Discussion: the ati/radeon open source drivers & radeon repository

handy · 2011-12-04 09:15:25

I'm pulling just shy of 2k fps out of glxgears on my RV630 on the current stable kernel in combination with the git packages from Perry3D's repo.

20k · 2011-12-05 00:16:12

Hey guys,

I've been rummaging through this thread in the hopes of finding a solution to my scolding hot HD5470.
I'm currently running the xf86-video-ati driver. I've got KMS enabled and I have forced the GPU into a low power profile.

# cat /sys/kernel/debug/dri/0/radeon_pm_info
default engine clock: 750000 kHz
current engine clock: 156990 kHz
default memory clock: 800000 kHz
current memory clock: 199960 kHz
voltage: 900 mV
PCIE lanes: 16

# sensors
radeon-pci-0200
Adapter: PCI adapter
temp1: +96.5°C

acpitz-virtual-0
Adapter: Virtual device
temp1: +26.8°C (crit = +100.0°C)
temp2: +0.0°C (crit = +100.0°C)

As you can see the GPU is succesfully undervolted and underclocked. Performance is not really that much of an issue for me as I use nearly exclusively 2D acceleration.

However. Temperatures of +96°C are just unacceptable. These temperatures are probably correct as the laptop surface is completely cool except for the part just above the GPU which is scolding hot and can not be touched for prolonged periods of time.

From this thread I understand there is a third alternative next to the classic open source driver and closed source proprietary driver (which I tried already: the performance is just horrid).

So if someone were to take the time to type out a comprehensive reply on how I get these drivers to work. That would be lovely. I can't get an overview in a thread that is nearly 100 pages long anymore.

handy · 2011-12-05 01:14:23

I don't know how old or what the dust level that your Notebook is used in, but it may be well worth your time to remove the keyboard & blow out all of the dust inside with compressed air. If you haven't got access to a compressor you can buy a can of compressed air to do the job.

If you do do that, make sure you don't let any fans be propelled by the air as they can be damaged by over revving.

Another thing, is you can buy high quality thermal paste, which you could use after you remove the heat-sink from the GPU, clean them both of all traces of the previous thermal paste & then use a thin even coating of the new paste.

Some people have incredibly good results from doing that.

As far as the drivers are concerned someone else will have to help you there as I've had no problems with my 2 Radeon cards - HD2600 Pro & a 6450 in that regard, though the 6450 doesn't support KMS, so I don't get to use all of the capabilities of the xf86-video-ati driver, but that's another story...

Good luck with it, I know you can spend a lot of time on problems like this.

20k · 2011-12-05 14:49:00

handy wrote:

I don't know how old or what the dust level that your Notebook is used in, but it may be well worth your time to remove the keyboard & blow out all of the dust inside with compressed air. If you haven't got access to a compressor you can buy a can of compressed air to do the job.

Thanks for your reply Handy.

It is most definitely a driver issue. The machine is running a dual boot setup with Windows 7. Under Windows the GPU stays at a consistent 50°C - 55°C. It will ramp up to 70°C under load which triggers the fans to go into 100% rotations. But as soon as the load is taking away the GPU will cool back to 55°C and the fans will spin down.

Under Arch the GPU goes up to 80°C within the first few minutes after a cold boot. Even with no load on the machine and a low clock profile.

edit: My plan right now is to install Ubuntu out of the box. It's my understanding that they provide proprietary drivers as part of their core package. I'll get Ubuntu running and see how the GPU behaves. I'm hoping I can close in on the source of the problem in this way. I'll report back here later with my test results.

edit2: I seem to have found a solution to the HD5470M problem. I'm going to move my post to the larger AMD Bar & Grill topic.

Last edited by 20k (2011-12-06 22:06:11)

simo3o · 2011-12-07 21:35:02

Hi I have a RADEON HD 6630 m card with an integrated card and if I switch from the Intel card to the RADEON with vgaswitcheroo, and restart the Xorg my computers gets frozen, and if I do it without X it says "client 0 refused to switch", any idea, please? I don't know what to do.
Thank you in advance.

Last edited by simo3o (2011-12-07 21:35:49)

agapito · 2011-12-07 23:33:46

Anyone has a 6790 card? I think this card not work in linux with open source drivers. I have continuous gpu lockups.

handy · 2011-12-08 13:47:55

agapito wrote:

Anyone has a 6790 card? I think this card not work in linux with open source drivers. I have continuous gpu lockups.

The following may or may not be of any use to you:-

Install hwinfo:

sudo pacman -S hwinfo

Then run the following command:

sudo hwinfo --framebuffer

The output will at least tell you if the resolution that you are trying to run on the open-source radeon driver is supported by your GPU.

