OSNews

This was actually one of the topics brought up in aseigo's TLLTS interview, which I highly recommend to people who want the 411 on KDE4.0 and the plans for the KDE4 series:

http://www.tllts.org/audio/tllts_226-01-02-08.ogg

Aaron's answer was disarmingly honest. Basically, if you hear claims of "KDE4 is 30-40% lighter/ faster than KDE3!!", take them with a pinch of salt: benchmarking is a surprisingly difficult and intricate business, and blanket claims like this (faster how? Faster in every conceivable way? What if KDE4 "does more" than KDE3 but takes slightly more time to do it - can it be called "faster" then?) are really quite naive.

Qt4's memory usage is a good example of this, which he brought up: most of Qt4 operations/ structures do indeed take up less memory than their Qt3 counterparts due to the extensive optimisation carried out by Trolltech - however, Qt4 also double-buffers all widgets, which takes up a fair amount of pixmap memory but also basically eliminates tearing, shearing and artifacting. Is Qt4 more or less memory hungry than Qt3 in this instance? In some ways yes, in some ways no, and in some ways it's apples-to-oranges, as Qt4 is doing a better job.

So don't believe all the performance claims that you hear - good or bad

Edit:

We can at least be assured that even with an unoptimised 4.0.0, performance and memory usage aren't completely diabolically awful, though

http://www.kdedevelopers.org/node/3137
http://www.jespersaur.com/drupal/node/36

Edited 2008-01-04 21:46