Canonical Publishes ATI Catalyst 8.10 Beta

[phoronix]

Phoronix: Canonical Publishes ATI Catalyst 8.11 Beta

X Server 1.5 was officially released last month with X.Org 7.4, but there had been server pre-releases going back to earlier this year. Fedora 9 had even shipped with an early version version of X Server 1.5. For those using the open-source X.Org drivers, running the latest server is not a big deal, but those with ATI or NVIDIA binary drivers they sometimes can be slow in supporting the latest version. NVIDIA has supported X Server 1.5 for a number of weeks now, but ATI has yet to update their Catalyst Linux driver with such support. With Ubuntu 8.10 being released in two weeks and it's using this newest X Server, how will ATI graphics cards be supported? Well, an interesting event has occurred and we will tell you what has happened in this article.

http://www.phoronix.com/vr.php?view=12957

[aaaantoine]

I didn't see this coming.

I was expecting to have to switch to xorg-driver-ati when I upgrade to Ubuntu 8.10. That may not have bothered me so much, because the latest version reportedly supports compositing on the Radeon Xpress 1100.

[m4rgin4l]

This is good news. If all goes well I'll be able to run glxears accelerated on Fedora 11 with my HD4850. Yeeeey!

Jokes aside, I'd like to see this happening more often. Given the amount of Ubuntu users, this could really help speed up the development of the driver.

[cjr2k3]

I'm going to buy a new PC now. Let me test this beta driver and then decide wether is ATI or ATI the graphic card :P. Just to know if this driver behave well with my ATI X1400.

I always on Ubuntu so, I hope this works fine or else I'm going to stick to my laptop to play WoW.

[d2kx]

Great move :)

But Michael, are you sure that you got the release streams right? Catalyst 8.8 was 8.52, 8.9 is 8.53 and so this driver is probably the Catalyst 8.10 driver (final driver I guess, or a Catalyst 8.10 hotfix). Correct me if I'm wrong, but I am... quite sure.

[bash]

Yea I think the article got the versions wrong. Or else that would also mean the if they need to ship the November driver for XServer 1.5 support, that the October still doesn't have it.

[Pfanne]

sweet! (ten characters)

[Michael]

But Michael, are you sure that you got the release streams right? Catalyst 8.8 was 8.52, 8.9 is 8.53 and so this driver is probably the Catalyst 8.10 driver (final driver I guess, or a Catalyst 8.10 hotfix). Correct me if I'm wrong, but I am... quite sure.

Yes, you are correct. There was a miscommunication. It should be fixed now.

[The_Monkey_King]

See my observation on an earlier topic. http://www.phoronix.com/forums/showthread.php?t=12700&page=3

Shuttleworth is presenting Ubuntu + AMD as the killer alternative to the Wintel platform. I believe he thinks he can sell to the HPs and the Dells of this world as a lower cost, higher margin alternative. This is why he gets more preferential treatment than Phoronix, OpenSUSE or RedHat.

[pierluc]

Is it possible that Catalyst 8.10 will be the beta on Catalyst 8.11? If yes, it's possible that the version of Ubuntu is not the really Catalyst 8.11 but Catalyst 8.10 beta.

[sundown]

Some good deeds by AMD. People should be really happy they took over ATI :D

[Kano]

That driver is crappy as all drivers since 8.40.4 which was the last which showed PointSprites, most simple test is to use gl2benchmark, which is even possible with intrepid using:

deb http://kanotix.com/files/fix/gl2benchmark.sid ./
deb-src http://kanotix.com/files/fix/gl2benchmark.sid ./

Try with nvidia to see how it should look like...

[Michael]

That driver is crappy as all drivers since 8.40.4 which was the last which showed PointSprites, most simple test is to use gl2benchmark, which is even possible with intrepid using:

deb http://kanotix.com/files/fix/gl2benchmark.sid ./
deb-src http://kanotix.com/files/fix/gl2benchmark.sid ./

Try with nvidia to see how it should look like...

It works with FlightGear... Can you find any other OpenSceneGraph program where PointSprites don't work?

[Kano]

Isn't one test app enough? That's a test tool, that should test some needed functions which definitely work with nvidia. Of course the production apps are not all opensource. Maybe somebody else knows additional examples.

[EmbraceUnity]

Intrepid is basically perfect now. I just wish it could have come with OpenOffice.org 3.0. I don't think that is possible now, but you never know.

[Michael]

Isn't one test app enough? That's a test tool, that should test some needed functions which definitely work with nvidia. Of course the production apps are not all opensource. Maybe somebody else knows additional examples.

gl2benchmark is a no longer maintained tool though...

[yoshi314]

i guess that's the end of the line in amd's beta catalyst program for canonical. i suspect ati's going to be pretty angry about this.

unless they changed their beta policy.

[Kano]

The source is there, everybody can add new features. And it just works the way it is.

[sundown]

i guess that's the end of the line in amd's beta catalyst program for canonical. i suspect ati's going to be pretty angry about this.

unless they changed their beta policy.

I bet that was a co-ordinate move.

[Kano]

Also I see no reason why the tool should get ATI extra workarounds, it worked before it has to work now!

[cjr2k3]

I bet that was a co-ordinate move.

I bet in a few hours there will be a news on the main page called:

ATI CATALIST 8.10 RELEASED

(Once michael gets UVD working :P)

[rbmorse]

Thursday, more likely, or maybe Thursday a week.

[bash]

According to the leaked ATI driver roadmap 8.54 should have been out for a while ;)

Link: http://forums.guru3d.com/showthread.php?p=2846561

[NeoBrain]

From the Catalyst 8.9 thread:
Wow, just read the phoronix article.
I'm really looking forward to the upcomming 8.10 driver, hoping I'll get XServer 1.5 support on my birthday (15th Oct 08)

Well, guess I was right actually :D
I'm really interested in what has been going on there "behind the scenes", hopefully the coming days clear that thing up.
Maybe ATI is really into public beta testing these days...

[Extreme Coder]

But what's so groundbreaking about this driver other than X Server 1.5 support? I don't see what the big issue is..

And also, I have to agree with Kano, 3D apps shouldn't have to make workarounds for ATI cards, one 3D rendering command should work for all..

[enzobelmont]

how can i use this driver on my gentoo system???

or should i wait for official release??

i'm tired of radeon driver...

