Fastest Perlin-Like 3D noise algorithm?

It's been well over 20 years since Ken Perlin first invented his noise. Has anybody managed to make a faster kind of 3D noise generator with properties close to Perlin's (procedural, natural-looking grouping, reduced banding, regular feature size, etc)?

I'm trying to build a procedural world generator but currently Perlin just isn't cutting it. I admit my implementation isn't the best it could be right now, but if I'm about to rewrite 开发者_Go百科it anyway I wondered if there was a better algorithm available.

You want Simplex Noise.

• less computationally expensive
• not based on a square grid, so no obvious directional artifacts
• scales better to higher dimensions: O(N^2) vs Classic Perlin's O(2^N) for N dimensions

There's a good explanation here. Apparently Ken Perlin's example implementation is not the most easy to understand code.