Do you have any results on this?

My intuition is that it'd be better for diffusers to denoise perlin noise rather than guassian noise, because perlin noise more closely approximates 'the real world' than does gaussian nosie.