-2 hours! How much faster is that? Well, running on the CPU was: 24 hours / 9 megapixels = 2.667 hours per megapixel But running on the GPU is: 2 hours / 64 megapixels = 0.03125 hours per megapixel So, the GPU is (ballpark) about 85 times faster - not too bad! I've also added upgraded the ambient occlusion to be directional, which produces more realistic shadow approximations. Hope you like the images! Click on any image for a larger (but still significantly scaled-down) version. |