I have float volume data. I dont have any further knowledge about them, so I am currious it it is good to render them as if it was RGB(A) image (and also use compressions designed for RGBA). If I compress it with some RGBA compressors and look at result on screen, it looks similar to raw data, but float representation as it was, is messed up and can never be reconstruct.

This is more like an academic question, but I am currious in general "work ewerywhere" approach where data are unknown.