The AMD/ATI Bar & Grill

IVIonty · 2013-02-21 12:51:49

cytrynek wrote:

IVIonty wrote:
So HD 5000 need catalyst-stable, HD 4000 needs catalyst legacy hd234k and Xorg older then 1.13.
I think you need to install Xorg 1.12 + catalyst-hd234k in order to use your cards without problems.
Still not working. I installed catalyst-hd234k (I don't know why beacuse i want run Radeon HD 5000) and Xorg 1.12 and its not working.
When i installed catalyst-utils and catalsyst-hook with Xorg 1.13 i log looks good but GDM don start ...
Can you help me?

Sorry I understood the problem was make the hd 4000 work.

For the HD 5000 I quote Ztjuh, you need catalyst-stable.

When you will do aticonfig --initial check that in xorg.conf under device section fglrx is linked to pci 02:00:00, your HD 5000 card

Last edited by IVIonty (2013-02-21 13:45:22)

Vi0L0 · 2013-02-21 16:25:42

IVIonty wrote:

cytrynek wrote:
(...)
For the HD 5000 I quote Ztjuh, you need catalyst-stable.

Actually he don't need, legacy driver is able to work with radeons hd 5 6 7 k.
I'm now running catalyst-utils and catalyst-hook 12.4 (yes, from catalyst-hd234k-12.4 repo) + xorg111 repo (last supported by 12.4) + 3.8 kernel on my hd 7850 and it's fine.
I thought that your's, IVIonty, first advice about using hd234k was good.
@cytrynek: ~~on your first post you mentioned that 12.4 is the last catalyst that should work for you, this could be true, did you test it after my update?~~ - sorry, im kinda exhausted today, still 12.4 is worth to check

IVIonty wrote:

When you will do aticonfig --initial check that in xorg.conf under device section fglrx is linked to pci 02:00:00, your HD 5000 card

this is also good advice and worth to check

Last edited by Vi0L0 (2013-02-21 16:32:55)

Anon11 · 2013-02-22 05:57:34

IVIonty wrote:

I tested the Xorg.conf for a day and seems that it fixed the kernel panic until now.
I share the modification and I hope it lasts. It's an ugly workaround according to me, but works:
I don't know X server behaviour very well, so probably I'm totally wrong, but I hypothesize that kernel panic with igpu active is caused by the system that isn't able to successfully unload fglrx module at shutdown/reboot. This error in unloading fglrx may be caused by the lacking of description of powersave gpu in Xorg. In fact with aticonfig --initial xorg.conf is configured for the first gpu ready, that is always the discrete gpu.
So I imagine the APU/GPU system as a multi-head system, but with two gpu and only a Xorg screen/monitor active from time to time.
I deleted Xorg.conf and created a new one with:
# aticonfig -f --initial --adapter=all
So aticonfig created two module sections for each gpu with fglrx as dedicated kernel module.
But with the default screen layout the system is unusable.
Then I tryed out some layout to manage the screen/monitor for each gpu, trying with one screen shared, two separated screens for each and so on. I found the following the most suitable ServerLayout section:
Section "ServerLayout"
	Identifier     "aticonfig Layout"
	Screen      0  "aticonfig-Screen[0]-0" 0 0
	Screen         "aticonfig-Screen[1]-0" Relative "aticonfig-Screen[0]-0" x y
EndSection
Where x and y must be the width and height of the monitor resolution (1024 and 768, 1366 and 768, 1920 and 1080, and so on).
After a X restart and
# aticonfig --px-igpu
to switch to powersave gpu the system seems to work properly.

I didn't get problem with this yet ( half day use only ) , but , this keeps the discrete gpu active after switching to integrated.
My laptop temp is up 10-15 C on desktop usage
Normally with integrated active catalyst does not show the discrete gpu in information tab but now i can see both of them.
(maybe thats ok but temperature verifies that dGPU is not sleeping) .

Switchable GFX shows powersaving GPU selected.
Do you get same problem?

EDIT : upgrade to latest catalyst + aticonfig -f --initial --adapter=all also broke using tty 1/2/3 etc and broke brightness changes for me.
i went back to 12.8 and everything is fine.

