Wednesday, May 20, 2009

Comparing KDE 3.5.10 and KDE 4.2.2 memory usage

Recently I have been checking out the memory usage of various window managers and desktop environments, concluding with a study of KDE. That led me to wonder how well KDE would work on my oldest working system, the Dell Dimension 4100 desktop. I often used KDE 3.5 on that system, and I had a KDE 3.5.9 and 3.5.10 system already installed, so last night I updated my PCLinuxOS partition on the 4100 to bring it up to date, used it for a while, then upgraded my sidux partition on the same system, bringing it up to KDE 4.2.2.

I found that both versions of KDE did swap, to a certain extent, on both PCLinuxOS (KDE 3.5.10) and sidux (KDE 4.2.2), but the memory management was excellent on each system, not too aggressive, nor too inactive, allowing swap where needed, keeping at least a modest amount of free memory at all times, and trimming swap usage when memory requirements were relieved. The result was acceptable performance, even on a system that is nearly a decade old, and contains a Pentium 3 processor with about a 1 GHz CPU and 256 MB of memory.

KDE remains a viable desktop environment in both the older legacy form and the newer Plasma based technology. Though a lightweight window manager definitely reduces the amount of swap space used, KDE still gets the job done, and surprisingly well with both systems. There was always at least 4 KB of available memory at all times, and when that level was reached, the scheduler reallocated memory and grew the active swap space, and when the memory usage declined, the swap space was cleanly released. That speaks very well for the Linux kernels on both PCLinuxOS and sidux systems, and it also demonstrates that KDE remains a viable desktop alternative, even for hardware that is nearly a decade old.


Aekold said...

I am interested how did you compared it? Memory usage of KDE itself is very low, most of memory is eaten by background processes like beagle, samba, power and network management tools, and compiler options. Posts like this is calculation of spherical horse in vacuum, but not real comparison. I'd recommend you to install some minimal distro on VirtualBox, turn off all possible backgruond tasks and uninstall all possible daemons (including pulseaudio) and then to low RAM of virtual machine until KDE will not start at all. I can tell you that my record minimum was for KDE 3.5.3 and it was 45MB.

zbog said...

Umm, could you also run some tests regarding memory usage. Although this is useful, I expected more of a benchmark comparison (Phoronix-like).

Anyway, have fun with the new KDE!