Now I do not use the AMD's closed source driver, I have no problems with the 3D performance, comparing my Nvidia GTX 560 with my other system with a Radeon 5870 for the type of work I do gives on the whole comparable good results with both of the closed src drivers. The reason I do not use the driver is a particular aspect of 2D performance and that is redrawing the background when a window closes or changing the back ground image, the performance can only be described as pathetic. On my 2560x1600 this takes around 1/5th to 1/6th of a second. Even on my 1920x1200 monitors you can see the background being redrawn. Consider on the larger monitor the card only has to transferr 16 megabytes of information a 1/6th of a second is patheticly slow basicly 100 megabytes /sec transfer rate, this on a system with 4 cpus running above 3gigahz. The Xorg.log file does say that 2D accelaratiion is enabled.
Note, the card performs well in both 2 and 3D under Windows XP.
For me this makes the card total unsuitable, but with the open source driver even without DRI support in the Xserver ATI driver (Debian stable), but of course with Radeon KMS support in the kernel the redrawing of the background is done in a blink of an eye and I assume this is being done just by the CPU. I find the difference between Catalyst and the open source driver inexplicable.
Hmmmm, the above is a bit of a ramble I am afraid, hope it makes sense, Now anybody have any idea of what is happenning here, No one else appears to have reported this phenomomnon but I have found this to be consistent with the Catalyst driver since I purchased the 2560x1500 monitor, a Samsung 305T.
This has happened in all versions of Catalyst I have tried including the lastest one.
I would really prefer to use the AMD cards for 3D development but this is for me anyway a real show stopper, hence the Nvidia GTX560 superover clocked card which performs well in both 2 and 3D.