Last edited by Anon11 (2013-02-22 06:18:19)

IVIonty · 2013-02-22 08:44:23

I didn't get problem with this yet ( half day use only ) , but , this keeps the discrete gpu active after switching to integrated.
My laptop temp is up 10-15 C on desktop usage
Normally with integrated active catalyst does not show the discrete gpu in information tab but now i can see both of them.
(maybe thats ok but temperature verifies that dGPU is not sleeping) .
Switchable GFX shows powersaving GPU selected.
Do you get same problem?
EDIT : upgrade to latest catalyst + aticonfig -f --initial --adapter=all also broke using tty 1/2/3 etc and broke brightness changes for me.
i went back to 12.8 and everything is fine.

Do you have and APU/GPU hybrid system or two mobile GPUs?

Unfortunatly I tested the workaround only with an APU/GPU system only, and the dGPU seems to stay out of the ring.

The problem of broken tty and brightness might be caused by the KMS, do you use nomodeset as kernel option at startup ?

Vi0L0 · 2013-02-22 08:51:12

I don't remember if there was a way of turning off one of the gpus with catalyst installed when it's not working by default, although this problem is showing up from time to time...
Maybe a good solution here would be to use -pxp package and perform the switching like:
- use catalyst for discrete gpu, when you need more power
- use radeon oss for integrated gpu (with accompaniment of vgaswitcheroo to disable discrete gpu), to incerase battery life
Ofcourse this solution is good only if radeon oss doesn't eat too much energy.

edit:

IVIonty wrote:

The problem of broken tty and brightness might be caused by the KMS, do you use nomodeset as kernel option at startup ?

We have already reported the problem with tty on 13.2 beta, it's catalyst fault. Don't know about brightness...

Last edited by Vi0L0 (2013-02-22 08:53:17)

IVIonty · 2013-02-22 09:16:41

Vi0L0 wrote:

I don't remember if there was a way of turning off one of the gpus with catalyst installed when it's not working by default, although this problem is showing up from time to time...
Maybe a good solution here would be to use -pxp package and perform the switching like:
- use catalyst for discrete gpu, when you need more power
- use radeon oss for integrated gpu (with accompaniment of vgaswitcheroo to disable discrete gpu), to incerase battery life
Ofcourse this solution is good only if radeon oss doesn't eat too much energy.

Good hint, I tried out that as first solution when I encounter the kernel panic problem. Unfortunatly using radeon for the APU cause system overheating and an energy consumption comparable with discrete cpu+catalyst.

Vi0L0 wrote:

edit:
IVIonty wrote:
The problem of broken tty and brightness might be caused by the KMS, do you use nomodeset as kernel option at startup ?
We have already reported the problem with tty on 13.2 beta, it's catalyst fault. Don't know about brightness...

Ok, I doesn't tested already cat 13.2 beta, only 13.1

Anon11 · 2013-02-22 10:20:47

IVIonty wrote:

I didn't get problem with this yet ( half day use only ) , but , this keeps the discrete gpu active after switching to integrated.
My laptop temp is up 10-15 C on desktop usage
Normally with integrated active catalyst does not show the discrete gpu in information tab but now i can see both of them.
(maybe thats ok but temperature verifies that dGPU is not sleeping) .
Switchable GFX shows powersaving GPU selected.
Do you get same problem?
EDIT : upgrade to latest catalyst + aticonfig -f --initial --adapter=all also broke using tty 1/2/3 etc and broke brightness changes for me.
i went back to 12.8 and everything is fine.
Do you have and APU/GPU hybrid system or two mobile GPUs?
Unfortunatly I tested the workaround only with an APU/GPU system only, and the dGPU seems to stay out of the ring.
The problem of broken tty and brightness might be caused by the KMS, do you use nomodeset as kernel option at startup ?

Actually I have 6755g2 - llano apu + 6750M gpu .
Yes i pass nomodeset in grub.

I think ill try it again later this week with maybe 13.1
How did you verify that dGPU is asleep ?