[Pfanne]

But what's so groundbreaking about this driver other than X Server 1.5 support? I don't see what the big issue is..

And also, I have to agree with Kano, 3D apps shouldn't have to make workarounds for ATI cards, one 3D rendering command should work for all..

i hope the ati developers were just trying to get on par with the windowsdrivers, featurewise...
so they are going to fix all the bugs they created on their way with the future releases...
otherwise they will go nowhere with their drivers having a lot of features with none working correctly!

[SheeEttin]

Thursday, more likely, or maybe Thursday a week.

Close. From the RSS feed:
Linux Proprietary Driver - Catalyst 8.10 Release: x86 and x86_64 variants
Driver version: 8.542
NEW FEATURES: (1) AMDTM CAL: Computer Abstraction Layer runtime and libraries support. (2) Support for Red Hat Enterprise Linux 4.7
ISSUES RESOLVED: (1) X startup failure on Radeon X1xxx and later AGP cards. (2) Xserver crash after logging out multiple times. (3) Bad Xv video playback performance on AGP cards. (4) Artefacts left on the screen when quickly moving an Xv video window.
:D

Installer: https://www2.ati.com/drivers/linux/ati-driver-installer-8-10-x86.x86_64.run

[maggo]

the 8.10 seems to be available (not announced yet)
https://a248.e.akamai.net/f/674/9206/0/www2.ati.com/drivers/linux/ati-driver-installer-8-10-x86.x86_64.run

[Michael]

Note that it's 8.542 and not 8.543.

[Yfrwlf]

Why is anyone praising this? Distro companies have started implementing additional ways to get users on *their* distro, now it's "I've got the exclusive driver over here!"? Exclusivity is the exact opposite of being open. All Linux users should have equal access to drivers for their cards. This makes me not want to support AMD if they are going to try to help further segregate the Linux platform. AMD should have a developer site of some sort where this driver is available, and everyone should be able to get it from there.

[bash]

Why is anyone praising this? Distro companies have started implementing additional ways to get users on *their* distro, now it's "I've got the exclusive driver over here!"? Exclusivity is the exact opposite of being open. All Linux users should have equal access to drivers for their cards. This makes me not want to support AMD if they are going to try to help further segregate the Linux platform. AMD should have a developer site of some sort where this driver is available, and everyone should be able to get it from there.

I don't think it's "I've got the exclusive driver over here!". Ubuntu is the first of the major distros that all release new version within the next months. And all of them will be using XServer 1.5. For example if another distro would have been out end of October instead Ubuntu, they would have gotten the new "beta" driver.

[biasquez]

make -C /lib/modules/2.6.27-7-generic/build SUBDIRS=/var/lib/dkms/fglrx/8.542/build modules
make[1]: Entering directory `/usr/src/linux-headers-2.6.27-7-generic'
CC [M] /var/lib/dkms/fglrx/8.542/build/firegl_public.o
/var/lib/dkms/fglrx/8.542/build/firegl_public.c: In function ‘__ke_printk’:
/var/lib/dkms/fglrx/8.542/build/firegl_public.c:1915: warning: format not a string literal and no format arguments
/var/lib/dkms/fglrx/8.542/build/firegl_public.c: In function ‘__ke_vm_phys_addr_str’:
/var/lib/dkms/fglrx/8.542/build/firegl_public.c:3522: warning: return makes pointer from integer without a cast
/var/lib/dkms/fglrx/8.542/build/firegl_public.c:3523: warning: return makes pointer from integer without a cast
/var/lib/dkms/fglrx/8.542/build/firegl_public.c:3524: warning: return makes pointer from integer without a cast
/var/lib/dkms/fglrx/8.542/build/firegl_public.c:3526: warning: return makes pointer from integer without a cast
/var/lib/dkms/fglrx/8.542/build/firegl_public.c: In function ‘KCL_enable_pat’:
/var/lib/dkms/fglrx/8.542/build/firegl_public.c:4063: error: too many arguments to function ‘smp_call_function’
/var/lib/dkms/fglrx/8.542/build/firegl_public.c: In function ‘KCL_disable_pat’:
/var/lib/dkms/fglrx/8.542/build/firegl_public.c:4082: error: too many arguments to function ‘smp_call_function’
/var/lib/dkms/fglrx/8.542/build/firegl_public.c: At top level:
/var/lib/dkms/fglrx/8.542/build/firegl_public.c:5774: warning: initialization from incompatible pointer type
/var/lib/dkms/fglrx/8.542/build/firegl_public.c:5800: warning: initialization from incompatible pointer type
make[2]: *** [/var/lib/dkms/fglrx/8.542/build/firegl_public.o] Error 1
make[1]: *** [_module_/var/lib/dkms/fglrx/8.542/build] Error 2
make[1]: Leaving directory `/usr/src/linux-headers-2.6.27-7-generic'
make: *** [kmod_build] Error 2

[nnotimportant]

I don't think it's "I've got the exclusive driver over here!". Ubuntu is the first of the major distros that all release new version within the next months. And all of them will be using XServer 1.5. For example if another distro would have been out end of October instead Ubuntu, they would have gotten the new "beta" driver.

It doesn't appear to actually BE a beta driver. Its published on amd's website. Someone at canonical probably just got lucky when browsing around AMD's ftp.

And the fact is that this is just when the driver is due. Its the first AMD driver released more than a couple of days after xserver 1.5.

[groo_pcd]

Anyone tested the tearing with video and compiz/kwin 3d?

Did they finally fixed it? (open source radeon (xf86-ati) fixed this eons ago, they even have xv AND texturedvideo suport across the 200 up to 600 series).

I'll just install it if/when this gets fixed, otherwise i'm very happy with the open source driver, the only downside is the lack of a memory manager ence 1.3 opengl only :( but this is about to change with GEM :) go open source radeon team!

[oyvind]

It's officially released now, at least. And is it just me, or has scrolling performance in Firefox w/Compiz been vastly improved in this release ? I'm on an X1400 mobile radeon.

