ATI Catalyst 8.10 Driver

[tramal]

Grab here:

https://a248.e.akamai.net/f/674/9206/0/www2.ati.com/drivers/linux/64bit/ati-driver-installer-8-10-x86.x86_64.run (https://a248.e.akamai.net/f/674/9206/0/www2.ati.com/drivers/linux/64bit/ati-driver-installer-8-10-x86.x86_64.run)


The release notes are still empty as of now:

https://a248.e.akamai.net/f/674/9206/0/www2.ati.com/drivers/linux/catalyst_810_linux.html (https://a248.e.akamai.net/f/674/9206/0/www2.ati.com/drivers/linux/catalyst_810_linux.html)


Cheers,

tramal

[Kano]

That driver is pure crap for serveral reasons

+ xorg 7.1 dri permissions are fixed (8-9 had wrong permissions)

- wine apps crash (for example trackmania)
- 2.6.27 kernel needs same patch as 8-9
- no x740 dirs like the driver for ubuntu 8.10
- as usual no pointsprites with gl2benchmark
- beryl not working with etch

so basically we are at the state of 8-8 again with a little bit different 2.6.27 patch.

[Fran]

- no support for xorg-server-1.5 yet.

[RealNC]

I'm starting to HATE Ubuntu for the same reasons I hate Windows.

Seriously.

[doubledr]

That driver is pure crap for serveral reasons

+ xorg 7.1 dri permissions are fixed (8-9 had wrong permissions)

- wine apps crash (for example trackmania)
- 2.6.27 kernel needs same patch as 8-9
- no x740 dirs like the driver for ubuntu 8.10
- as usual no pointsprites with gl2benchmark
- beryl not working with etch

so basically we are at the state of 8-8 again with a little bit different 2.6.27 patch.

I will be very disappointed if this is true.
:(:eek:

[porksome]

Tried the 8.10 driver on the ATI website.

ATIConfig results in a Segmentation Fault when I run it. Have tried manually adding the driver to xorg.conf and doing the "aticonfig -tls=1" trick but when I reboot I end up going back to the command line as Xserver says it cannot find any screens. Using the open-source drivers again now until some more people have tried 8.10 incase I had installed it wrong.

One thing I did notice is that an extra "modaliases" .deb package is generated when I build the packages.

Running ubuntu hardy, kde 4.1 and Mobility Radeon M10 (9600)

[Thetargos]

Oops there go my hopes for Xorg 7.4 support :rolleyes:

I will have to roll back to F8, or switch distributions entirely... not a good thing either way. F9 (Xorg 7.4) is unstable as the result of XServer 1.5 - radeon(hd) drivers interaction crashing X (when using EXA the crash is very well defined to pan images larger than the viewing area, without it, the crashes are more "random"), though I seem to be of the few (very few it would seem) experimenting this issue with both drivers (radon and radeonhd), so no fgrlx for me, means I most likely will roll back. :-/

[grantek]

So is Ubuntu using a different binary driver or something? You'd think if it was a patch to the distributed-as-source bits, that Ubuntu could/would/should divulge what they did while playing with the prerelease driver.

[Thetargos]

So is Ubuntu using a different binary driver or something? You'd think if it was a patch to the distributed-as-source bits, that Ubuntu could/would/should divulge what they did while playing with the prerelease driver.

Ubuntu? Get serious! They do not contribute much back upstream (until fairly recently, that is)

[Kano]

For Xserver 1.5 packages on other systems then Ubuntu you can try this hack:

Extract the ati-installer 8-10 and combine it with

http://archive.ubuntu.com/ubuntu/pool/multiverse/f/fglrx-installer/fglrx-installer_8.543.orig.tar.gz

this way:

a) remove all old arch common x* dirs completely
b) create a new common dir, put etc lib opt usr from 8.543 in there the rest just copy to extracted 8-10 dir
c) rename x740 to x710 and x740_64a to x710_64a
d) now you can execute

./ati-installer.sh 8.542 --install

or whatever option you usually use.

Have fun ;)

[nnotimportant]

Another thing you can do is to just install the 8.10 version from amd's website and then copy the files in the x740 directory into the appropriate locations.

