What does that mean?
IMO, if there is not a measurable improvement, then it is not worth it. Are you using actual empirical data? That is, are you taking before and after benchmark measurements? Or are you just going by what you [think you] see with your eyes?
I am reminds of some folks who disable their page files, then report, "nothing broke" so they leave the page file disabled, thinking that is better. Or as suggested above, it is just the placebo affect where we want it work great so we believe it is working great. Yet if one to do some real scientific measuring, or "double-blind" A/B comparisons, there would be no difference - or things might even have degraded a little bit.