[pjjanak]

make -C /lib/modules/2.6.27-7-generic/build SUBDIRS=/var/lib/dkms/fglrx/8.542/build modules
make[1]: Entering directory `/usr/src/linux-headers-2.6.27-7-generic'
CC [M] /var/lib/dkms/fglrx/8.542/build/firegl_public.o
/var/lib/dkms/fglrx/8.542/build/firegl_public.c: In function ‘__ke_printk’:
/var/lib/dkms/fglrx/8.542/build/firegl_public.c:1915: warning: format not a string literal and no format arguments
/var/lib/dkms/fglrx/8.542/build/firegl_public.c: In function ‘__ke_vm_phys_addr_str’:
/var/lib/dkms/fglrx/8.542/build/firegl_public.c:3522: warning: return makes pointer from integer without a cast
/var/lib/dkms/fglrx/8.542/build/firegl_public.c:3523: warning: return makes pointer from integer without a cast
/var/lib/dkms/fglrx/8.542/build/firegl_public.c:3524: warning: return makes pointer from integer without a cast
/var/lib/dkms/fglrx/8.542/build/firegl_public.c:3526: warning: return makes pointer from integer without a cast
/var/lib/dkms/fglrx/8.542/build/firegl_public.c: In function ‘KCL_enable_pat’:
/var/lib/dkms/fglrx/8.542/build/firegl_public.c:4063: error: too many arguments to function ‘smp_call_function’
/var/lib/dkms/fglrx/8.542/build/firegl_public.c: In function ‘KCL_disable_pat’:
/var/lib/dkms/fglrx/8.542/build/firegl_public.c:4082: error: too many arguments to function ‘smp_call_function’
/var/lib/dkms/fglrx/8.542/build/firegl_public.c: At top level:
/var/lib/dkms/fglrx/8.542/build/firegl_public.c:5774: warning: initialization from incompatible pointer type
/var/lib/dkms/fglrx/8.542/build/firegl_public.c:5800: warning: initialization from incompatible pointer type
make[2]: *** [/var/lib/dkms/fglrx/8.542/build/firegl_public.o] Error 1
make[1]: *** [_module_/var/lib/dkms/fglrx/8.542/build] Error 2
make[1]: Leaving directory `/usr/src/linux-headers-2.6.27-7-generic'
make: *** [kmod_build] Error 2

Ditto. Anyone know why this is happening?

[Michael]

It doesn't appear to actually BE a beta driver. Its published on amd's website. Someone at canonical probably just got lucky when browsing around AMD's ftp.

And the fact is that this is just when the driver is due. Its the first AMD driver released more than a couple of days after xserver 1.5.

No, the official Catalyst 8.10 is 8.54.2. Canonical is using 8.54.3.

[groo_pcd]

It's officially released now, at least. And is it just me, or has scrolling performance in Firefox w/Compiz been vastly improved in this release ? I'm on an X1400 mobile radeon.

How about xv/textured video?
if you do a xvinfo does it show anything?
Does it tear when playing video with mplayer/xine/whatever? Or does it just work with the X11 driver?

Cheers.

[bash]

It's officially released now, at least. And is it just me, or has scrolling performance in Firefox w/Compiz been vastly improved in this release ? I'm on an X1400 mobile radeon.

I can confirm the scrolling. Didn't notice any real difference between the open-source driver and fglrx. Haven't owned an ATI card in a while, so I just thougth it was due to my more powerfull machine.

Haven't tested any video playing, but at least Flash video playback (tested on YouTube) seemed to be kinda choppy/slow/laggy. Funnily enough it actually seem to run smoother fullscreen, than in the normal size.

[conholster]

Video tearing, yes.

xvinfo:

[conny@gomez ~]$ xvinfo
X-Video Extension version 2.2
screen #0
no adaptors present


I wonder when the changelog will be available, dont think its very long...

For the record HD3850 pcie

[oyvind]

How about xv/textured video?
if you do a xvinfo does it show anything?
Does it tear when playing video with mplayer/xine/whatever? Or does it just work with the X11 driver?

Cheers.

Textured Video in Catalyst 8.10 works fine and is v-synced, but not in Compiz. Then you get your nice flickering mess. Didn't really expect a fix for this though, since AFAIK fglrx uses the standard X DRI architecture, which does not support things such as accelerated redirected rendering (aka DRI2). It would be nice if AMD could implement some sort of work-around just for Textured Video output, but I'm not getting my hopes up. Using the "Unredirect Fullscreen Windows" option in Compiz for now, which makes fullscreen XV-output OK. I'm on Ubuntu Hardy, btw.

[oyvind]

[...]

Haven't tested any video playing, but at least Flash video playback (tested on YouTube) seemed to be kinda choppy/slow/laggy. Funnily enough it actually seem to run smoother fullscreen, than in the normal size.

That might be the Compiz "Unredirect Fullscreen Windows" option in action. If fullscreen, your window updates aren't routed through compositing (unredirected), and thus the raw performance "unfortunately" improves :). But hey, fancy 3D-effects aren't for free, sadly. They come at a performance cost for all things being composited. Whether noticeable or not depends heavily on the hardware in question. I have no trouble noticing a general loss of 2D performance with Compiz on my ATI X1400 mobile Radeon (128MB RAM) and Intel Core Duo 2GHz laptop (compared to a plain 2D desktop with Metacity).

[oyvind]

Haven't tested any video playing, but at least Flash video playback (tested on YouTube) seemed to be kinda choppy/slow/laggy. Funnily enough it actually seem to run smoother fullscreen, than in the normal size.

Oh, and if you haven't already, make sure you upgrade to the latest official Flash player 10 from Adobe. I find the performance increase in the latest Flash-releases to be rather nice.

[groo_pcd]

Textured Video in Catalyst 8.10 works fine and is v-synced, but not in Compiz. Then you get your nice flickering mess. Didn't really expect a fix for this though, since AFAIK fglrx uses the standard X DRI architecture, which does not support things such as accelerated redirected rendering (aka DRI2). It would be nice if AMD could implement some sort of work-around just for Textured Video output, but I'm not getting my hopes up. Using the "Unredirect Fullscreen Windows" option in Compiz for now, which makes fullscreen XV-output OK. I'm on Ubuntu Hardy, btw.

Not completely true for video, radeon (opensource) also uses the default DRI arch and is HAS xv/textured that DOES WORK with compiz and every other composited window manager) WITHOUT the flickering.

If i do a xvinfo with xf86-radeon-ati:

X-Video Extension version 2.2
screen #0
Adaptor #0: "ATI Radeon Video Overlay"
number of ports: 1
port base: 57
operations supported:
(snip)
Adaptor #1: "Radeon Textured Video"
number of ports: 16
port base: 58
(snip)

I little diferent, dont you think?
Now, if dave airlie and friends can do that for the open source driver (video without flickering and video without flickering RENDERED by gpu (aka texturedvideo) why cant all mighty AMD which has 1 - the specs, 2 - the man power?

