We implemented a fast Euclidean Distance Transform method for the VdW, solvent-accessible and solvent-excluded surfaces. The result meant that molecular surfaces were kinda blocky - even if slightly smoothed:
I've been playing with some updates because the render benchmarks showed that a lot of the time was spent in creating the mesh and smoothing the mesh rather than the initial distance calculations.
It's particularly nice in translucency:
I'm currently working (with Claude reviewing my code) on updating the solvent-excluded surface.
Discuss this on our forum.