Phoronix has used its Test Suite to compare the memory and power consumption of different desktop environments. However, the results should be handled with care.

The "Power & Memory Usage" test was done to evaluate whether XFCE and LXDE consume less power and memory compared with their "big" siblings KDE and Gnome. The tests were done on a stock Ubuntu installation. At first glance the results suggest that KDE consumes much more power than the others. However, these results are misleading.

For a start, measuring memory consumption is all but an easy task and requires a lot of thought. The problem is that many applications share a certain amount of memory and this is especially true for KDE. Very few programs out there can handle this shared memory properly while still showing the memory consumption of a process. "Top" and "free" are worthless in this regard; if you need to use any tool, take "exmap"! A more detailed analysis of this problem was done here.

Yet even if the memory consumption could be measured perfectly, there is still the matter of what this memory is being used for. Gnome and KDE do need quite a chunk of memory right at the start, but that is mainly due to the larger libraries, which offer a number of functions for other programs as well. It is for this reason that the memory measurement done with a set of programs shows different results, indicating that a Desktop with larger libraries has a much slower growing fingerprint for each additional application compared to a Desktop with smaller libraries.

However, even after taking this point into account, you might still ask, "what do you get?" KDE and Gnome usually offer a file indexer and tagger by default - whether these are running or sleeping can massively influence the measurements in tests such as these. Also, both desktops come with some heavy 3D effects. If you turn these off by default the memory footprint and the power consumption is of course much lower.

A comparison between a full featured desktop vs a Desktop with much less functionality is brainless - you could also add a turned off machine to the test and declare it the winner because it has the smallest memory footprint and power consumption!

To clarify: in general the Phoronix Test Suite can be used to evaluate certain data or at least trends in certain data - but data acquisition requires fixed surrounding conditions as well as a detailed discussion of the compared objects and their features, not to mention a detailed (read: scientific) analysis of the results. Throwing numbers around is not enough - if you are going to do so, you should at least include accurate details!

The Power and Memory Usage test is lacking in all these factors, so it is unfortunately worthless. :/