Violo - Both of them work fine with catalyst 12.8 , i just like to run integrated one as i dont game much in linux and its a lot cooler.
Using oss driver is actually quite usable but temperature increases by about 10C.
Your 12.8 repo works great with everything updated

thanks

IVIonty · 2013-02-22 15:51:17

Anon11 wrote:

Actually I have 6755g2 - llano apu + 6750M gpu .
Yes i pass nomodeset in grub.
I think ill try it again later this week with maybe 13.1
How did you verify that dGPU is asleep ?

The same way you described some post ago:

1) checking output of aticonfig --pxl;
2) checking information tab in aticccle;
3) checking the power-consumption and system heating levels

As additional info, I tried out the edited xorg.conf in the following configurations:

- Arch / linux 3.7.9-1 / catalyst 13.1-2 / xorg 1.13.2-1
- Chakra / linux 3.7.6-1 / catalyst 13.1-3 / xorg 1.12.3-3
- Debian Testing / linux 3.2.0-38 / catalyst 13.1-1 / xorg 1.11.1-1

At this point I think that this workaround does not work in a wide range of system configuration (hardware and software I mean)

I'll continue testing (13.2 beta for example) and hope AMD will fix soon is log time open bugs

IVIonty · 2013-02-23 11:15:56

@ Anon11

I would to install and test Catalyst 12.8 to make a comparison with 13.1 and 13.2 beta from Vi0l0 repositoires.
My stupid question is: I must take old Catalyst packages from A.R.M. package snapshot http://arm.konnichi.com/search/index.ph … ommunity=1 or there is another/better source as a dedicated repository for older Catalyst versions?

EDIT: argh all I need was there http://catalyst.wirephire.com/repo/catalyst/x86_64/ ...
Another stupid question, I have to take and install old catalyst packages with pacman -U or there is a way to take them via pacman -S ?
Added the repository but with pacman -Ss catalyst I see only 13.1 ones

Thanks in advance

Last edited by IVIonty (2013-02-23 12:14:23)

Unia · 2013-02-23 12:46:19

^ Pacman -U should do the trick.

IVIonty · 2013-02-23 12:57:48

Unia wrote:

^ Pacman -U should do the trick.

Done, thanks

Vi0L0 · 2013-02-23 13:55:19

IVIonty wrote:

Unia wrote:
^ Pacman -U should do the trick.
Done, thanks

There was an info about archive catalyst repositories on the wiki page, but was removed guess in the time when wiki was cleaned up.
Using archive repo on a top of pacman.conf file is easier than pacman -U, also when using archive repo pacman won't warn you that you're using older versions of packages and so won't try to update it.
Just look at the list of files on the repo page ie http://catalyst.wirephire.com/repo/catalyst/x86_64/ - you will find files with *.db extension, those files are repo databases.
Now you can copy name of this database, without extension, and paste it into pacman.conf as a repo ie there is catalyst-12.8.db file with informations about catalyst 12.8 packages, pacman.conf entry should look like:

[catalyst-12.8] 
Server = http://catalyst.wirephire.com/repo/catalyst/$arch

then
pacman -Sy && pacman -Suu
should do the job

Last edited by Vi0L0 (2013-02-23 14:02:14)

IVIonty · 2013-02-23 22:38:31

Vi0L0 wrote:

IVIonty wrote:
Unia wrote:
^ Pacman -U should do the trick.
Done, thanks
There was an info about archive catalyst repositories on the wiki page, but was removed guess in the time when wiki was cleaned up.
Using archive repo on a top of pacman.conf file is easier than pacman -U, also when using archive repo pacman won't warn you that you're using older versions of packages and so won't try to update it.
Just look at the list of files on the repo page ie http://catalyst.wirephire.com/repo/catalyst/x86_64/ - you will find files with *.db extension, those files are repo databases.
Now you can copy name of this database, without extension, and paste it into pacman.conf as a repo ie there is catalyst-12.8.db file with informations about catalyst 12.8 packages, pacman.conf entry should look like:
[catalyst-12.8] 
Server = http://catalyst.wirephire.com/repo/catalyst/$arch
then
pacman -Sy && pacman -Suu
should do the job

Clean and easy, thank you very much Vi0l0.

LithiumPT · 2013-02-25 15:16:10

Hello!

