The AMD/ATI Bar & Grill

DarksideEE7 · 2010-09-17 04:54:26

Vi0L0 wrote:

DarksideEE7 wrote:
Vi0L0 wrote:
Are you sure you enabled ATi 2D Acceleration?
# cat /var/log/Xorg.0.log | grep 2D
[ 16.373] (==) fglrx(0): ATI 2D Acceleration Architecture enabled
No it's not enabled....I wasn't aware that you had to manually enable it. i'm going to give that a shot. Thanks.
once you typed `aticonfig --set-pcs-str=DDX,ForceXAA,TRUE` - it will always work (its stored in /etc/ati/amdpcsdb).
I suggest to remove amdpcsdb in the way i posted here.

All right I did that and I've been doing some work with several windows open, and I've noticed that at first resizing and moving is very fast, however after a short period of time it begins to delay and resize slowly. Even maximizing programs from Smooth Tasks has a delay of ~3s at times. Strange.

After closing the other windows it goes back to normal speed. This is on a very powerful system, mind you. I'm going to compare the two and see which performs better.

foutrelis · 2010-09-17 08:22:51

@Vi0L0: The catalyst package says it's been built on September 28. This is also the mtime of /lib/modules/2.6.35-ARCH/video/fglrx.ko, which I believe causes depmod to regenerate the module dependencies on every boot, since it's called from /etc/rc.sysinit and the system time is older than that of the fglrx module.

If you have invented a time machine, it'd be awesome if you would share the schematics. Otherwise, a rebuild of the catalyst package with correct system time would be fine too!

DarksideEE7 · 2010-09-17 08:31:47

Well I just noticed that Crossfire is no longer working on my 4870x2..............it was working with a simple xorg.conf with 10.8 and 10.7:

sudo aticonfig --initial --adapter=all

foutrelis · 2010-09-17 09:57:43

Does anyone else get his screen all garbled up when switching out of a full screen YouTube video? Same thing was happening yesterday with full screen videos in mplayer and smplayer. A Catalyst 10.9 bug maybe?

DarksideEE7 · 2010-09-17 10:07:19

foutrelis wrote:

Does anyone else get his screen all garbled up when switching out of a full screen YouTube video? Same thing was happening yesterday with full screen videos in mplayer and smplayer. A Catalyst 10.9 bug maybe?

No, but sometimes when I watch a video and go full screen I have to minimize Firefox and the video will be on my desktop (so to speak).

Vi0L0 · 2010-09-17 11:31:22

foutrelis wrote:

@Vi0L0: The catalyst package says it's been built on September 28. This is also the mtime of /lib/modules/2.6.35-ARCH/video/fglrx.ko, which I believe causes depmod to regenerate the module dependencies on every boot, since it's called from /etc/rc.sysinit and the system time is older than that of the fglrx module.
If you have invented a time machine, it'd be awesome if you would share the schematics. Otherwise, a rebuild of the catalyst package with correct system time would be fine too!

Well... it's looking like my VirtualBox is working on Time Machine without even notifying me...
If VB will invent such great thing i'm gonna take volunteers into a journey full of dinosaurs, gnomes and space ships

Thanks, it's fixed now.

Last edited by Vi0L0 (2010-09-17 11:31:50)

Vi0L0 · 2010-09-17 11:39:50

DarksideEE7 wrote:

Well I just noticed that Crossfire is no longer working on my 4870x2..............it was working with a simple xorg.conf with 10.8 and 10.7:
sudo aticonfig --initial --adapter=all

AMD brakes crossfire support for 4850 and 4870 in 10.8 and 10.9 - that's why they bring 10.8a and 10.9a HotFixes for Windows.
I didn't know it's touching also Linux.
I suggest you to fill bug report at ati.cchtml.com

Vi0L0 · 2010-09-17 11:47:29

foutrelis wrote:

Does anyone else get his screen all garbled up when switching out of a full screen YouTube video? Same thing was happening yesterday with full screen videos in mplayer and smplayer. A Catalyst 10.9 bug maybe?