This command gave me the information that I needed to know that my 6450 GPU does not support resolutions higher than 1600 x 1200. It also does not support KMS; therefore the radeon drivers won't work properly with it so I have to use vesa & 1600 x 1200 !!

Vesa sure ain't what it used to be... [edit:] Which is certainly a good thing.

Last edited by handy (2011-12-08 23:02:14)

dw · 2011-12-09 08:04:34

20k wrote:

Hey guys,
I've been rummaging through this thread in the hopes of finding a solution to my scolding hot HD5470.
I'm currently running the xf86-video-ati driver. I've got KMS enabled and I have forced the GPU into a low power profile.

Hi 20k,

my girlfriend's notebook (samsung e272) does have this card too and i am experiencing exactly the same problems.

however, whenever i try to put the card to respect powersaving (as suggested in the wiki) using either

echo dynpm > /sys/class/drm/card0/device/power_method

or

echo profile > /sys/class/drm/card0/device/power_method
echo XY> /sys/class/drm/card0/device/power_profile

i instantly get screen artifacts when using gnome-shell in a way that i cannot use gnome (or any tty) at all. i have to ssh into the laptop and reboot it. can you reproduce this as well?

Cdh · 2011-12-10 07:52:00

Anybody wondering why lib32-mesa-... is not compiling but 64 bit mesa is?

chris@chrisl ~ % pacman -Si llvm | grep Version
Version              : 3.0-1
chris@chrisl ~ % pacman -Si lib32-llvm | grep Version
Version              : 2.9-2
chris@chrisl ~ %

Has been flagged out of date for a week now, nothing has happened so far.

Last edited by Cdh (2011-12-10 07:52:17)

Ejik · 2011-12-12 05:04:19

Hi.
Is there any way to include libglut.so.3 in mesa-full?
Torcs wont run due to libglut.so.3 => not found
freeglut and mesa-full(currently i use) are in conflict.
Thx 4help.

Cdh · 2011-12-12 10:23:56

Is probably by mistake in the mesa-full PKGBUILD.
Try pacman -Sdd freeglut (and remember that you have done it ).

Ejik · 2011-12-12 17:43:21

Cdh wrote:

Is probably by mistake in the mesa-full PKGBUILD.
Try pacman -Sdd freeglut (and remember that you have done it ).

thx for help.
pacman -Sdd freeglut
:: freeglut и mesa-full конфликтуют (glut). Удалить mesa-full? [y/N] (still in conflict)

Cdh · 2011-12-12 22:33:23

I think according to the manpage pacman is supposed to ignore dependencies in such a case... Well, it doesn't.

Easy way: Use mesa-r600g-git from AUR: https://aur.archlinux.org/packages.php?ID=47817 which doesn't have that wrong dependency (I assume you use the repository from the first post here?).
Quick-and-dirty way: Download the package from the repository (pacman -Sw mesa-full... ), unpack it with any archive program, remove freeglut from conflicts, replaces etc from the .PKGINFO in the root of the package, repack it to tar, tar.gz, etc. and install it then.
Dirty way: Edit /var/lib/pacman/local/mesa-full-whatever-you-have-installed/desc directly to remove the wrong dependency...

Ejik · 2011-12-13 05:52:17

Thx 4 help, Cdh.
I fix it such way:
Download and untar freeglut. Then copy lib files to /usr/lib.
Now torcs run!!!

lordheavy · 2011-12-18 20:19:53

Cdh wrote:

Anybody wondering why lib32-mesa-... is not compiling but 64 bit mesa is?
chris@chrisl ~ % pacman -Si llvm | grep Version
Version              : 3.0-1
chris@chrisl ~ % pacman -Si lib32-llvm | grep Version
Version              : 2.9-2
chris@chrisl ~ %   
Has been flagged out of date for a week now, nothing has happened so far.

lib32-llvm hasn't been updated to 3.0 because current lib32-mesa doesn't built without patching.

You can try to build mesa in extra with llvm-3.0, it will fail :-)

lordheavy · 2011-12-19 00:11:42

I've just created a repo for mesa from git and following schema from extra/testing.
These packages are built inside a clean testing chroot, so you need to have testing repo enabled.

[mesa-git]
Server = http://pkgbuild.com/~lcarlier/$repo/$arch

then pacman -Syu mesa-git yourvideodriver (intel-dri-git or ati-dri-git or nouveau-dri-git)

No yet lib32 but it should follow soon

++

Cdh · 2011-12-19 13:15:21

Yay, lib32-llvm has been updated and lib32-mesa is compiling again.
BUT: I have tried three games with wine and none of them start: Skyrim, Trackmania Nations and warcraft 3.

