Qt/GTK Speed On Unity/KDE/Xfce/LXDE/GNOME/Razor

[przemoli]

1. the way it is done here it is not pointless at all
2. your article is WAY too outdated. gnome shell with mutter is using unredirected for fullscreen already since a long time. in this article it still talks about a mutter without that.

That article only meantion Gnome Shell as NOT BEING TESTED.

[drago01]

1. the way it is done here it is not pointless at all
2. your article is WAY too outdated. gnome shell with mutter is using unredirected for fullscreen already since a long time. in this article it still talks about a mutter without that.

1. It isn't most of this benchmarks are not fullscreen and therefore the FPS they report might not be the FPS they actually draw on screen (please read the (non outdated) article).
2. I know (I wrote that code after all) but this is irrelevant see 1.

Also unredirected windows should have more or less the same performance everywhere assuming no background work taking up resources.

[ickle]

Was the KDE system configured properly?

Yes, that result did surprise me. Running the test locally points towards a pixmap leak in the GTK+ oxygen-gtk theme - performance is trashed when you are running from swap. Assuming that the KDE system is configured similarly, ofc.

[ickle]

Isn't QGears2 GLORIFIED "Do I have OpenGL AT ALL?" ???

Hmm. The QGears2 is not a very good benchmark in that the Image/Render/GL are not comparable and thanks to Qt do not measure what you think. The GL variant of the benchmark has no antialiasing, sacrificing visual quality for speed. That would be reasonable if the other two methods were also configured similarly. The Qt Render backend does all the rasterisation in the client on the CPU before sending images to X and compositing them. Qt does not have a fast raster engine, and this entirely misses the point of using the GPU to do the operations for which it is extremely fast. In essence, the Render backend is just a slow way of acting like its Image backend. If Qt were any good at all, the Render backend should be faster than the GL backend for all operations.

[varikonniemi]

Unity seems to be mature and performing well. It only took 2 years of public beta testing

Better late than never i guess..

[xeekei]

Way to go, Xfce!

[dee.]

Why no Cinnamon or MATE?

[leidola]

I'm wondering how the GTK+speed can vary that much across the desktops. I guess it comes all down to the speed of the default theme engine of the desktop and less to the speed of the compositor - if any.

[a user]

1. It isn't most of this benchmarks are not fullscreen and therefore the FPS they report might not be the FPS they actually draw on screen (please read the (non outdated) article).

aaaaand???

the benchmarking was not done on my pc so all its results are useless.... no!

that's exactly what was benched: some specific circumstances. not being the scenario you desired to bench does not make it less usefull. you seem to expect a simple answer from a benchmark: what is the best xy?

sounds like the typical kids question, who has the bigger one?

the test provides some information about the influece of compositors in this test setup. it is neither ment as an all overall test, nor testing the impact of compositors during fullscreen etc.

i read the artical you mentioned. while makes some good technical points it simply doesn't fit to this test.

2. I know (I wrote that code after all) but this is irrelevant see 1.

actually your article is irrelevant!

when you bechmark a specifc scenario you get results that have to be interpred in context of it. doing a benchmark in another scenario than you wished doesn't make it useless or pointless.

Also unredirected windows should have more or less the same performance everywhere assuming no background work taking up resources.

also this is not true and enver was for none of the DE we have on linux. and btw. it is quite obvious why, and should be for the guy who wrote that code.

[a user]

Unity seems to be mature and performing well. It only took 2 years of public beta testing

Better late than never i guess..

way faster than kde needed.