I usually use a displacement shader on large surface sometimes 1KMm x 1Km wide (working in real world units and in cm so 1KM=100 000 units) to mimic rocky, snowy or sandy surfaces. I always use the mental ray approximation node to achieve descent results. Now if I apply the displacement to a 1Km x 1Km wide plane (), I get ugly/weird results even with high mr_displace_aproxumation settings and the render time takes forever. On the other hand if I separate my 1km x 1Km patch into a number of smaller 10 meters x 10 meters patches, the displacement looks neat without artifact/weird results and the render time is way faster. Why is the first method giving bad results and why does separating the big patch into smaller ones give better results?
http://sor.typepad.com/.a/6a00d8350600cb53ef01a511ca0b5e970c-pi