My time:
CPU: Intel i7 3.5Ghz
GPU: Nvidia GTX 970
OS: Windows 10 64bit
Time: 1:17.52 (GPU - CUDA)
Time: 5:14.17 (CPU)
I seemed to get the best rendering time when I set the Tiles to quarter size of the finished render (so 480 x 270). I tried rendering the whole thing in one shot but it was about 10 seconds slower. Going by the default 128x128 netted me a time of around 1:40.
Very interesting seeing what differences the Tile size makes. I wonder what the mathematical explanation is behind the differences, and if there’s an easy-to-define sweet spot.
