Dual-source driver: catalyst 10.6 Linux 2D acceleration performance test

AMD's recently released catalyst 10.6 driver package not only brings a lot of new features and game performance improvements to Windows systems, but also changes significantly in Linux systems, for example, the new 2D acceleration architecture is used by default, Red Hat Enterprise Linux 5.5 is officially supported, and OpenGL 4.0/3.3 is officially supported. In particular, the 2D acceleration architecture attracts attention.

　　In fact, AMD started to use Direct2D code in the new Linux 2D acceleration architecture from the catalyst 10.2 in July. This is the first time AMD shared Windows 2D acceleration code to the Linux driver, however, this function is hidden by default. You must add an AMDPCSDB key value named Direct2DAccel to enable this function.

AMD was not satisfied with the new acceleration architecture until four months later,It is enabled by default from catalyst 10.6 and a line of "ATI 2D Acceleration Architecture enabled" appears in the/var/log/Xorg.0.log log file, instead of "Using XFree86 Acceleration Architecture (XAA )".

The Linux community welcomed AMD's practice,Many users are very satisfied with the performance and function performance of the new architecture, and the window is minimized and rolled faster and smoother.Of course, some people are in trouble and even some cannot work at all.

In today's test, we put the user experience aside, but only focus on performance. The three drivers involved in the comparison are:Catalyst 10.6And built-in Ubuntu 10.04Catalyst 10.4And open-source driversATI Radeon DRI2 + KMSThe versions of catalyst 10.6 and 10.4 fglrx are 8.74.4 and 8.72.11, respectively. Test Platform configuration includes: Core i3-530 3.32GHz processor, elite H55H-M motherboard, 2 GB memory, OCZ Vertex 64GB solid state drive, Radeon HD 4650 graphics card, Ubuntu 10.04 LTS x86_64 Operating System (Linux 2.6.32 ).

　　Test results:

Catalyst 10.4 is faster than open-source drivers, but the catalyst 10.6 is down by about 18%.

The open-source driver is superior to the catalyst 10.4 and 10.6, respectively, leading 84% and 213%. At the same time, it can be seen that the catalyst 10.6 declined significantly.

Catalyst 10.6 Finally won the open-source driver, but it is still slower than catalyst 10.4.

The column charts of catalyst 10.6 and catalyst 10.4 are longer, but the shorter the catalyst, the better. It takes about 40% of the time for the open-source driver to complete the test.