Anyway, just my 2 cents..

[oyvind]

No, the official Catalyst 8.10 is 8.54.2. Canonical is using 8.54.3.

I've just installed the official Catalyst 8.10 from http://ati.amd.com/support/drivers/linux/linux-radeon.html, and the AMD control panel reports driver version as 8.54.3.

[bash]

That might be the Compiz "Unredirect Fullscreen Windows" option in action. If fullscreen, your window updates aren't routed through compositing (unredirected), and thus the raw performance "unfortunately" improves :). But hey, fancy 3D-effects aren't for free, sadly. They come at a performance cost for all things being composited. Whether noticeable or not depends heavily on the hardware in question. I have no trouble noticing a general loss of 2D performance with Compiz on my ATI X1400 mobile Radeon (128MB RAM) and Intel Core Duo 2GHz laptop (compared to a plain 2D desktop with Metacity).

Ah ok. So I actually get the benefit of GPU acceleration now when playing fullscreen videos. Already wondered why fullscreen flash got so much faster on my other test box when I upgraded to Ubuntu 8.10 alpha. At least with my hardware I don't notice a big impact with compositing enabled. Although I find Compiz to be more laggy on my ATI machine than on the other Nvidia one, even though the ATI card is brand new.

Oh, and if you haven't already, make sure you upgrade to the latest official Flash player 10 from Adobe. I find the performance increase in the latest Flash-releases to be rather nice.

Yea already using the Flash 10 final release.

[Fran]

I've installed xorg-server-1.5 and ati-drivers-8.10 in gentoo, but I get the same error as with 8.9 (undefined symbol miZeroLineScreenIndex). Oh, and I had to keep the patch for 2.6.27 kernels, otherwise it won't compile (does this driver fix ANYTHING? :confused:).

Reverting now to xorg-server-1.4

[Yfrwlf]

I don't think it's "I've got the exclusive driver over here!". Ubuntu is the first of the major distros that all release new version within the next months. And all of them will be using XServer 1.5. For example if another distro would have been out end of October instead Ubuntu, they would have gotten the new "beta" driver.

And I'm just saying anyone should be able to get it is all. On a related note/rant, I'd also go a step further and *gasp* ask for easy installation of that driver as well as the new version of X, for those who want to upgrade, instead of waiting around for some distro to install it for them, but I guess that's asking for too much and it won't happen til distro package managers get over themselves and adopt compatibility with at least one standard common format. I know I know, I'm such a rebel for challenging the status quo. Dear lord he wants Linux to be easy, what a freak. ^^

[oyvind]

Not completely true for video, radeon (opensource) also uses the default DRI arch and is HAS xv/textured that DOES WORK with compiz and every other composited window manager) WITHOUT the flickering.

If i do a xvinfo with xf86-radeon-ati:

X-Video Extension version 2.2
screen #0
Adaptor #0: "ATI Radeon Video Overlay"
number of ports: 1
port base: 57
operations supported:
(snip)
Adaptor #1: "Radeon Textured Video"
number of ports: 16
port base: 58
(snip)

I little diferent, dont you think?
Now, if dave airlie and friends can do that for the open source driver (video without flickering and video without flickering RENDERED by gpu (aka texturedvideo) why cant all mighty AMD which has 1 - the specs, 2 - the man power?

Anyway, just my 2 cents..

I agree, that's why I mentioned they should implement some sort of work-around for Textured Video, so that it does work in a composited environment. I would be satisified if they could just eliminate the flickering (through some hack or not, I don't care :). What happens with the video window when using the open source driver in a composited environment, and you drag another window over it ? Do, for instance, the drop shadows render correctly onto the video ? If so, I would be really surprised .. But hey, AMD: eliminate the flickering, and I would be satisfied.

Oh, and just to be clear, have you tested both of these XVideo adaptors in a composited environment ? AFAIK, classic video overlay usually has dedicated video-hardware parts (older graphics cards) which works at a level above everything else (i.e. app/X just draws a certain color, and the GFX hardware scales and puts video images in the area covered by this color), while textured video uses, well 3D textures and the GPU .. I also have an older laptop with an Intel 915 graphics card in it, and this renders XVideo fine in a composited environment, using the Intel drivers. But rendering is not correct when the video window is partially covered by another window. Then you get artifacts/solid coloring where drop shadows should be, which proves that it is in fact a hack.

[Kano]

And does _not_ include the x740* dirs! Not even 2.6.27 support! At least the old patch for 8-9 works...

[SheeEttin]

Hmm... 3D doesn't seem to work. "LIBGL_DEBUG=verbose glxinfo" gives this:
libGL: XF86DRIGetClientDriverName: 8.54.3 fglrx (screen 0)
libGL: OpenDriver: trying /usr/lib/dri/fglrx_dri.so
libGL error: driver exports no extensions (/usr/lib/dri/fglrx_dri.so: undefined symbol: __driDriverExtensions)
libGL: OpenDriver: trying /usr/lib/dri/swrast_dri.so
libGL error: dlopen /usr/lib/dri/swrast_dri.so failed (/usr/lib/dri/swrast_dri.so: undefined symbol: _glapi_tls_Context)
libGL error: unable to load driver: swrast_dri.so
libGL error: reverting to indirect rendering
I've managed to get this close to getting 3D acceleration again... Anyone know what's up with this?

[groo_pcd]

I agree, that's why I mentioned they should implement some sort of work-around for Textured Video, so that it does work in a composited environment. I would be satisified if they could just eliminate the flickering (through some hack or not, I don't care :). What happens with the video window when using the open source driver in a composited environment, and you drag another window over it ? Do, for instance, the drop shadows render correctly onto the video ? If so, I would be really surprised .. But hey, AMD: eliminate the flickering, and I would be satisfied.

Oh, and just to be clear, have you tested both of these XVideo adaptors in a composited environment ? AFAIK, classic video overlay usually has dedicated video-hardware parts (older graphics cards) which works at a level above everything else (i.e. app/X just draws a certain color, and the GFX hardware scales and puts video images in the area covered by this color), while textured video uses, well 3D textures and the GPU .. I also have an older laptop with an Intel 915 graphics card in it, and this renders XVideo fine in a composited environment, using the Intel drivers. But rendering is not correct when the video window is partially covered by another window. Then you get artifacts/solid coloring where drop shadows should be, which proves that it is in fact a hack.