Try with re-installation of mplayer-vaapi and libva packages.
Edit: not, i guess it won't help... i just woke up and not thinking in a right way
umm... so maybe with redirect/un-redirect fullscreen option of your compositing manager.

Last edited by Vi0L0 (2010-09-17 11:50:16)

foutrelis · 2010-09-17 11:56:12

Vi0L0 wrote:

foutrelis wrote:
Does anyone else get his screen all garbled up when switching out of a full screen YouTube video? Same thing was happening yesterday with full screen videos in mplayer and smplayer. A Catalyst 10.9 bug maybe?
Try with re-installation of mplayer-vaapi and libva packages.
Edit: not, i guess it won't help... i just woke up and not thinking in a right way
umm... so maybe with redirect/un-redirect fullscreen option of your compositing manager.

Not sure how to do that. Never had this issue before Catalyst 10.9. I'm using plain Gnome, so I suppose my compositing manager would be metacity. I'll look for options named like redirect.

Vi0L0 · 2010-09-17 12:37:28

foutrelis wrote:

Vi0L0 wrote:
foutrelis wrote:
Does anyone else get his screen all garbled up when switching out of a full screen YouTube video? Same thing was happening yesterday with full screen videos in mplayer and smplayer. A Catalyst 10.9 bug maybe?
Try with re-installation of mplayer-vaapi and libva packages.
Edit: not, i guess it won't help... i just woke up and not thinking in a right way
umm... so maybe with redirect/un-redirect fullscreen option of your compositing manager.
Not sure how to do that. Never had this issue before Catalyst 10.9. I'm using plain Gnome, so I suppose my compositing manager would be metacity. I'll look for options named like redirect.

If you aren't using any "Desktop Effects" then it should work fine, i just switched to my plain gnome and it's working as it should...
So it must be a 10.9 bug in your case.
Maybe check fresh flashplugin with x86_64 support.

foutrelis · 2010-09-17 12:53:22

Vi0L0 wrote:

foutrelis wrote:
Not sure how to do that. Never had this issue before Catalyst 10.9. I'm using plain Gnome, so I suppose my compositing manager would be metacity. I'll look for options named like redirect.
If you aren't using any "Desktop Effects" then it should work fine, i just switched to my plain gnome and it's working as it should...
So it must be a 10.9 bug in your case.
Maybe check fresh flashplugin with x86_64 support.

I switched to the new prerelease version of Flash earlier today with no improvement. Also, I don't use any desktop effects.

The issue isn't occurring anymore with mplayer/smplayer, only with Flash, and it's reproducible every time I leave full screen mode while watching a video. After that, I need to restart X to fix the corrupted display. It's also happening on my laptop which uses the same IGP (Radeon HD 4200).

I'll do some more testing. Thanks for the suggestions though.

Last edited by foutrelis (2010-09-17 13:09:09)

Vi0L0 · 2010-09-17 13:20:33

Any x86_64 user want fresh chromium-dev 7.0.525? I built it yesterday and you may download it from here.

In this new build i modified "Exec=" line of /usr/share/applications/chromium-dev.desktop, so it now looks like this:

Exec=chromium-dev %U --enable-gpu-plugin --enable-accelerated-compositing --enable-webgl --enable-extensions

Feel free to modify it if it's not working fine for you.

New build is working far better than 7.0.509 (that i previously uploaded) with accelerated 2d canvas, so you may safely add
--enable-accelerated-2d-canvas
into "Exec=" line

@foutrelis:
maybe this chromium-dev will help you with youtube, because it's rendering flash through OpenGL (+ no-flickering if v-sync is enabled in amdcccle)

foutrelis · 2010-09-17 13:31:16

Vi0L0 wrote:

@foutrelis:
maybe this chromium-dev will help you with youtube, because it's rendering flash through OpenGL (+ no-flickering if v-sync is enabled in amdcccle)

That sounds nice, but I'll stick with Firefox for the moment. I found another mention of the issue I'm experiencing and will shortly file a report on http://ati.cchtml.com/.

