Linked by Thom Holwerda on Sun 27th Dec 2009 14:12 UTC
KDE Since the stream of news is still pretty much dry, I figured I'd throw in something I've been meaning to talk about for a while now, but really didn't dare to: KDE4's performance. Since experiences with KDE4 seem to widely differ between people, it might be a good idea if we, together, can find a common cause among those of us having problems.
Thread beginning with comment 401549
To view parent comment, click here.
To read all comments associated with this story, please click here.
harryF
Member since:
2005-07-06

Unfortunately, GtkPerf is useless to measure performance. A desktop's performance is not measured by how often it can redraw a simple control in a tight loop.

Reply Parent Score: 1

lemur2 Member since:
2007-02-17

Unfortunately, GtkPerf is useless to measure performance. A desktop's performance is not measured by how often it can redraw a simple control in a tight loop.


Nevertheless, a desktop graphics system and its configuration (incorporating GPU, card, driver and Xorg and WM software stack) which takes less time to render the GtkPerf tests is surely faster than another such system and configuration which takes longer.

Reply Parent Score: 2

Rahul Member since:
2005-07-06

Your tests are correct. There is indeed a performance hit on 3D with ATI when KMS is enabled. Dave Airlie from Red Hat is working on getting this fixed.

In Fedora 12, 3D support for ATI with Radeon is still considered experimental for this reason. You will have to install mesa-dri-drivers-experimental package from the repo.

Reply Parent Score: 2

saynte Member since:
2007-12-10

"Unfortunately, GtkPerf is useless to measure performance. A desktop's performance is not measured by how often it can redraw a simple control in a tight loop.


Nevertheless, a desktop graphics system and its configuration (incorporating GPU, card, driver and Xorg and WM software stack) which takes less time to render the GtkPerf tests is surely faster than another such system and configuration which takes longer.
"

Sure, something got faster. However, if it doesn't translate into a faster experience for the user then it doesn't matter either.

That's the main problem with GtkPerf, you can't be sure it will mean that the interface has gotten faster under real-world conditions.

That's the same reason no one uses glxgears for benchmarking, it's useless. You use a game, or something that represents the way the graphics card is actually utilized to benchmark 3d performance.

Reply Parent Score: 1