Fedora ATi driver performs much worse than the vanilla code

[Joshua C.]

2009/3/4 Roberto Ragusa <mail at robertoragusa.it>:
> Joshua C. wrote:
>
>> I did some test and here are the results:
>>
>> "These results are not representative for all the fedora community.
>> They just show how the free radeon driver performs on my pc"
>>
> [snip]
>>
>> These results represent a regression in my case.
>
> Thank you for your detailed tests.
>
> I've also noticed very bad performance on a r300 based laptop
> which used to work well with older Fedora releases (up to F8
> for sure), but it is now practically unusable with rawhide.
>
> My tests were not as extensive as yours, but I didn't manage
> to find an acceptable combination of parameters.
>
> It was particularly confusing for me as my intention with
> rawhide was to test KDE 4 and it was not too clear if the
> slugginess was related to the new KDE stuff (widgets,
> compositing,...) or something more low level.
> But it's low level: moving windows burns CPU and is slow.
>
> I'm really looking forward to having a solution for this issue.
>
> Best regards.
> --
>   Roberto Ragusa    mail at robertoragusa.it
>
> --
> fedora-devel-list mailing list
> fedora-devel-list at redhat.com
> https://www.redhat.com/mailman/listinfo/fedora-devel-list
>

F9 is the best (for my RS482) but because of package updates I need to
upgrade it. Working on F11 with xserver-staff from f9 isn't
encouraged.
I'M wondering why adding new features (like exa, kms) makes the
x-staff work slower. if those new features aren't supported on older
hardware, then the driver should know this and acts accordingly. I
think that a driver-update should at least keep the old level of
functionality and usability. I can enable the compiz staff with kde on
f9 (which is fancy/nice) but the same combination is useless on f11,
despite that both use the same kde.

Maybe the developers should concentrate the radeon driver on r100-r400
and radeonhd on r500++. Or maybe I/we should buy new hardware.