Did you consider that radeon drivers have vsync and vblank on by default limiting the maximum framerates which leads into a lower average?
i expected that to be honest
thanks for the answer and the codes mr Airlie
but...
isn't there any ways to attract more developers in that area ie. an fsf/linux foundation campaign, or try to raise more money (for paying devs) from the industry, or propose some Thesis on Universities on the subject or whatever![]()
Did you consider that radeon drivers have vsync and vblank on by default limiting the maximum framerates which leads into a lower average?
omh again no information about color tiling and pageflipping...
im using color tiling and get much better framerates...
what a bad article...
Why X1000 was not included as base reference?
They're not all mainline yet... FTA: "The Gallium3D driver performance will also improve once the latest color tiling and page-flipping patches have been merged, which should happen soon. Another article is planned at this time looking at the Radeon page-flipping performance as it may bring sizable performance boosts."
FTA: "Unfortunately since AMD dropped support for the R500 hardware and earlier back in 2009 with a Catalyst driver release that is no longer compatible with today's Linux distributions and hardware, we are unable to compare the relative performance directly of the R300g driver + Catalyst to that of today's combination."
The article says clearly:
The Gallium3D driver performance will also improve once the latest color tiling and page-flipping patches have been merged, which should happen soon. Another article is planned at this time looking at the Radeon page-flipping performance as it may bring sizable performance boosts.
Thanks for the benchmark.
Updating some packages to latest rawhide on Fedora 14 works quite nice to test Gallium3D.
Code:su -c 'yum update --enablerepo=rawhide kernel libdrm mesa-* xorg-x11-drv-ati cairo pixman'