I did a little *subjective* battery testing on my laptop (Core Solo 1.06Ghz ULV, Intel 945GM, 2GB RAM, SATA HDD), pitting KDE 4.4.5 Kubuntu 10.04 vs. latest (stable) gnome version in Ubuntu 10.04. I noticed essentially no difference between the two. Compositing enabled in both, doing basic things like reading/scrolling/opening documents (mostly pdfs), text editing, and web browsing mostly. I imagine the important thing is that I try to use programs intelligently, which means almost never scrolling (I use page downs instead) and disabling/quitting programs that cause a lot of wake ups according to powertop. Maybe if I ran my laptop on the battery as I would from a power source things would be different, but do (sensible) people do that?
I find this quite satisfying actually because I like KDE quite a bit more than gnome (especially because of Okular, Kile, Kopete, Dolphin, Gwenview and plasma-desktop vs their gnome counterparts), and while KDE feels sometimes a bit more sluggish than gnome (I think mainly for opening more feature-filled apps), it's not a problem if you already have them open when you move to battery power.
It would be nice to see this test redone from a more realistic usage perspective. E.g., it would be nice to see the computers set to do some light/moderate web browsing with some text editing/word processing, file navigation, etc. in between.
The tests done here are so far away form real world usage that one has to wonder how useful they are.
These tests need to be redone today: power and memory usage of gnome, kde, mate, xfce... and no ubuntu this time please!