Is there a vast difference between CFG scale and de-noising for Diffusers Vs. Non-Diffusers Pipeline using Img 2 Img ?

[swamp-cabbage]

Using SD 1.5 models in the original recipe backend, a cfg scale and denoise of 7.5 produces massive variation in images generated. But using cfg scale and denoise of 7.5 when using SDXL and diffusers backend produces almost no variation at all in images generated.

Am I missing something ?

Advice ?

Thanks !