[Moronix]

I'm having better luck with this release. My plan was to stick with a more proven kernel, usually I went with whatever is latest. I stuck with a more proven kernel 2.6.26.x and I am having better luck with fglrx than I have before. I don't know if this release is actually better or if it's just because I'm sticking with the old kernel. More proven old version of Fedora as well, 8, rolled a kernel with the deprecated calls stuff and installed this driver manually (GPL hacks needed). No segfaults so far, of course I've probably jinxed it by saying that. Hope not. My problem wasn't performance for quite a few releases now, it was that segfaults made it useless, brought the whole system down at times. My advice is to forget about bleeding edge kernels and Xorg if you're going to try using this driver, maybe that is the key.

[Fran]

a) remove all old arch common x* dirs completely
b) create a new common dir, put etc lib opt usr from 8.543 in there the rest just copy to extracted 8-10 dir
c) rename x740 to x710 and x740_64a to x710_64a
d) now you can execute

./ati-installer.sh 8.542 --install

Thanks, following the first 2 steps I created an archive ati-driver-installer-8-10b-x86.x86_64.tar.gz that can be installed with an ebuild. It compiles fine with 2.6.27 kernels without patching and works ok with xorg-server-1.5. The only problem I have is that now amdcccle segfaults.

The diff for the ebuild:
--- /usr/portage/x11-drivers/ati-drivers/ati-drivers-8.532.ebuild 2008-09-30 01:24:49.000000000 +0200
+++ ati-drivers-8.543.ebuild 2008-10-16 21:23:02.000000000 +0200
@@ -9,7 +9,7 @@
DESCRIPTION="Ati precompiled drivers for recent chipsets"
HOMEPAGE="http://www.ati.com"
ATI_URL="https://a248.e.akamai.net/f/674/9206/0/www2.ati.com/drivers/linux/"
-SRC_URI="${ATI_URL}/ati-driver-installer-8-9-x86.x86_64.run"
+SRC_URI="${ATI_URL}/ati-driver-installer-8-10b-x86.x86_64.tar.gz"

LICENSE="AMD GPL-2 QPL-1.0 as-is"
KEYWORDS="~amd64 ~x86"
@@ -114,7 +114,7 @@
fi

# Only support xorg-server >=1.1
- BASE_DIR="${S}/x710"
+ BASE_DIR="${S}/x740"

# This is used like $(get_libdir) for paths in ati's package.
if use amd64 ; then
@@ -130,8 +130,7 @@
src_unpack() {
#Switching to a standard way to extract the files since otherwise no signature file
#would be created
- local src="${DISTDIR}/${A}"
- sh "${src}" --extract "${S}" 2&>1 /dev/null
+ unpack "${A}"

# These are the userspace utilities that we also have source for.
# We rebuild these later.
@@ -160,7 +159,7 @@
|| die "Replacing 'finger' with 'who' failed."
# Adjust paths in the script from /usr/X11R6/bin/ to /opt/bin/ and
# add funktion to detect default state.
- epatch "${FILESDIR}"/8.532/ati-powermode-opt-path-2.patch
+ epatch "${FILESDIR}"/8.543/ati-powermode-opt-path-2.patch
fi

pushd ${MODULE_DIR} >/dev/null

[grantek]

For Xserver 1.5 packages on other systems then Ubuntu you can try this hack
Thought so - it's a shame AMD couldn't release it this way, but hopefully all the other distros that package binary drivers can set it up for 1.5 as well. I hope the reason AMD didn't flick the "enable" switch isn't too serious though...

[Inkaine]

Did anyone get the 8.10 release working with kernel 2.6.27 yet? I tried both the old, working 8.9 patch with addapted lines as well as a diff of the ubuntu firegl_public.c to the original one. Both with same result sadly - a black screen after running X. Ctrl+Alt+Backspace doesn't quit the X server, Ctrl+Alt+Delete results in a clean shutdown though. The X logfile didn't ring anything for me. You can check the Xorg.0.log (http://www.manuelgaul.eu/archlinux/Xorg.0.log) here. Anyone been successful yet or got a clue?

[djdoo]

Hey this is the first time something like that happens!!
It reminds me something very older when a driver from the old codebase used to display a watermark with AMD's logo at the lower right corner of the screen(I think so..) and AMD took the driver back after 1-2 days and gave out a new with the problem fixed! But this was not that serious!!