I've tested it yes, since i use it daily (even openned a few bug reports for radeon).

here it how it works:

If i activate the first adaptor (XV) i get hardware 2d accelerated video, i can out windows over it, but the shadows etc arent rendered and if for example i use freewins (compiz rotaded window plugin) the video doesnt follow the window. Now is i use the second adaptor (textured video), i get 3d aceleratet video WITHOUT flickering and the windows+video respects the compositing, i can rotate the window + video, change opacity, etc etc.. quitte impressive.. i have all my video players configured for textured now.

So thats my point.. textured should be enough for fglrx but although fglrx suports textured for my rs485 it doesnt talk with the composite (dont know how the open source crowd did it).

Hope this clarifies things.

[d2kx]

No, the official Catalyst 8.10 is 8.54.2. Canonical is using 8.54.3.

Nearly. Having been offline for over a week, I have a huge need to post anything, so here is some boring stuff:

Catalyst 8.10 Release/Final
Release Identifier: 8.542
Base Driver Version: 8.54.3

Catalyst 8.10 Hotfix
Release Identifier: 8.543
Base Driver Version: 8.54.3

[oyvind]

I've tested it yes, since i use it daily (even openned a few bug reports for radeon).

here it how it works:

If i activate the first adaptor (XV) i get hardware 2d accelerated video, i can out windows over it, but the shadows etc arent rendered and if for example i use freewins (compiz rotaded window plugin) the video doesnt follow the window. Now is i use the second adaptor (textured video), i get 3d aceleratet video WITHOUT flickering and the windows+video respects the compositing, i can rotate the window + video, change opacity, etc etc.. quitte impressive.. i have all my video players configured for textured now.

So thats my point.. textured should be enough for fglrx but although fglrx suports textured for my rs485 it doesnt talk with the composite (dont know how the open source crowd did it).

Hope this clarifies things.

Yes it does, cool, I didn't know they've made textured video work that well with the open source radeon driver. How about OpenGL-applications in composited environment with the radeon-drivers, do they work just fine ? Just curious :) ..

[pjjanak]

Anyone have any clue as to what those of us getting compile errors for the kernel module should do?

Peter

[groo_pcd]

Yes it does, cool, I didn't know they've made textured video work that well with the open source radeon driver. How about OpenGL-applications in composited environment with the radeon-drivers, do they work just fine ? Just curious :) ..

No, for that you'll need DRI2.. should be good to go for xorg 1.6 (fingers crossed), airlie and friends are changing radeon to the new GEM api (actually they are wrapping TTM around it), so with a bit of luck, we might see DRI2 radeon drivers around early 2009. :)

cheers

[energyman]

I've installed xorg-server-1.5 and ati-drivers-8.10 in gentoo, but I get the same error as with 8.9 (undefined symbol miZeroLineScreenIndex).

when do you get that error? on loading or when compiling?

[d2kx]

Catalyst 8.10 does not support Xorg 7.4, Fran.

[Gordy]

The driver (8.54.3) does not work for my sapphire 9600XT 128MB AGP card..back to driver 8.37..the last semi stable driver for my hardware..

The tests run:
- glxgears runs
- flight gear seems to work but there are gitters/pauses..not smooth..usable
- Lego StarWars II under wine 1.1.5 with useGLSL enabled..lines/triangles drawn to the center of screen. Also, gitters/pauses..not smooth..basically unusable
- video seems to work window/fullscreen but I see the cpu usage jump from 20% to 40%
- compile of wine does not cause random GCC crashes like the 8.9 driver did

BTW, I am not planning to upgrade my hardware until the bugs get worked out..or in other words I need to see a full functioning driver first before upgrading..still waiting four years later..

Cheers,
Gordy


System config:
SAPPHIRE ATI RADEON 9600XT 128MB AGP (8x) - (ATI Technologies Inc RV350 AR [Radeon 9600])
CRT 19" monitor: Philips 109B (109B10) with a resolution of 1600x1200 or 1280x1024
Motherboard: ASUS K8V SE Deluxe/ VIA K8T800 & VIA VT8237
Processor type: AMD64 3000+
Amount of memory: 1GB
Distro used: Ubuntu 7.10 Gutsy Gibbon (32-bit)
Kernel version: 2.6.22-15(.22)

[Spitfire]

What is this driver good for if it doesn't support 7.4? I thought this should go into Intrepid which uses 7.4 right?


irrlicht@home:~/Documents/ati$ sudo dpkg -i fglrx-kernel-source_8.542-0ubuntu1_amd64.deb
Selecting previously deselected package fglrx-kernel-source.
(Reading database ... 182769 files and directories currently installed.)
Unpacking fglrx-kernel-source (from fglrx-kernel-source_8.542-0ubuntu1_amd64.deb) ...
Setting up fglrx-kernel-source (2:8.542-0ubuntu1) ...
Adding Module to DKMS build system
Doing initial module build

Error! Build of fglrx.ko failed for: 2.6.27-7-generic (x86_64)
Consult the make.log in the build directory
/var/lib/dkms/fglrx/8.542/build/ for more information.
Installing initial module

Error! Could not locate fglrx.ko for module fglrx in the DKMS tree.
You must run a dkms build for kernel 2.6.27-7-generic (x86_64) first.
Done.


Daniel

[Fran]

when do you get that error? on loading or when compiling?
When trying to start X. It compiles fine (but only with the patch for 2.6.27 kernels).

Catalyst 8.10 does not support Xorg 7.4, Fran.
So it seems. I assumed that the official version would be the same as the ubuntu version released earlier. This is perfect, is ubuntu the only distribution that supports fglrx + xorg-server-1.5 now? WTF?

[Kano]

You can use the multiverse repository with intrepid but not 8-10 final.

