| 1 |
On Fri, July 3, 2009 20:30, Volker Armin Hemmann wrote: |
| 2 |
> Well, AMD uses a fixed cycle - one release a month. They also have a |
| 3 |
> fixed development cycle that takes roughly 3 month (all explained on |
| 4 |
> phoronix a long time ago). So supporting the latest kernels is not easy or |
| 5 |
> even possible. |
| 6 |
|
| 7 |
Very well for them, I don't know how that is relevant, since 2.6.29 was |
| 8 |
released 23-Mar-2009 23:30. So, by now, both the 1 month and the 3 month |
| 9 |
cycles have passed. Still no support. Or maybe my maths are not correct. |
| 10 |
|
| 11 |
> They also only support a few distros. If you insist on using a |
| 12 |
> non-supported kernel, you are on your own. |
| 13 |
|
| 14 |
Crap. All the distros use the same kernel. If it compiles against a vanilla |
| 15 |
kernel, it will compile against a vanilla kernel on *every* distro under |
| 16 |
the sun. There can be problems with the location of the libraries, versions |
| 17 |
of xorg and mesa and the like, that that's out of the scope of this thread, |
| 18 |
which is basically that we will get a working driver for 2.6.30 6 months |
| 19 |
after the kernel is released. By then we will have .35, which won't work |
| 20 |
with fglrx either. That the distro is listed as supported or not doesn't |
| 21 |
change the scenery if you want to use a current kernel for some reason. |
| 22 |
|
| 23 |
> That said - Nvidia might be 'better' when it comes to binary drivers, but |
| 24 |
> they suck when it comes to open source drivers. |
| 25 |
|
| 26 |
Agreed. But at least, you have *one* driver that works. With my hd2600, |
| 27 |
fglrx sucks big time, and neither radeon nor radeonhd works (unless I |
| 28 |
disable dri, and then vesa performs better). |
| 29 |
|
| 30 |
If I have to choose between one closed driver that works perfectly and |
| 31 |
one close driver that works bad+2 open ones that simply don't work, I |
| 32 |
choose the closed driver that works. |
| 33 |
|
| 34 |
> AMD is working on free |
| 35 |
> drivers for all cards - that is something you have to recognize. |
| 36 |
|
| 37 |
As far as I know (I might be wrong) amd isn't working in anything. They |
| 38 |
just released the specifications and some technical sheets. The community |
| 39 |
is doing all the work. |
| 40 |
|
| 41 |
> Besides - if you use the ebuild, even 2.6.30 works. Yes, dmesg is spammed |
| 42 |
> with ugly messages. But who cares? Do you really monitor dmesg non stop? |
| 43 |
|
| 44 |
Who cares?!? mmmm, my /var partition gets filled with +1gb logs everyday, |
| 45 |
I have to delete the logs (yes, losing everything in them) and then |
| 46 |
restart |
| 47 |
the sys logger because while the files are open the space won't be freed. |
| 48 |
The driver does that, and the rest of my programs fail. But, who cares... |
| 49 |
The cpu usage is insane when using fglrx on .30, 60-100% on X when I play |
| 50 |
something in mplayer, so while watching a movie I can't do anything else, |
| 51 |
with the same fglrx version on linux 2.6.28 it takes around 10%. |
| 52 |
|
| 53 |
Who cares? I do. The driver just doesn't work, these are not minor |
| 54 |
annoyances, as you paint them, these are big problems. And the ebuild |
| 55 |
should be eliminated, and the old one depending on <2.6.29 should be |
| 56 |
restored. |
| 57 |
|
| 58 |
I really advice to stay in .28 if you are going to use fglrx. |
| 59 |
-- |
| 60 |
Jesús Guerrero |
Archives