Guys from AMD please do the same for this thing cause 8.10 driver from the website doesn't offer us anything new actually except for me that I don't have to disable my onboard HD 3200 chip everytime I boot into openSUSE cause screen used to remain black with 8.9 but still my on board 3200 is not a candidate for a Crossfire chain so still not usable!
-XvMC enabled but not usable...
-I still use 7.3 Xorg (xserver 1.4.0) although all the other Xorg packages are of 7.4 version from Xorg:X11 repo of suse 11(except for driver-input package cause otherwise mouse and keyboard don't work)
-I just noticed that Anisotropic Filtering at AMDCCCLE reached x16 level but without HD 3200's support my HD 3450 cannot handle these levels easily...

So nothing new that it is vitaly usable!

You must replace it with Canonical's driver version even if it is in beta stage!!

[djdoo]

Well I tried Xine with XvMC (xxmc xine driver) and it is showing a correct menu at the settings after loading and also it is playing video!! DVD checked.
In the past, I mean 8.9 and earlier, xxmc Xine driver could not be load at all...

Seems that Mplayer is not usable! (Which I don't care too much about)

Check Xine guys it might be a good surprise!

[bulletxt]

I just love these threads! they just remind me how shitty AMD and their drivers are! I LOVE YOU AMD I NEVER GET BORED WITH YOU.

[djdoo]

It was to good to be true sadly...:(
Video plays and after going down all DVD menus the playback starts and stops after a few moments freezing xine.

But try it! It might work better for you...

[Pulfer]

Check Xine guys it might be a good surprise!

I checked Kaffeine with xine engine and it still freezes the system with XV output. It can play video for several seconds but hangs anyway. And the picture gets screwed up when I stop or pause playback. With xshm it works fine.

[Thetargos]

How is performance with Xorg 1.5? Especially EXA accel and compositing?

[hansch]

I have installed the 8.10 driver on MythBuntu 8.04 with kernel 2.6.24.19-generic. It builds and installs fine, but starting the computer with the new driver results in a black screen. This after some resolution changes in the monitor (Samsung LCD tv through HDMI) It settles on 640x480 through VESA, but the screen remains black. Before with 8.9 it correctly set output to 1920x1080 through EDID. The output had to be corrected for overscan, but it did work.

8.10 is much worse for me then 8.9, so I just reverted bsck. I have workarounds for almost all issues, except for the displaying of TV through mythtv. This gives garbled output when changing channels.

8.10 is disappointing sofar.

Hans

[Petteri]

I think that the only reason there is 8.10 driver is the 'one driver every month'-policy. This 8.10 driver seems to be just 8.9 with some quick fixes to justify new driver and amd has really been working on something like 8.11 with support for new xorg and kernel. If some driver version has planned features that can't be done in less than month, driver team just puts out something like 8.10 to keep amd's driver policy going while they are all the time (for longer period of time) developing the "real" driver update.

This is how I see things. :D

[mirv]

I think that the only reason there is 8.10 driver is the 'one driver every month'-policy. This 8.10 driver seems to be just 8.9 with some quick fixes to justify new driver and amd has really been working on something like 8.11 with support for new xorg and kernel. If some driver version has planned features that can't be done in less than month, driver team just puts out something like 8.10 to keep amd's driver policy going while they are all the time (for longer period of time) developing the "real" driver update.

This is how I see things. :D

If I remember an article about how the amd driver development occurs, basically yes. An "old" code path is kept before any major changes, and is used until the changes are ready for release. There's an article somewhere on phoronix which explains how it all goes.

Here it is:

http://www.phoronix.com/scan.php?page=article&item=735&num=1

[madman2k]

I'm starting to HATE Ubuntu for the same reasons I hate Windows.

Seriously.
good third party software support and beginner friendlieness?

[porksome]

Anyone else have a problem with aticonfig segfaulting. Every time i try "sudo aticonfig --initial" I get a segfault. Never had this before.

Even tried removing everything fglrx after uninstalling previous driver using "find" command

If I reboot Xserver gives me a VESA driver. 8.9 worked fine apart from flickering 3d when using compositing (to be fair this is unlikely to be fixed until DRI2 as ATI APPEAR (?) to follow the protocols and not write their own versions of xserver parts- correct me if I'm wrong)

Running Ubuntu Hardy 8-04. Latest updates. Radeon Mobility 9600 (M10)

[Kano]

You can use my script - it works with xserver 1.4.2 directly. No aticonfig tool needed, it can directly use the xorg.conf created by

dpkg-reconfigure xserver-xorg

(you can even remove all xorg.conf* files and run that the reconfigure command then execute the script)

run in vt1 (ctrl-alt-f1):

http://kanotix.com/files/install-fglrx-debian.sh

[djdoo]

I said Xine(Xine-UI)! Not Kaffeine! Kaffeine could not load xxmc driver even with nvidia cards that could play XvMC video in the past here!