[Bigon]

I cannot load the driver on my laptop with a mobility 9700 :/
[ 3638.630594] fglrx: module license 'Proprietary. (C) 2002 - ATI Technologies, Starnberg, GERMANY' taints kernel.
[ 3638.713007] [fglrx] Maximum main memory to use for locked dma buffers: 1896 MBytes.
[ 3638.714365] [fglrx:drm_alloc] *ERROR* [driver] Allocating 0 bytes
[ 3638.714377] [fglrx:firegl_init_device_list] *ERROR* Out of memory when allocating device heads
[ 3638.714389] [fglrx:firegl_init_module] *ERROR* firegl_init_devices failed

[korpenkraxar]

My $0.02 about running 8.542 on a Debian Thinkpad with ATI X1400:

* Builds and installs nicely on Debian/Sid i386.
* mplayer plays Big Buck Bunny 1080p MP4 great using -vo xv (but how can we use UVD?)
* glxgears fps is down from about 4200 in 8.9 to 3000-3500
* Can not exit fullscreen Nexuiz, mouse moves but no desktop.
* Compiz is still unusable due to painfully slow window resizing (does anyone know if there are any hacks to improve this?) and flickering video

While video is more important than 3D for me and does seem to perform just fine when compositing is off, I still have to confirm that ATI's fglrx driver remains among the biggest obstacles for achieving Linux on the desktop. The driver is so rough around the edges and still prone to unpredictable regressions in each release. But least it feels like there is a trend towards releasing slightly improved (on average) drivers since AMD got involved.

[Pulfer]

While video is more important than 3D for me and does seem to perform just fine when compositing is off, I still have to confirm that ATI's fglrx driver remains among the biggest obstacles for achieving Linux on the desktop. The driver is so rough around the edges and still prone to unpredictable regressions in each release. But least it feels like there is a trend towards releasing slightly improved (on average) drivers since AMD got involved.

I totally agree with you. 99% problems that I meet using linux on my desktop are caused by very poor fglrx driver quality. I've been buying only ATI cards since they released Radeon 7500. But looking at the driver's extremely slow progress I guess 4850 is the last ATI card I bought. :(

[energyman]

Note that it's 8.542 and not 8.543.

[ 580.240478] [fglrx] module loaded - fglrx 8.54.3

says the module from the driver I downloaded from amd's website.

[linuxjacques]

Anyone tested the tearing with video and compiz/kwin 3d?

Did they finally fixed it? (open source radeon (xf86-ati) fixed this eons ago, they even have xv AND texturedvideo suport across the 200 up to 600 series).




I don't know what you're talking about with the "up to 600 series" quote.

According to http://www.x.org/wiki/RadeonFeature R600 (what I have)
has none of these features implemented.

[rbmorse]

How about xv/textured video?
if you do a xvinfo does it show anything?
Does it tear when playing video with mplayer/xine/whatever? Or does it just work with the X11 driver?

Cheers.

The Ubuntu version, whatever that really is, returns this from xvinfo:X-Video Extension version 2.2
screen #0
Adaptor #0: "ATI Radeon AVIVO Video"
number of ports: 4
port base: 131
operations supported: PutImage
supported visuals:
depth 24, visualID 0x23
depth 24, visualID 0x24
depth 24, visualID 0x25
Lots and lots more up to 0x72

...

[Draco-LVNH]

Hello everyone, is there a way gentoo users can have that "hotfix" version of the 8.10 driver release ?, because i can not find it, can someone write the URL for the package of the "hotfix" for Ubuntu for 64 Bit ?, so i can extract the binary parts and place them by hand on my system, thanks by the way.

[Nille_kungen]

For me the new driver seems to work good.
But i can't start amdcccle

amdcccle: xcb_io.c:352: _XReply: Assertion `!dpy->xcb->reply_data' failed.
Aborted

Not that i use it but atleast it should work.
HD3200 integrated
Slackware 12.1

[cruiseoveride]

doesnt build on vanilla 2.6.27

complains about:

/tmp/ATI-fglrx-8.542-1-21319-root/tmp/ATI-fglrx-8.542-1.f9-root-root/lib/modules/fglrx/build_mod/2.6.x/firegl_public.c:4063: error: too many arguments to function 'smp_call_function'


It has not been updated for 2.6.27

[bulletxt]

why is it that every AMD/fglrx driver creates a thread with 200 pages posts? maybe because they are full of issues? WHY? always the same bulls**t.
a publicity on tv says: "change happens" . it seems AMD doesn't.

[Odur]

I wrote this in the thread about the 8.9 driver, and I got the same issue with the new 8.10 downloaded from AMD. Can someone else confirm, or has anybody got a clue what the h-ll is going on?


I wrote this about 20 minutes ago in bug report #112518 at Launchpad:

I too got this bug and haven't found a solution yet.

Dist: Kubuntu 8.04.1 Fully upgraded
Graphics card: PowerColor (ATI) HD3870 SCS3
Tested drivers: fglrx 8.3 through 8.9

Symptoms: When switching user the screen goes black and the system is totally unresponsive, including the keyboard. Not even caps-lock trigges the caps-lock light. The only thing that works is the reset button and to hold the powerbutton until it shuts down. Shutdown or restart works without problem (I have fixed the problem with atieventsd, see bug 118605).
If i use the vesa or radeonhd drivers everything is ok. But then I got no 3D acceleration, which I think would be nice...

This bug is really annoying to me, as we are switching users a lot on this computer. It started when I bought this card. I got a ATI X600 before, and that didn't suffer from this bug at all.

[xeros]

I wrote this in the thread about the 8.9 driver, and I got the same issue with the new 8.10 downloaded from AMD. Can someone else confirm, or has anybody got a clue what the h-ll is going on?

Have you modified /etc/ati/atieventsd.sh (or something like that, I'w writing from memory - don't have AMD/ATI card with me for now)?
In the fglrx drivers from AMD there's a bug with the wrong path for Xauthority file for KDM. If you change it to the real Xauthority file it should work without such problems. That helped me solve this problem sometime ago for Kubuntu.

[Odur]

Have you modified /etc/ati/atieventsd.sh (or something like that, I'w writing from memory - don't have AMD/ATI card with me for now)?
In the fglrx drivers from AMD there's a bug with the wrong path for Xauthority file for KDM. If you change it to the real Xauthority file it should work without such problems. That helped me solve this problem sometime ago for Kubuntu.
Yeah I've done that, and that's a different bug. With the wrong path to Xauthority the computer hangs on shutdown, but the keyboard is still responsive. The freeze when switching user is a whole different story.

[rektide]

Running 2.6.27 on a 7.1 xorg. This is bull@#$%.

