Gtk Engines Benchmarks – What’s the fastest?
Final Chart (not sorted for a better results view):
Fastest: Bluecurve —-> 6.11 secs
Slowest: Aurora ——> 18.95 secs
If you had other great engines… just comment it and i’ll update benchs. :)
Gianvito's
Stay hungry, stay foolish.
What about the Xfce-gtk-engine?
Ok… I will try as soon as possible to add xfce gtk engines…
Updated with xfce engines
Nodoka is almost murrine engine with a different name (Nodoka), its performance are nearly the same.
If you chose a Murrine theme with rounded borders, you’ll see it.
By the way, the problem is not on +1second, -1second etc etc… the thing is that Aurora, for example, is exactly twice slower than murrine with small roudness :)
This will be noticeable when using many windows.
Talking about nodoka… i added it to the test because it is part of Fedora… I wanted to give a general purpose to the test…
However… We all know Murrine is a great engine… It combines beauty and speed together :)…
Aurora is handsome to see… but it has to do more if it wants to be a very usable engine… especially on old machines
What about rezlooks gtl-engine???
@pibarnas
Updated with rezlooks engine
Here’s some more. http://art.gnome.org/themes/gtk_engines/
You got most of them, but I think you’re missing eXperience-engine, and gtk-industrial-engine. There may be others.
Updated with experience & industrial engines
this is an interesting benchmark, however when it comes to gtk-engines I’m all about stability. I have seen so many engines causing crashes of applications in weird spots. In many cases these engines use cairo_* function calls, where as engines with native gtk calls like xfce seem to be less error prone. I don’t blame it on the engine entirely, in many cases that might be caused by some intercation of libraries. So I’m surprised to see that xfce engine is actually that fast because it is also very very stable! I say that out of the experience of closing bugs in ArchLinux with: “Use another gtk engine and you’ll be fine.”
No wonder why the Xfce Engines yield all the same results; (a nanosecond here and there is in the area of statistical error) They are different colour schemes for the same engine (gtk-xfce-engine).
So you could just leave all but one out and spare a whole page from the article :)
(Whoops… I was looking at the tests visible on the window, and not the total time)
Interesting how different colours & sprites affect performance.
Can you please add charts for each individual test?
Or just a csv file with the data would do, too.
Regards
@stavrosg
Hi,
Excuse me I didn’t understood what you want to say…
What do I have to add to the chart to every individual engine test?
Post updated
Nodoka —–> 0.6.99.1-1
Added charts to gtk-engines and xfce-gtk-engines pages
How about testing QtCurve for GTK? I’ve found this engine to be extremely fast, on par with Simple and Mist!
Post updated
Nodoka —-> 0.7 version
Added QtCurve (@Chris)
thnx man..
this is very usefull
Why not 1) sort the graph by time (it doesn’t seem to be sorted at all right now) and 2) label each bar with number of seconds? I’m not sure what software you’re using, but it ought to be feasible.
Hi gianvito,
you should test the Bluecurve engine. Fastest on my system (Q9450).
Urfe
You should do the following test: modify the Nodoka engine, find the definition (‘engine “nodoka”‘) and replace it with something like this:
engine “nodoka”
{
animation = FALSE
contrast = 1.0
gradients = FALSE
listviewheaderstyle = 0
listviewstyle = 1
menubarstyle = 0
roundness = 0
scrollbarstyle = 2
stripes = FALSE
}
I get 5.3 sec. after the modifications (removing gradients and setting the roundness of the buttons to 0 (none)), 10.8 before.
The roundness of the buttons is the key performance factor in this tests.
And MythBuntu engine?
Post updated (sorry for the long time i’ve missed here)
Excellent blog. good luck.
Can you add excelsior engine? =)
http://www.nanolx.org/index.php?option=com_content&view=article&id=45:excelsior-301-is-out&catid=34:newsnanolx&Itemid=53
Have you just ran the perf program once per theme? It would be interesting to run it 100s of times, perhaps at different warm points in operation, with swap and caches cleared, etc. and take averages.
New post… http://gianvito.wordpress.com/2008/08/20/gtk-engine-benchmarks-2-the-update/2/
@Jon
10 times each themes with averages… (100 is too much :D)
Вау
Never give in to adversity
How about Candido engine? (http://candido.berlios.de/pages/engine.php)
It’s also there in ArchLinux repos
hcengine : High Contrast Engine
used in the high contrast ‘accessibility’ themes.
Would be interesting to see how it compares.
Please include equinox engine http://gnome-look.org/content/show.php?content=121881
It’s funny that Aurora in my computer “feels” the fastest of all?
What with GTK3 now ?