And NOT XV but XXMC.

[Nille]

I have now Test the New Driver with Ubuntu Hardy and Intrepid and on Both Systems the Driver runs fine for me.

XvMC has no Problem ( now i can Play 1080p *.* and 720 with CnQ on the lowest Multi (200x5) )

[Methylene]

Um I just want to say... I did have a few hitches in installation, but I think they were very specific to my circumstances... However, if you like compiz, this driver in 8.10 ubuntu is the way to go... I mean I have compiz settings really turned up, and it works, completely, and utterly, perfectly. My framerate never drops below my refreshrate (60hz).

Mipmaps work and everything... WTF JUST HAPPENED. I like it... even if it was a little unstable at first.

[djdoo]

Hey guys one question...
You the ones with ubuntu do you use the driver from Canonical??

[Methylene]

I had to, generating packages from the ati installer was 100% fail and I never trust the installer's automatic methodology. I have had a few problems now with frequent crashes... Sucks cuz compiz really does work way better, and OpenGL is just plain faster.

I think this may be a setup related issue though, I did an install with a home directory from 8.04.1 32bit, to 8.10b amd64 and there was a lot of things wrong from just configuration. I'm trying to sort it all out and remove any configurations that may be problematic.

Oh, BTW I had 0 problems whatsoever just upgrading from the 8.9 driver to 8.10 before I upgraded ubuntu from 8.04.1.

[mirak63]

thanks ati, now I have a direct crash at X server start
thanks for make me going back to windows
it's been 4 years I use Linux only, and since I have an this hd4850 I am forced to use windows

[highlandsun]

Hey guys one question...
You the ones with ubuntu do you use the driver from Canonical??

Yes. It mostly works for me, but when I log out from X and try to get back to a text console (I don't use gdm) the X server hangs and the console is stuck with multicolored text-mode hash all over it.

I can ssh in from another machine and see that the X server is running full bore, 100% CPU and uninterruptable. kill -9 doesn't affect it, and ptrace can't attach to it so gdb/gcore don't work either. Quite annoying; it also prevents a clean shutdown so I wind up having to do a hard reset.

I tried xine xxmc and when it didn't just hang on me, it would display the top left corner of a video, stretched out about 8x normal size. The rest of the video frame would be lost/truncated. totem-xine would hang a lot too, but sometimes it would work, and then it appeared to be pretty decent. Playing a 1280x720 x264 movie, I'd see CPU usage around 65% and the processor stayed at it's lowest clock speed most of the time (500MHz.) There'd be occasional jumps to 1.1GHz and 2.1GHz, but most of the time it stayed at 500Mhz and temperatures stayed pretty cool. With vlc, the same video would peg a CPU core at 90-100% and the speed would stay at 2.1GHz.

[yorik]

I saw a couple of changes in blender: the good point: opengl dotted lines are back, and rendered at decent speed. the bad point: opengl texts, that were rendering very slowly but at least they were rendering correctly, now present quite a lot of artifacts (disappearing, changing size, etc).

[dscharrer]

The Release Notes (https://a248.e.akamai.net/f/674/9206/0/www2.ati.com/drivers/linux/catalyst_810_linux.html) for 8.10 are online now. Not that there's anything interesting about them.

I don't think "Quickly moving a X video window no longer results in occasional corruption being left behind on the desktop on Radeon X1000 series hardware or later." belongs under Known Issues though. Unless AMD now considers fixed bus an issue? Which would actually explain a lot. ;)

Anyway, why is it that only such old versions of glibc as 2.2 and 2.3 are supported? The oldest glibc in Gentoo Portage that is not masked by the default 2008.0 amd64 profile is 2.4 while the newest 'stable' one is 2.6.1, which seems to be what Ubuntu is using (http://packages.ubuntu.com/source/gutsy/glibc).

[Heiko]

For me the 8.10 (downloaded from AMD website) is one of the worst drivers in a while. I tried to install it on my AMD780g integrated graphics with an AMD3450 graphics card. But the only thing I got was a blank screen, no possible way to start X and I also couldn't get into a command line using Ctrl-Alt-Fn.

So I went back to the 8.9 driver, which also has some issues for me, but at least it gives me an image on my screen.