I found the farbring's patch (http://www.phoronix.com/forums/showpost.php?p=46823&postcount=4) for fixing the fglrx module compilation error and applied it before the kernel compiled and still got the following:

[Message] Kernel Module : Trying to install a precompiled kernel module.
[Message] Kernel Module : Precompiled kernel module version mismatched.
[Message] Kernel Module : Found kernel module build environment, generating kernel module now.
ATI module generator V 2.0
==========================
initializing...
doing Makefile based build for kernel 2.6.x and higher
rm -rf *.c *.h *.o *.ko *.GCC* .??* *.symvers
make -C /lib/modules/2.6.27.1-deneb/build SUBDIRS=/lib/modules/fglrx/build_mod/2.6.x modules
make[1]: Entering directory `/usr/src/linux-2.6.27.1'
CC [M] /lib/modules/fglrx/build_mod/2.6.x/firegl_public.o
/lib/modules/fglrx/build_mod/2.6.x/firegl_public.c: In function ‘__ke_vm_phys_addr_str’:
/lib/modules/fglrx/build_mod/2.6.x/firegl_public.c:3528: warning: return makes pointer from integer without a cast
/lib/modules/fglrx/build_mod/2.6.x/firegl_public.c:3529: warning: return makes pointer from integer without a cast
/lib/modules/fglrx/build_mod/2.6.x/firegl_public.c:3530: warning: return makes pointer from integer without a cast
/lib/modules/fglrx/build_mod/2.6.x/firegl_public.c:3532: warning: return makes pointer from integer without a cast
/lib/modules/fglrx/build_mod/2.6.x/firegl_public.c: At top level:
/lib/modules/fglrx/build_mod/2.6.x/firegl_public.c:5780: warning: initialization from incompatible pointer type
/lib/modules/fglrx/build_mod/2.6.x/firegl_public.c:5806: warning: initialization from incompatible pointer type
CC [M] /lib/modules/fglrx/build_mod/2.6.x/kcl_io.o
CC [M] /lib/modules/fglrx/build_mod/2.6.x/kcl_pci.o
CC [M] /lib/modules/fglrx/build_mod/2.6.x/kcl_agp.o
LD [M] /lib/modules/fglrx/build_mod/2.6.x/fglrx.o
Building modules, stage 2.
MODPOST 1 modules
WARNING: modpost: module fglrx.ko uses symbol 'init_mm' marked UNUSED
CC /lib/modules/fglrx/build_mod/2.6.x/fglrx.mod.o
LD [M] /lib/modules/fglrx/build_mod/2.6.x/fglrx.ko
make[1]: Leaving directory `/usr/src/linux-2.6.27.1'
build succeeded with return value 0
duplicating results into driver repository...
done.
You must change your working directory to /lib/modules/fglrx
and then call ./make_install.sh in order to install the built module.
- recreating module dependency list
- trying a sample load of the kernel modules
failed.
[Error] Kernel Module : Failed to install compiled kernel module - please consult readme.

Obviously the patch was not so good.

DAAMIT why do you hate your customers so? Hopefully now that Ubuntu's done it, it will be easy for you to implement quickly.

[mirza]

Question: As opose to Intel, ATI has 2 drivers: Open and closed source (lets ignore fact there are several FOSS versions too). Obviously, closed version is way ahead, so I wonder why they maintain 2 versions instead of having one (open), just like Intel? I understand there are some IP issues, but those can be offset to small binary module. That would meke possible for community to replace closed module with open source version for people in countries where those SW-patent laws are not valid (EU) or for users that just don't care. I would like to have ATI card, but it looks that FOSS drivers will always play catch up game, performance-wise and feature-wise. Closed source driver is OK but this is what I have already with nVidia. So, currently best option I see is switching to Intel and hoping they will continue in performance improvements started with X4500HD.

[yoshi314]

I understand there are some IP issues, but those can be offset to small binary module. That would meke possible for community to replace closed module with open source version for people in countries where those SW-patent laws are not valid (EU)
- that would be no small module.
- i don't know of a country where drm solutions are not valid. what about HDCP, for example? is that illegal somewhere in the world?

[ptitgnu]

Hi,

for Gentoo users :

If you want to test the new ati-drivers 8.543 with xorg-server-1.5 , I made some ebuilds this night... There are on my gentoo overlay (ftp://ftp.ptitgnu.be/gentoo-overlay).

To add the xorg-server and ati-drivers overlay in one line, you can use :

wget -nH --cut-dirs=1 -r -l inf -P /usr/local/portage ftp://ftp.ptitgnu.be/gentoo-overlay/

It's based on the ubuntu 8.543 source package and it's working for me with a Radeon HD3300 :)

I have more FPS with glxgears too (~3170 with 8.543 vs ~3000 with 8.532)


Troubleshooting :

It seems that there is a problem with 'fglrx_dri.so' (don't forget, this is a beta version)..
This causes a problem with AIGLX :
(EE) AIGLX error: fglrx exports no extensions (/usr/lib64/dri/fglrx_dri.so: undefined symbol: __driDriverExtensions)
It seems to make 'amdcccle' segfault too..

Does anyone have an idea for that ?

[grantek]

so I wonder why they maintain 2 versions instead of having one (open), just like Intel?
The way I understand it, AMD/ATI are worried that their closed-source driver has code in it that reveals tricks about what the silicon is actually doing. They can release full documentation/APIs for the hardware, but knowing exactly how to squeeze the most performance out of a card isn't always obvious from a datasheet, and AMD think that revealing exactly how their driver works will help competitors learn how the hardware is put together.

From what bridgman's said, they do release example code along with documentation, so that example code is probably from fglrx. But I agree that the closed-source driver shouldn't be necessary in the long term :)

[SheeEttin]

Hello everyone, is there a way gentoo users can have that "hotfix" version of the 8.10 driver release ?, because i can not find it, can someone write the URL for the package of the "hotfix" for Ubuntu for 64 Bit ?, so i can extract the binary parts and place them by hand on my system, thanks by the way.

http://packages.ubuntu.com/search?keywords=fglrx&searchon=names&suite=intrepid&section=all
For dependencies, you're on your own.

[bash]

Just gave the Ubuntu version a test on video playback. Still is b0rked. Only output mode that has no tearing is still X11. GL/XV remains unusable.