I'd just like to know if anyone with a HD4870 and using the FOSS driver (xf86-video-ati) gets 'playable' fps in 3D games...
I've seen some youtube videos of people running UT2k4 on a HD4650 with playable fps... but I can't get past 10fps even on UT99, and 4fps on CS1.6 (steam), I'm not expecting great performance, but I think something is off on my setup.. other than 3D performance, 2D performance is nice even flash videos play acceptably smooth..

DarksideEE7 · 2013-02-26 06:38:39

Looks like I'm experiencing the bug where X crashes when I minimize (or restore possibly) konsole from the task bar. I'm not sure if it's limited to konsole or other apps too....pretty sure it's just konsole though. Pretty annoying :\ It seemed to be fixed for a while, then I think the last catalyst-dkms update caused it to come back.

Last edited by DarksideEE7 (2013-02-26 06:39:32)

ChemBro · 2013-02-26 16:39:19

I have an update problem. lib32-mesa-libgl and lib32-catalyst-utils are in conflict and I don't know, why it wants lib32-mesa-libgl installed.

Okay, looks like it is because of Steam. It also wants to install the nvidia-drivers for my AMD card? What?

Last edited by ChemBro (2013-02-26 16:43:56)

Vi0L0 · 2013-02-26 18:31:21

looks like lib32-mesa-libgl is doing some mess with it's
Replaces : lib32-libgl

Does this mean that lib32-catalyst-utils should provide lib32-mesa-libgl from now on?
Edit: lol, i firstly thought that this would be a joke, but no - it really should provide lib32-mesa-libgl (lib32-nvidia-utils was just updated with it), kinda tricky-hacky-not-true resolution...
i will update last mentioned repos today

Last edited by Vi0L0 (2013-02-26 18:53:10)

Vi0L0 · 2013-02-26 18:33:23

Edit: sorry for double posting, miss-click

let's use this miss-click for something better...

MrDeepPurple have found a nice solution to turn off discrete graphics card in a dual graphics environment - acpi_call (you can find it out on AUR), it worked well for him on he's amd-intel powerXpress hybrid laptop with the company of catalyst-total-pxp package

Last edited by Vi0L0 (2013-02-26 18:43:16)

Unia · 2013-02-26 19:19:21

Vi0L0, perhaps this is of interest: https://bbs.archlinux.org/viewtopic.php?id=158744

Vi0L0 · 2013-02-26 20:54:30

Unia wrote:

Vi0L0, perhaps this is of interest: https://bbs.archlinux.org/viewtopic.php?id=158744

Kinda is.

I knew about incoming mesa 9.1 update from arch-dev-public, thought the way of how this happen surprised me.
I still don't know which was better:
- pulling mesa-libgl whithout replace=libgl option, so package wasn't updated automatically (i guess) edit: ah wait mesa-libgl was updated and now in 9.1-2 it has replace=libgl;
- pulling lib32-mesa-libgl with replace=lib32-libgl option, which causes lib32-nvidia/catalyst-utils looks silly with provides=lib32-mesa-libgl (silly because those packages are not providing mesa's libgl)

Anyway i have updated lib32-catalyst-utils and lib32-catalyst-utils-pxp in following x86_64 repos: [catalyst] [catalyst-stable] [catalyst-12.8] [catalyst-hd234k] and [catalyst-hd234k-12.4]
If someone wants other repos to be updated - feel free to post for it (or mail me).
I've also updated AUR's: catalyst-test, catalyst-total, catalyst-total-pxp, catalyst-total-hd234k

edit: hah, it could turn out that those updates were unnecessary, because now freshly updated lib32-nvidia-utils is free of provides=lib32-mesa-libgl , so things were cleaned up, nice...
well... my update should not harm anything, i'm not gonna waste my and your time for another update which will revert the provides=lib32-mesa-libgl changes, i will just clean it up next time

edit: nope, the bug was elsewhere - it was steam that was depending on lib32-mesa-libgl :> and now with -4 it is fixed (i guess, still cannot get it from mirror)
edit: i'm too old for this and too exhausted latelly... i should think on myself instead of blindlly looking at mainline updates...

btw: have you already updated you kernel? kernels from 3.3 to 3.8 including have got horrible security hole : http://www.h-online.com/open/news/item/ … 10597.html
linux 3.7.9-2 is already patched with CVE-2013-1763.patch, don't forget to update your custom kernel.

