Fedora 23: KDE vs. Xfce vs. GNOME vs. LXDE vs. MATE

[shawnsterp]

Would you mind retesting with the "unredirect and disable compositing when there's a fullscreen application running" features enabled in Plasma 5? It takes two mouse clicks in a config window.

Would you mind tellng me where the "unredirect" is in kde. I have had the disable compositing for a while, but am not sure what the unredirect is (unless they are the same thing?).

[edoantonioco]

- cancels Kubuntu 15.10 download -

just disable compositing in fullscreen and you will be ok on kde

[humbug]

The reason it's not enabled by default in kwin is that it can cause flickering when you switch between the desktop and your fullscreen app (alt+tab).
This is why I leave it disabled and just switch off compositing if I actually need the best performance (alt+shift+f12) and reenable it when I'm finished playing (alt+shift+f12).

Is this flickering only an issue with KDE?

[sunweb]

The reason it's not enabled by default in kwin is that it can cause flickering when you switch between the desktop and your fullscreen app (alt+tab).
This is why I leave it disabled and just switch off compositing if I actually need the best performance (alt+shift+f12) and reenable it when I'm finished playing (alt+shift+f12).

Kubuntu 14.04, KDE 4, HD5770, no flickering.

Would you mind tellng me where the "unredirect" is in kde. I have had the disable compositing for a while, but am not sure what the unredirect is (unless they are the same thing?).

In KDE4 go to System Settings -> Desktop Effects -> Advanced check "Suspend desktop effects for fullscreen windows"
In KDE5 they moved it to System Settings -> Display -> Compositor. Though KDE5 issue buggy on many occasions for different sets of hardware.

[schmalzler]

I think it is very fair to run the software as is. That's the defaults any user are going to get, than you test what the users are going to experience...
What you are asking is not fair to the other ones. It's like you would have a performance test and your application runs faster because of aggressive caching(Chrome for example), you are satisfied with the result and then somebody has a memory test and sees that your application consumes a lot more memory and you ask them to disable caching and then compare...
You decided to have that tradeoff, than it is absolutely fair to everybody else to run the defaults on everybody and not tune for each case separately...

The test as it was done here was NOT fair. Because gnome-shell actually DOES unredirect fullscreen windows, at least some. The only difference is you can't find a button to disable it in gnome-shell. AFAIR I also read about unredirect feature in xfwm.Those things really should be evaluated BEFORE doing tests and the reader should be informed about the way the test is done. The best thing of course would be several tests for one DE to cover the different possible settings.

[mgraesslin]

Concerning "Suspend desktop effects for fullscreen windows" in Plasma: please note that this option does nothing for Intel hardware. Due to crashers in the past in the Intel driver we deactivated that option for Intel and won't ever turn it on again.

We consider this a useless option for Plasma. If one wants to run games: turn of desktop effects completely. Alt+Shift+F12 to do so, use a kwin rule to set it automatically for a window or use a "game script" (see http://kde-look.org/content/show.php...content=156659). Even games can indicate to KWin that they want desktop effects disabled by a simple property.

Given that we consider turning desktop effects off the correct way to play games on Linux (ideally you launch your own dedicated X Server), we don't care at all about benchmarks like that. We consider them actually a joke. They are done in a wrong way (hey let's test only one hardware and tell everyone that this is the ultimate way to do benchmarks!) and don't say anything useful anyway.

[carewolf]

Would you mind retesting with the "unredirect and disable compositing when there's a fullscreen application running" features enabled in Plasma 5? It takes two mouse clicks in a config window.

Isn't that default? At least it used to be.

[mgraesslin]

Isn't that default? At least it used to be.

No that has never been a default for 5.x and not for as long as I can remember in 4.x

[sorab]

I really wonder why sometimes your results are so no reproduceable for me. XFCE being slower than GNOME 3? Sorry, but I dont need any Phoronix Test Suite to prove that this cant be representative, because I can see with my bare eyes. On my slow machine (EeePC 1005) certain tasks, such as replaying an HD video, cant even be performed under GNOME (neither under KDE, Unity or Cinnamon) without major(!) framedrops; while XFCE is doing just foine with that (using the exact same video playback application on the exact same OS with the exact same drivers on the exact same hardware). OK, you tested a completely different hardware here, but can it make such a difference?
Also: In your benchmark has the suspend desktop effects on fullscreen opiton been enabled in KDE? (not saying that is has to be in order to be comparable, but it makes a huge difference. I 'd like to suggest always including both options for KDE in bechmarking for comparison)

[sorab]

Martin, you suggest to run a dedicated XServer for Games. Just curious: do you mean running a dedicated XServer alongside (i.e. at the same time) with the XServer KDE is running on, or to shutdown the desktop XServer first (in other words: log out, then log in again on a DE-less session)?