Performance change from X in Fedora Core 4 to Fedora Core 5

Carsten Haitzler (The Rasterman) raster at
Sun Jul 9 06:21:11 PDT 2006

On Sun, 09 Jul 2006 13:56:27 +0100 Roo <oscosc at> babbled:

sorry to say that chances are is its the cairo move - cairo using xrender and
xrender basically being horrible in terms of speed on all but the barest few
chipsets if you are lucky. this is not something new - it has been hashed and
rehashed. the problem is basically xrender is falling back to software
rendering as there is no hardware accel - and software xrender is like deciding
to go from los angeles to seattle by relying on continental drift. it has
improved - but it is still bad - and the only major savior is a lot of
extremely funky caching code to shadow drawables in system ram and smartly
transfer only the updated sections TO the card as writes to avoid reads, or to
move al drawables to system ram (try shadowfb if its available for mga drivers
- haven't used them in years so i can't say), change gfx card (ati r200 for
example has render accel (albeit a little buggy)), downgrade gtk back to before
cairo, oe start helping to add acceleration to all the x drivers for xrender.

> Machine: Athlon XP 2000+
> Graphics card: Matrox G550 (AGP, 32Mb).
> Distro: Fedora Core 5
> This machine used to run Fedora Core 4 and the graphics, while still being
> slow compared to stuff under Windows, were tolerable. After upgrading to
> Fedora Core 5, things seem to become almost unbearably slow -- GTK apps in
> particular became even more sluggish than they usually are. I strongly
> suspect the introduction of Cairo (and the change to GTK to use it
> exclusively), but obviously I can't prove that since there is no facility
> to tell GTK not to use Cairo and test it. All of the claimed performance
> improvements in the FC5's version of GNOME were sucked up by this change,
> and more. It's now slower than ever -- you can practically see the window
> contents redrawing. This PC can dual boot with Windows XP, and the
> difference is embarrassing.
> Performance is the elephant in the room when it comes to GTK... so
> discussing it with GTK developers is a pointless task. However, I would
> like to know from X developers: has there been a serious performance
> regression in the support for Matrox cards? Has the RENDER extension lost
> much of its hardware support under these cards in the FC5 version of xorg?
> Or is Cairo just really slow?
> Are only Intel motherboard chipsets, ATI and NVIDIA cards getting
> attention these days?
> _______________________________________________
> xorg mailing list
> xorg at

------------- Codito, ergo sum - "I code, therefore I am" --------------
The Rasterman (Carsten Haitzler)    raster at
Tokyo, Japan (東京 日本)

More information about the xorg mailing list