Last edited by Vi0L0 (2013-02-26 21:59:45)

Vi0L0 · 2013-02-27 07:39:24

DarksideEE7 wrote:

Looks like I'm experiencing the bug where X crashes when I minimize (or restore possibly) konsole from the task bar. I'm not sure if it's limited to konsole or other apps too....pretty sure it's just konsole though. Pretty annoying :\ It seemed to be fixed for a while, then I think the last catalyst-dkms update caused it to come back.

I doubt it's catalyst-dkms fault, you can easily check it out by switching to ie. catalyst-hook.
Since you are experiencing this bug from time to time i would rather say that you have got problem with some package' update, it could also be catalyst' fault.
Look at ~/.xsession-errors after a crash, maybe it will tell something.

Last edited by Vi0L0 (2013-02-27 07:39:52)

DarksideEE7 · 2013-02-27 07:47:30

Vi0L0 wrote:

DarksideEE7 wrote:
Looks like I'm experiencing the bug where X crashes when I minimize (or restore possibly) konsole from the task bar. I'm not sure if it's limited to konsole or other apps too....pretty sure it's just konsole though. Pretty annoying :\ It seemed to be fixed for a while, then I think the last catalyst-dkms update caused it to come back.
I doubt it's catalyst-dkms fault, you can easily check it out by switching to ie. catalyst-hook.
Since you are experiencing this bug from time to time i would rather say that you have got problem with some package' update, it could also be catalyst' fault.
Look at ~/.xsession-errors after a crash, maybe it will tell something.

I'll check it out if it keeps happening. The reason I thought it was related to catalyst-dkms (or other such ATI related package) is because I never got the crash when I was running integrated Ivy Bridge (Intel HD4000) graphics. When I installed the 7950 it started happening often, then it stopped. I haven't seen it happen since the last catalyst-dkms upgrade. I'll check it out and see if the server logs yield any useful info.

I know in the past konsole has been a problem child :\

Vi0L0 · 2013-02-27 08:16:42

DarksideEE7 wrote:

I know in the past konsole has been a problem child :\

Indeed it was, but remember - with catalyst if problem was once solved it can and probably will return in the future
I in example noticed similar bug in one of the 12.11 betas - while resizing konsole X was crashing, but only sometimes, it was actually pretty hard to reproduce - this problem is returning to me from time to time like the boomerang, the good thing is that sometimes i need to wait 1,5 y for it to get back

LithiumPT wrote:

Hello!
I'd just like to know if anyone with a HD4870 and using the FOSS driver (xf86-video-ati) gets 'playable' fps in 3D games...
I've seen some youtube videos of people running UT2k4 on a HD4650 with playable fps... but I can't get past 10fps even on UT99, and 4fps on CS1.6 (steam), I'm not expecting great performance, but I think something is off on my setup.. other than 3D performance, 2D performance is nice even flash videos play acceptably smooth..

i would suggest asking on this forum

Vi0L0 · 2013-02-27 08:43:33

btw: tressfx - yet another technology basing on DirectCompute... Why promoting opencl if you gonna use ms' **** any way?

darn... my work' pc is freaking out, i'm sure now i clicked 'edit', gotta format it guess, anyway sorry for another double post LOL

Last edited by Vi0L0 (2013-02-27 08:46:08)

PhotonX · 2013-02-27 11:30:00

I have following problem: All versions of catalyst-hook up to 13.1 work fine but the update to 13.2-6 breaks X with the error

