03-12-2010, 09:12 AM
Basically it is possible that KDE 4 effects (or Compiz) can show memory leaks in in grafic drivers, in that case the memory usage for Xorg will increase dramatically. Mostly older Nv drivers seemed to be affected by that so running KDE 4 on old machines with low memory was much slower than with KDE 3.5. I tested my own live systems now with a very simple test: show memory usage directly after boot using the integrated infobash tool (inside vbox with 980 mb ram) - rounded, basically it is changeing a bit with each run, but you see the tendency.
32 bit: 129 mb
64 bit: 212 mb
kde 4.3 (backport for lenny):
32 bit: 179 mb
64 bit: 276 mb
03-12-2010, 09:24 AM
03-12-2010, 09:25 AM
PS: Or... "What the fsck are we doing here?"
03-12-2010, 09:33 AM
03-12-2010, 03:10 PM
As far as memory use goes it looks like the Gnome guys are creamin the KDE guys. At least when it comes to the buntus anyways.
Initial blimpage in memory for Kubuntu 10.04 might be due to the Nepomuk Strigi indexer auto starting on login every session.
03-12-2010, 04:50 PM
Awesome, mugginz. I think you earned (almost) everybody's cookies.
The Kubuntu figure is in agreement with what Kano reported for 64 bits. Which makes sense if you are also on 64 bits...which you are, I think?
So your tests give 19% of Ubuntu-KDE over Ubuntu-Gnome, and 33-64% (!) of Kubuntu over Ubuntu-Gnome, if I understood your numbers right. Phoronix results, including caches, give an average of 30% for Ubuntu-KDE over Ubuntu-Gnome. Now, the reason why the numbers are reduced is because KDE 4 caches more than Gnome. If the amount of cached memory was the same, the differences would be of course greater, according to the supreme laws of division. And the same would happen if you subtracted the memory consumed by all the other processes in the system, which are common to both DEs--adding a constant changes the percentages, only that in favour of the loser, reducing the differences ; )
Very lucid, V!NCENT. Does that mean that you now realize why these past comments of yours were so out of place that I didn't bother replying to them?
Originally Posted by V!NCENT
Originally Posted by V!NCENT
03-12-2010, 05:26 PM
I should've noted that the above figures are as follows:
All distros are x64
Above tests are with all current updates applied
Configuration, software, etc are as per default install.
Memory usage is as per that reported by the free command.
Buffers and cache are not included and don't belong in that measurement in my view.
Overhead of the script performing the bench is approx 1.5MB for bash, etc.
The Ubuntu 10.04 test is the standard Gnome desktop on Ubuntu 10.04
The Kubuntu 9.04 test is the standard KDE desktop on Kubuntu 9.04
The Kubuntu 10.04 test is the standard KDE desktop on Kubuntu 10.04
My objective is to try and determine how much RAM is made unavailable by the standard software suite of a particular distro.
For the Gnome test the terminal used is gnome-terminal and the Video playback is performed by Totem.
For the KDE test the terminal used is Konsole and the Video playback is performed by Dragon Player.
The video file is a 320x240 ogg theora video.
The desktop is running at 1024x768 24bits/pixel
Tests werre run in a KVM virtual machine as per:
- 1024M RAM
- 8096M HDD
- Cirrus CL 5446 VGA
- ENSONIQ AudioPCI ES1370 sound card
A virtual machine was used because the only fast machine I have here at the moment is my desktop and I need that up at all times. Now that I'm happy with the test methodology I'll be running them on a Celeron 1.2G, 512M RAM, 40G HHD machine in order to get an idea for 32bit results. I've got some figures for unpatched Kubuntu 10.04 x64 but the KDE fans out there wouldn't want to read those.
03-12-2010, 05:57 PM
Is Compiz and/or the KWin compositor enabled? This can affect memory usage quite a bit (around 30MB or so for 1024x768).
03-12-2010, 06:01 PM
Not for those tests but I'll soon have numbers including with & without compiz when I finish the benches on bare metal.
Originally Posted by BlackStar
03-13-2010, 04:36 AM
It's interesting to note that when you start an application, it takes a larger chuck of memory in GNOME than KDE. Not enough to make up the difference, but I wonder if you could come up with a heavier use scenario that had a few apps running at the same time. Maybe try a console app + file manager + video player and see what happens.