Warcraft 3 shows "Couldn't open Game.DLL" so I was curious what that means and started with WINEDEBUG=+all. Naturally it was very slow, BUT: it started with no problems with debug messages enabled, but not with debug messages disabled? What is this sorcery?

pkgname=lib32-mesa-full-r600g
pkgver=20111219
_realver=7.12
pkgrel=1
pkgdesc="Full Mesa 3D graphics library with all its components, built from the git master branch (mesa 7.12). Compiles mesa for r600g (gallium)."
arch=(x86_64)
url="http://mesa3d.org/"
license=('LGPL')
depends=('lib32-libdrm-git' 'dri2proto>=2.1' 'glproto>=1.4.10' 'libxxf86vm' 'libxdamage' 'expat>=2.0.1' 'libxmu' 'talloc' 'lib32-llvm')
makedepends=('pkgconfig' 'imake' 'xorg-server-devel')
provides=("lib32-libgl=${_realver}" "lib32-mesa=${_realver}" "lib32-ati-dri=${_realver}" 'lib32-libglapi' 'lib32-libegl' 'lib32-libgles')
replaces=('lib32-libgl' 'lib32-mesa' 'lib32-ati-dri' 'lib32-libglapi' 'lib32-libegl' 'lib32-libgles')
conflicts=('lib32-libgl' 'lib32-mesa' 'lib32-ati-dri' 'lib32-libglapi' 'lib32-libegl' 'lib32-libgles')

_gitroot="git://anongit.freedesktop.org/git/mesa/mesa"
_gitname="mesa"

build() {
   msg "Connecting to git.freedesktop.org GIT server...."

#   if [ -d $startdir/src/$_gitname ] ; then
#      cd $_gitname && git pull origin
#      msg "The local files are updated."
#   else
#      git clone $_gitroot
#   fi

   rm -rf $startdir/src/$_gitname-build
   cp -rH $startdir/src/$_gitname $startdir/src/$_gitname-build
   cd ${srcdir}/${_gitname}-build

  export CC="gcc -m32"
  export CXX="g++ -m32"
  export PKG_CONFIG_PATH="/usr/lib32/pkgconfig"
  # for our llvm-config for 32 bit
  export LLVM_CONFIG=/usr/lib32/llvm/llvm-config

    ./autogen.sh --prefix=/usr \
    --with-dri-driverdir=/usr/lib32/xorg/modules/dri \
    --with-gallium-drivers=r600 \
    --with-dri-drivers= \
   --enable-texture-float \
   --enable-gles1 \
   --enable-gles2 \
   --enable-osmesa \
   --enable-egl \
   --enable-xorg \
   --enable-xa \
   --enable-xvmc \
   --enable-vdpau \
   --enable-gallium-g3dvl \
   --enable-xcb \
   --enable-glx-tls \
   --enable-glu \
   --enable-glw \
   --enable-motif \
   --enable-gbm \
   --enable-gallium-gbm \
   --enable-shared-glapi \
   --enable-xorg \
   --enable-gallium-llvm \
   --enable-openvg \
   --enable-gallium-egl \
   --enable-shared-dricore \
   --enable-32-bit \
   --libdir=/usr/lib32 || return 1

   make || return 1
   make DESTDIR="${pkgdir}" install || return 1

   cd ${srcdir}/${_gitname}-build
   install -m755 -d "${pkgdir}/usr/lib32/xorg/modules/extensions"
  ln -s libglx.xorg "${pkgdir}/usr/lib32/xorg/modules/extensions/libglx.so"
  rm -rf "${pkgdir}"/usr/{include,share,bin}
   install -d ${pkgdir}/etc/profile.d
   echo '#!/bin/bash' > ${pkgdir}/etc/profile.d/lib32-mesa.sh
   echo 'export LIBGL_DRIVERS_PATH=$LIBGL_DRIVERS_PATH:/usr/lib32/xorg/modules/dri' >> ${pkgdir}/etc/profile.d/lib32-mesa.sh
   chmod +x ${pkgdir}/etc/profile.d/lib32-mesa.sh
}

Can anybody confirm or see whether anything is wrong in that pkgbuild?

Last edited by Cdh (2011-12-19 13:22:56)

lordheavy · 2011-12-19 15:48:10

lordheavy wrote:

I've just created a repo for mesa from git and following schema from extra/testing.
These packages are built inside a clean testing chroot, so you need to have testing repo enabled.
[mesa-git]
Server = http://pkgbuild.com/~lcarlier/$repo/$arch
then pacman -Syu mesa-git yourvideodriver (intel-dri-git or ati-dri-git or nouveau-dri-git)
No yet lib32 but it should follow soon
++

Ok, now lib32-* is also available for x86_64 of course, it need multilib-testing to be enabled and can be installed like this:

pacman -Syu lib32-mesa-git yourvideodriver (lib32-intel-dri-git or lib32-ati-dri-git or lib32-nouveau-dri-git)

sources files are here:
http://pkgbuild.com/~lcarlier/mesa-git/sources

I've tested Ennemy Territory:Quake Wars, no problems here.