[   109.895] (WW) fglrx(0): Kernel Module version does *not* match driver.
[   109.895] (EE) fglrx(0): incompatible kernel module detected - HW accelerated OpenGL will not work
[   109.895] 
Backtrace:
[   109.895] 0: /usr/bin/X (xorg_backtrace+0x28) [0x561068]
[   109.895] 1: /usr/bin/X (0x400000+0x164e29) [0x564e29]
[   109.895] 2: /usr/lib/libpthread.so.0 (0x7fdb6e16a000+0xf1e0) [0x7fdb6e1791e0]
[   109.895] 3: /usr/lib/xorg/modules/drivers/fglrx_drv.so (swlDrmFreeSurfaces+0x42) [0x7fdb6a9798d2]
[   109.896] 4: /usr/lib/xorg/modules/drivers/fglrx_drv.so (xdl_xs111_atiddxDriCloseScreen+0x14d) [0x7fdb6a8d8d5d]
[   109.896] 5: /usr/lib/xorg/modules/drivers/fglrx_drv.so (xdl_xs111_atiddxDriScreenInit+0x8ad) [0x7fdb6a8d841d]
[   109.896] 6: /usr/lib/xorg/modules/drivers/fglrx_drv.so (xdl_xs111_atiddxScreenInit+0xb18) [0x7fdb6a8d24a8]
[   109.896] 7: /usr/bin/X (AddScreen+0x188) [0x4341d8]
[   109.896] 8: /usr/bin/X (InitOutput+0x296) [0x470086]
[   109.896] 9: /usr/bin/X (0x400000+0x23036) [0x423036]
[   109.896] 10: /usr/lib/libc.so.6 (__libc_start_main+0xf5) [0x7fdb6cffaa15]
[   109.896] 11: /usr/bin/X (0x400000+0x234ad) [0x4234ad]
[   109.896] Segmentation fault at address 0x888
[   109.896] 
Fatal server error:
[   109.896] Caught signal 11 (Segmentation fault). Server aborting
[   109.896] 
[   109.896] 
Please consult the The X.Org Foundation support 
	 at http://wiki.x.org
 for help. 
[   109.896] Please also check the log file at "/var/log/Xorg.0.log" for additional information.
[   109.896] 
[   110.444] Server terminated with error (1). Closing log file.

Any hints, how I could analyse the problem?

Arch Linux

#4826 2013-02-21 12:51:49

Re: The AMD/ATI Bar & Grill

#4827 2013-02-21 16:25:42

Re: The AMD/ATI Bar & Grill

#4828 2013-02-22 05:57:34

Re: The AMD/ATI Bar & Grill

#4829 2013-02-22 08:44:23

Re: The AMD/ATI Bar & Grill

#4830 2013-02-22 08:51:12

Re: The AMD/ATI Bar & Grill

#4831 2013-02-22 09:16:41

Re: The AMD/ATI Bar & Grill

#4832 2013-02-22 10:20:47

Re: The AMD/ATI Bar & Grill

#4833 2013-02-22 15:51:17

Re: The AMD/ATI Bar & Grill

#4834 2013-02-23 11:15:56

Re: The AMD/ATI Bar & Grill

#4835 2013-02-23 12:46:19

Re: The AMD/ATI Bar & Grill

#4836 2013-02-23 12:57:48

Re: The AMD/ATI Bar & Grill

#4837 2013-02-23 13:55:19

Re: The AMD/ATI Bar & Grill

#4838 2013-02-23 22:38:31

Re: The AMD/ATI Bar & Grill

#4839 2013-02-25 15:16:10

Re: The AMD/ATI Bar & Grill

#4840 2013-02-26 06:38:39

Re: The AMD/ATI Bar & Grill

#4841 2013-02-26 16:39:19

Re: The AMD/ATI Bar & Grill

#4842 2013-02-26 18:31:21

Re: The AMD/ATI Bar & Grill

#4843 2013-02-26 18:33:23

Re: The AMD/ATI Bar & Grill

#4844 2013-02-26 19:19:21

Re: The AMD/ATI Bar & Grill

#4845 2013-02-26 20:54:30

Re: The AMD/ATI Bar & Grill

#4846 2013-02-27 07:39:24

Re: The AMD/ATI Bar & Grill

#4847 2013-02-27 07:47:30

Re: The AMD/ATI Bar & Grill

#4848 2013-02-27 08:16:42

Re: The AMD/ATI Bar & Grill

#4849 2013-02-27 08:43:33

Re: The AMD/ATI Bar & Grill

#4850 2013-02-27 11:30:00

Re: The AMD/ATI Bar & Grill

Board footer