Small typo, I meant that I could not get into a command line. My problem is the as the one Mirak63 has, it crashes the computer completely.

[mirak63]

For me the 8.10 (downloaded from AMD website) is one of the worst drivers in a while. I tried to install it on my AMD780g integrated graphics with an AMD3450 graphics card. But the only thing I got was a blank screen, no possible way to start X and I also could get into a command line using Ctrl-Alt-Fn.

So I went back to the 8.9 driver, which also has some issues for me, but at least it gives me an image on my screen.


for me it totally cash the computer at X start
so I can't even know what's wrong or debug at all

However I don't understand how can Linux kernel let a video driver crash everything.

So here I can't not blame Linux as well as ATI.

[Kano]

The fglrx KERNEL module can crash...

[mirak63]

The fglrx KERNEL module can crash...

bringing the operating system down with him
that's not normal
it's a conception flaw
of both

[Kano]

Well you can try:

http://en.wikipedia.org/wiki/Magic_SysRq_key

in that case.

[energyman]

oh realy? a conception flaw?
read up on modules. They can do all and everything. After that you might be able to help people to make it better.

[bugmenot]

The 8.10 version still crashes Softimage XSI. Even selecting points can lead to an error:
http://vimeo.com/2034870
http://vimeo.com/2034885

I hope they'll fix that! Its a pain to work like this. Very old driver worked fine but they won't support a newer kernel so I can't get back to them without changing kernels!

[Methylene]

Anyone who is still having problems with a crash with the new xorg, insure backing store is disabled, it seems to be broken in the new xorg version. (at least I'm quite sure it worked fine with catalyst 8.10 and the 7.3 xorg /1.4 xserver.)

[mirak63]

oh realy? a conception flaw?
read up on modules. They can do all and everything. After that you might be able to help people to make it better.

oh really ?
anyone have a possibility to improve fglrx here ?

let me know when this happens lol

[energyman]

see, the problem is that a module can do whatever it wants. If you want to be save from module misdeeds, either don't use them more contribute patches for a module infrastructure where modules can't do anything wrong.

[Vighy]

uh! strange bug here!

it's related to randr 1.2

if after launching xorg i run xrandr, i can see some less resolutions than with previous drivers but everything is quite fine:

ben@Santa^s_lil^_helper ~ $ xrandr
Screen 0: minimum 320 x 200, current 1280 x 800, maximum 1280 x 1280
LCD connected 1280x800+0+0 (normal left inverted right x axis y axis) 331mm x 207mm
1280x800 60.0*+
1024x768 60.0
720x480 60.0
512x384 60.0
400x300 60.0
320x240 60.0
DFP1 disconnected (normal left inverted right x axis y axis)
CRT1 disconnected (normal left inverted right x axis y axis)
TV disconnected (normal left inverted right x axis y axis)


re running it, it gives the same output.

if I then run amdcccle, it shows me every resolution i was used to be able to select....
but now if I run xrandr again.....

ben@Santa^s_lil^_helper ~ $ xrandr
Screen 0: minimum 320 x 200, current 1280 x 800, maximum 1280 x 1280
LCD connected 1280x800+0+0 (normal left inverted right x axis y axis) 331mm x 207mm
320x240 60.0
320x200 60.0
DFP1 disconnected (normal left inverted right x axis y axis)
CRT1 disconnected (normal left inverted right x axis y axis)
TV disconnected (normal left inverted right x axis y axis)
1280x800 (0x8d) 71.1MHz
h: width 1280 start 1328 end 1360 total 1440 skew 0 clock 49.4KHz
v: height 800 start 803 end 809 total 823 clock 60.0Hz


it shows this mess.... and now amdcccle doesn't show right resolutions anymore. (just leaves the 320x240 choice).

so i think it's an amdcccle related issue, since xrandr didn't bring that after I launched it.

sorry if I reported it just now, but I didn't play much with the new driver this month.
Bridgman, do you think there will be time to fix this bug in the next release? :D

[nstamoul0]

I apologize in advance if this has already been answered.I have been following the progress of the ATI fglrx driver for some time now,upgrading every month to the newest driver (with the exception of 8.10).

I am running Debian (Sid) and I am getting pretty tired of the well known bug where you can't load the fglrx driver on a vanilla debian kernel due to the lack of the export flush_tlb line in the kernel source.I have to recompile every time there is a newer kernel source and I am sick of it.

My question is if there is a workaround on this by patching the driver and not the kernel.