[rbmorse]

I don't notice any tearing or other artifacts when using VLC in Xv on my HD3870. CPU is down around 10% - 15%

GL crashes when the source goes to wide-aspect ratio (i.e., plays the the FBI warning and the duplicator's logo but crashes when it tries to open the file with the main program).

I wonder what's really in repository as opposed to installed by Ubuntu's restricted device manager. From X.org.log:


(II) LoadModule: "fglrx"

(II) Loading /usr/lib/xorg/modules/drivers//fglrx_drv.so
(II) Module fglrx: vendor="FireGL - ATI Technologies Inc."
compiled for 1.4.99.906, module version = 8.54.3
Module class: X.Org Video Driver
(II) Primary Device is: PCI 01@00:00:0
(WW) Falling back to old probe method for fglrx
(II) ATI Proprietary Linux Driver Version Identifier:8.54.3
(II) ATI Proprietary Linux Driver Release Identifier: UNSUPPORTED-8.543.2
(II) ATI Proprietary Linux Driver Build Date: Oct 10 2008 21:37:39
(WW) This ATI Proprietary Linux Driver does not guarantee support of video driver ABI higher than 2.0
(WW) Video driver ABI version of the X server is 4.1

[Tares]

Well, I've installed the 8.10 from ubuntu repos, but there's something wrong with the performance here. On q3 i get 70fps on my HD4850 :x I've tried to install the driver manually, but I've failed ;-)

[cruiseoveride]

Built on Fc9/x86_64, using a little hot-wiring and the ubuntu orig file
But cannot load module.


fglrx: module license 'Proprietary. (C) 2002 - ATI Technologies, Starnberg, GERMANY' taints kernel.
[fglrx] Maximum main memory to use for locked dma buffers: 1886 MBytes.
[fglrx] vendor: 1002 device: 9440 count: 1
[fglrx] ioport: bar 4, base 0xa000, size: 0x100
ACPI: PCI Interrupt Link [APC3] enabled at IRQ 18
pci 0000:01:00.0: PCI INT A -> Link[APC3] -> GSI 18 (level, low) -> IRQ 18
pci 0000:01:00.0: setting latency timer to 64
[fglrx] PAT is enabled successfully!
[fglrx:firegl_init_module] *ERROR* firegl_stub_register failed

[atippett]

Driver seems to be posted now at http://ati.amd.com/support/drivers/linux/linux-radeon.html

[Fran]

Well, I've installed the 8.10 from ubuntu repos, but there's something wrong with the performance here. On q3 i get 70fps on my HD4850 :x I've tried to install the driver manually, but I've failed ;-)

Maybe something related to vsync? With my 3850 I get 140fps in timedemo demo1 at 1680x1050 with AA4x and AF8x (IIRC, amdcccle segfaults with 8.10 so I can't check :P).

[Tares]

Maybe something related to vsync? With my 3850 I get 140fps in timedemo demo1 at 1680x1050 with AA4x and AF8x (IIRC, amdcccle segfaults with 8.10 so I can't check :P).

Not really, cause I've checked that first. The other thing, quake wars are terrible :( in menu i get like 15-20 fps, feels really crappy. But when i want to connect to any server i get seg fault :( so something is really not right with the driver ;x

Atleast i can read temp ;-)

There is an update for the fglrx now, I hope it will fix my issues ;x

EDIT :

well, I've found why I get so poor performance :/ Control Center reports that my OpenGL is mesa, not the ATI. Also CCC version is 1.0, not the 2.1 or smth like that :(

Anyone got something similar ? or Ideas how to fix it ?

[Xtigyro]

Flickering, flickering, flickering... Mobility X1600 256 MB DDR3 !!!!!!!!!!!!

AMD, I want to be absolutely clear, THIS IS THE LAST CARD I SPENT MONEY FOR till you make some REAL progress with your your BUGGY, LAGGY, FLICKERING drivers (both OPENSOURCE an' CLOSESOURCE). :)))))))))))))))))))

:)

[Tares]

My CCC looks like this :
http://img220.imageshack.us/img220/4561/cccgc5.th.png (http://img220.imageshack.us/my.php?image=cccgc5.png)http://img220.imageshack.us/images/thpix.gif (http://g.imageshack.us/thpix.php)
its in polish, but I've marked the mesa thingy.

Any ideas how to fix it ?

[Kano]

Maybe you did install xserver-xgl? Remove that or the fglrx mode way not loaded.

[Tares]

Maybe you did install xserver-xgl? Remove that or the fglrx mode way not loaded.

I didn't install that, but to make sure I've checked synaptic and my guess was right. Its not installed.

My system is Ubuntu 8.10 beta + all updates there were released.

EDIT :

The funny thing is, that fglrxinfo returns :

display: :0.0 screen: 0
OpenGL vendor string: ATI Technologies Inc.
OpenGL renderer string: ATI Radeon HD 4800 Series
OpenGL version string: 1.4 (2.1.8087 Release)

[nphoronix]

Tearing/Flickering still there... Are they even fixing anything...?

[Kano]

Not much, maybe it is a part time job ;)

[bububabu]

My CCC looks like this :
http://img220.imageshack.us/img220/4561/cccgc5.th.png (http://img220.imageshack.us/my.php?image=cccgc5.png)http://img220.imageshack.us/images/thpix.gif (http://g.imageshack.us/thpix.php)
its in polish, but I've marked the mesa thingy.

Any ideas how to fix it ?

i get something like this too, plus my card is 256megs afaik but only 128 are showed, see screenshot
http://img119.imageshack.us/img119/1679/immaginejq5.th.jpg (http://img119.imageshack.us/my.php?image=immaginejq5.jpg)http://img119.imageshack.us/images/thpix.gif (http://g.imageshack.us/thpix.php)

[Kano]

lsmod|grep -e fglrx -e radeon -e drm

check this, only fglrx should be seen loaded, if not reboot. I tested my script with a modified ati installer on intrepid and had the correct output.

[energyman]

maybe some symlinks are messed up? check the stuff in /usr/lib64 and /usr/lib32

[Tares]

lsmod|grep -e fglrx -e radeon -e drm

check this, only fglrx should be seen loaded, if not reboot. I tested my script with a modified ati installer on intrepid and had the correct output.

fglrx 1813960 28
agpgart 42184 2 fglrx,intel_agp

Dunno what I have to check in /usr/lib/

I'll try to install the driver manually and use your script Kano ;-)

[rbmorse]

That's what I get, too