++

lordheavy · 2011-12-19 15:56:47

Cdh wrote:

Can anybody confirm or see whether anything is wrong in that pkgbuild?

You should try with a native 32 bits game to see if the problem is coming from mesa or wine, or you can try glxinfo32 from lib32-mesa-demos.

Cdh · 2011-12-19 16:59:53

Interesting... Return to Castle Wolfenstein runs fine, but doom3 (binaries from back then) won't start with

libGL error: dlopen /usr/lib32/xorg/modules/dri/r600_dri.so failed (libstdc++.so.6: version `GLIBCXX_3.4.9' not found (required by /usr/lib32/xorg/modules/dri/r600_dri.so))

lordheavy · 2011-12-19 18:13:51

Cdh wrote:

Interesting... Return to Castle Wolfenstein runs fine, but doom3 (binaries from back then) won't start with
libGL error: dlopen /usr/lib32/xorg/modules/dri/r600_dri.so failed (libstdc++.so.6: version `GLIBCXX_3.4.9' not found (required by /usr/lib32/xorg/modules/dri/r600_dri.so))

You need to remove the libstdc++.so (something like that) installed with doom3

Cdh · 2011-12-19 20:53:44

Oh, thanks, didn't quite notice that there was one. Haven't played it in a while and the serial number seems to be gone so I need to search it before I can actually test gameplay.

So at least return to castle wolfenstein works well. I'm curious about the games in wine. For warcraft3 maybe it's a race condition so it runs when the wine debug messages slow its startup down?

Cdh · 2011-12-20 10:11:55

I have a question to lib32-libgl...

 ln -s libglx.xorg "${pkgdir}/usr/lib32/xorg/modules/extensions/libglx.so"

What?

I mean,

$ pkgfile libglx.xorg
extra/xorg-server
$

There is no libglx.xorg for /usr/lib32/xorg/modules/extensions... Is this symlink supposed to be broken?

lordheavy · 2011-12-20 10:21:34

Cdh wrote:

I have a question to lib32-libgl...
 ln -s libglx.xorg "${pkgdir}/usr/lib32/xorg/modules/extensions/libglx.so"
What?
I mean,
$ pkgfile libglx.xorg
extra/xorg-server
$
There is no libglx.xorg for /usr/lib32/xorg/modules/extensions... Is this symlink supposed to be broken?

You're right, this link is useless :-)

Cdh · 2011-12-20 13:08:40

So it's not needed for lib32? I still have trouble finding out what the cause for its causing wine games to not work is... I have also tried the stable wine I built some days ago but it's exactly the same problem...
LIBGL_DEBUG=verbose doesn't tell me about problems...

The last lib32 mesa I had was from perry3d's repo and it worked fine: http://spiralinear.org/perry3d/x86_64/l … pkg.tar.xz
This is built against lib32-llvm 2.9.

So it has to be my mesa's fault. Now I could accept just a broken mesa... but why on earth does it work with WINEDEBUG=+all?

CFLAGS="-march=native -fomit-frame-pointer -O2 -pipe -fstack-protector --param=ssp-buffer-size=4 -D_FORTIFY_SOURCE=2"
CXXFLAGS="-march=native -fomit-frame-pointer -O2 -pipe -fstack-protector --param=ssp-buffer-size=4 -D_FORTIFY_SOURCE=2"
LDFLAGS="-Wl,-O1,--sort-common,--as-needed,-z,relro,--hash-style=gnu"
MAKEFLAGS="-j5"

Arch Linux

#2151 2011-12-04 09:15:25

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2152 2011-12-05 00:16:12

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2153 2011-12-05 01:14:23

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2154 2011-12-05 14:49:00

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2155 2011-12-07 21:35:02

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2156 2011-12-07 23:33:46

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2157 2011-12-08 13:47:55

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2158 2011-12-09 08:04:34

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2159 2011-12-10 07:52:00

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2160 2011-12-12 05:04:19

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2161 2011-12-12 10:23:56

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2162 2011-12-12 17:43:21

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2163 2011-12-12 22:33:23

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2164 2011-12-13 05:52:17

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2165 2011-12-18 20:19:53

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2166 2011-12-19 00:11:42

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2167 2011-12-19 13:15:21

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2168 2011-12-19 15:48:10

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2169 2011-12-19 15:56:47

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2170 2011-12-19 16:59:53

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2171 2011-12-19 18:13:51

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2172 2011-12-19 20:53:44

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2173 2011-12-20 10:11:55

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2174 2011-12-20 10:21:34

Re: Discussion: the ati/radeon open source drivers & radeon repository

#2175 2011-12-20 13:08:40

Re: Discussion: the ati/radeon open source drivers & radeon repository

Board footer