Kde On The Pandora

[Tor]

I use many KDE applications, but I don't use the KDE window manager. That one I don't like. I can't seem to get the configuration I want, and even on the fastest hardware it's way WAY too slow to start. So I use the infinitely configurable fvwm window manager. It starts in 1 second. You don't need the KDE desktop environment to use KDE applications.

And lastly, KDE 3.5 has been very much usable on the good old Nokia N8x0 tablets for what must be a couple of years now. And there was even some 4.0 testing. The Pandora is much faster and has more memory resources. If you want KDE you shouldn't have any problems methinks.

 

[bzar]

The last time I gave KDE a shot was when 4.2 was released. I'd really like to use it as my main DE for obvious reasons, but it always seems to have something I can not live with or lack something I can not live without. There might be workarounds, though. My main concern with 4.2 was performance. Everything seemed to use more resources than on gnome + compiz fusion. And I mean everything. Some applications jumped from previous 10% CPU usage to 30% (like some emulators). The desktop was also in some ways less configurable in terms of input. I couldn't find any way to map my mouse buttons to wm functions like I can in compizconfig. This is a minor nuisance, but I'm really used to having expose/scale on a specific button on my mouse. I guess I could just swap kwin for compiz, but somehow I don't think it would fix the resource hogging.

 

[dflemstr]

My main concern with 4.2 was performance. Everything seemed to use more resources than on gnome + compiz fusion. And I mean everything. Some applications jumped from previous 10% CPU usage to 30% (like some emulators).

Are you really sure about that? Because:
- If the emulator uses Qt, it should run faster on KDE than on other DE's for obvious reasons (one of which is that the Qt libs are loaded in the DE already so the emulator doesn't have to load them)
- If the emulator uses GTK+ and glib, that shouldn't be a problem because Qt can run using the glib main loop too, so your application will actually run under exactly the same circumstances on KDE as in Gnome.
- KDE can't interfere with any other part of the program, really. The GUI would be the only bottleneck, but I seriously doubt that it indeed is one.

Open the system monitor
(Open KRunner with Alt+F2 and press on the "Show System Activity" button, or search for System Monitor: ) and see which of your applications takes the most resources. Nepomuk can take a lot of CPU while it is indexing files, but that should be understandable...

 

[bzar]

My main concern with 4.2 was performance. Everything seemed to use more resources than on gnome + compiz fusion. And I mean everything. Some applications jumped from previous 10% CPU usage to 30% (like some emulators).

Are you really sure about that? Because:
- If the emulator uses Qt, it should run faster on KDE than on other DE's for obvious reasons (one of which is that the Qt libs are loaded in the DE already so the emulator doesn't have to load them)
- If the emulator uses GTK+ and glib, that shouldn't be a problem because Qt can run using the glib main loop too, so your application will actually run under exactly the same circumstances on KDE as in Gnome.
- KDE can't interfere with any other part of the program, really. The GUI would be the only bottleneck, but I seriously doubt that it indeed is one.

Open the system monitor
(Open KRunner with Alt+F2 and press on the "Show System Activity" button, or search for System Monitor: (image)) and see which of your applications takes the most resources. Nepomuk can take a lot of CPU while it is indexing files, but that should be understandable...
(image)

Ok, someone has done something very right to kde since I last tried it. I repeated the test a couple of times to make sure. I used two applications to test the performance. First, I used visualboy advance (with super 2xSAI and frameskip 0), which is one of the emulators for which I last saw big differences in cpu usage and the other was one of my own pygame creations called torqueball (a project I may never finish ). I measured the peak cpu% for both applications for an autoplaying scenario. For gnome/compiz I got: vba 77%/1% and torqueball 9%/3%, and for kde/kwin: vba 47%/7% and torqueball 9%/10%. In other quick tests (like moving a window around with similar effect settings) kwin seemed to almost always use more cpu. On the other hand, vba results may indicate that this cpu usage is well spent.

I may have to give kde another go sometime soon.