Edit: Oh, and switching to a tty and then back to X like the post I linked above says, actually works.

Edit 2: Filed bug report here: http://ati.cchtml.com/show_bug.cgi?id=1913. Let's see what happens.

Last edited by foutrelis (2010-09-17 14:08:24)

codycarey · 2010-09-17 15:52:51

Experienced a single short-lived instance of black box artifacts in Chromium-dev this morning with the latest drivers. It appeared at the bottom of the browser, grew a little bit as I scrolled down the page, and then *poof*. I switched back to 'extra/xorg-server' yesterday to test it out and it's perfect except for this one incident.

ragnarok2040 · 2010-09-17 20:23:11

Wow, I didn't notice this before, but every window now has vsync, not just fullscreen windows. The standalone flash player is tear free, too, and other games and such running in a window are tear free as well.

I did experience some black box problems, when I pushed F11 to render Firefox in fullscreen then switched back. All the other windows except Firefox had turned black, but it only happened once, as doing it repeatedly only results in a split second of black, then the windows redraw.

Last edited by ragnarok2040 (2010-09-17 20:23:38)

LeCrayonVert · 2010-09-17 22:11:09

Vi0L0 wrote:

LeCrayonVert wrote:
Vi0L0 wrote:
I think that removing /etc/ati/amdpcsdb is a "must do" after main catalyst update.
Could this removal be included in the post-install function ?
No, you must kill X/DE before removing file. Simple remove won't give anything because this file will be immediately restored by catalyst drivers.

I've upgraded to 10.9 by using your previously described method (to install catalyst-hook and catalyst-utils). Everything works fine, but there is no /etc/ati directory anymore...and the version displayed in catalyst control center is still the old one...

edit : fixed by reinstalling catalyst-utils

Last edited by LeCrayonVert (2010-09-18 06:36:17)

DarksideEE7 · 2010-09-18 05:32:35

I ended up fixing Crossfire, however I may need to test further for stability. I shouldn't have to do this since I've always had Crossfire working with a simple xorg.conf generated by aticonfig. This is what I did:

rm /etc/X11/xorg.conf && rm /etc/ati/amdpcsdb
aticonfig --initial --adapter=all

Then I noticed this section in xorg.conf that shouldn't be there (at least it hasn't been in the past).

Section "ServerLayout"
        Identifier     "aticonfig Layout"
        Screen      0  "aticonfig-Screen[0]-0" 0 0
        Screen         "aticonfig-Screen[1]-0" RightOf "aticonfig-Screen[0]-0"
EndSection

I guess it was configuring it for dual head with a monitor to the right of my main. I then removed that section and started x, then checked Crossfire with:

aticonfig --lscs

It showed to be enabled, great.
Well I was watching some 1080P SC2 videos on youtube and finished the last video then exited full screen and shifted my attention to my laptop. When I looked back I was at a bash prompt at tty1......X crashed.

Upon inspecting /var/log/messages.log I found:

Sep 17 23:07:00 waterbeast -- MARK --
Sep 17 23:10:40 waterbeast kernel: [fglrx] IRQ 49 Disabled
Sep 17 23:10:40 waterbeast kernel: [fglrx] IRQ 48 Disabled
Sep 17 23:14:54 waterbeast kernel: [fglrx] Firegl kernel thread PID: 5691
Sep 17 23:14:54 waterbeast kernel: [fglrx] IRQ 48 Enabled
Sep 17 23:14:54 waterbeast kernel: [fglrx] Firegl kernel thread PID: 5692
Sep 17 23:14:54 waterbeast kernel: [fglrx] IRQ 49 Enabled
Sep 17 23:14:55 waterbeast kernel: [fglrx] Gart USWC size:1280 M.
Sep 17 23:14:55 waterbeast kernel: [fglrx] Gart cacheable size:508 M.
Sep 17 23:14:55 waterbeast kernel: [fglrx] Reserved FB block: Shared offset:0, size:1000000 
Sep 17 23:14:55 waterbeast kernel: [fglrx] Reserved FB block: Unshared offset:fb7c000, size:484000 
Sep 17 23:14:55 waterbeast kernel: [fglrx] Reserved FB block: Unshared offset:3fffb000, size:5000 
Sep 17 23:14:56 waterbeast kernel: ata1.00: configured for UDMA/100
Sep 17 23:14:56 waterbeast kernel: ata1: EH complete
Sep 17 23:14:56 waterbeast kernel: EXT4-fs (sda3): re-mounted. Opts: commit=0
Sep 17 23:14:56 waterbeast kernel: EXT4-fs (sda4): re-mounted. Opts: commit=0
Sep 17 23:36:39 waterbeast kernel: lo: Disabled Privacy Extensions
Sep 17 23:37:18 waterbeast kernel: DeQueueRunning[0]= TRUE!
Sep 18 00:15:24 waterbeast kernel: [fglrx] IRQ 48 Disabled
Sep 18 00:15:24 waterbeast kernel: [fglrx] IRQ 49 Disabled
Sep 18 00:28:36 waterbeast kernel: [fglrx] Firegl kernel thread PID: 29273
Sep 18 00:28:36 waterbeast kernel: [fglrx] IRQ 48 Enabled
Sep 18 00:28:36 waterbeast kernel: [fglrx] Firegl kernel thread PID: 29274
Sep 18 00:28:36 waterbeast kernel: [fglrx] IRQ 49 Enabled
Sep 18 00:28:38 waterbeast kernel: [fglrx] Gart USWC size:1280 M.
Sep 18 00:28:38 waterbeast kernel: [fglrx] Gart cacheable size:508 M.
Sep 18 00:28:38 waterbeast kernel: [fglrx] Reserved FB block: Shared offset:0, size:1000000 
Sep 18 00:28:38 waterbeast kernel: [fglrx] Reserved FB block: Unshared offset:fb7c000, size:484000 
Sep 18 00:28:38 waterbeast kernel: [fglrx] Reserved FB block: Unshared offset:3fffb000, size:5000

I'll see how it goes from here on out, but if it keeps crashing I'll need to fill out a bug report with ATI or something. I know the 4870x2 isn't their flagship card anymore, but it's still a fine video card, and shouldn't be crippled like it is atm.

Last edited by DarksideEE7 (2010-09-18 05:35:48)

LeCrayonVert · 2010-09-18 10:30:57

Since 10.8, powerplay feature seems to be broken... After intensive GPU usage, fan doesn't spin down as quickly as with the 10.7 driver. Can anyone can confirm that ?

snack · 2010-09-18 10:32:58

LeCrayonVert wrote:

Since 10.8, powerplay feature seems to be broken... After intensive GPU usage, fan doesn't spin down as quickly as with the 10.7 driver. Can anyone can confirm that ?

My laptop runs cool as before with 10.9; no noticeable powerplay issue here (Mobility HD 3470).

mrweasel · 2010-09-18 11:25:36

LeCrayonVert wrote:

Since 10.8, powerplay feature seems to be broken... After intensive GPU usage, fan doesn't spin down as quickly as with the 10.7 driver. Can anyone can confirm that ?

I can confirm that (Lenovo T400, HD3470), although powerplay does not seem to be broken completely. Still the GPU gets slightly hotter than before and the fan keeps spinning more often.

LeCrayonVert · 2010-09-18 14:43:23

So I've done some tests to measure this time. (For example, play a video using GPU acceleration (vaapi) for even just a few seconds)
10.7 : less than 10 seconds
10.8 and 10.9 : between 2 and 5 minutes according to the current load.
radeon free driver (with latest version from git) : between 10 and 15 minutes

Last edited by LeCrayonVert (2010-09-18 14:50:44)

monshin · 2010-09-18 15:53:10

Hi. I would like to have some advice for "optimal" settngs for kwin's compositing. I've noticed slowdowns when using desktop effects (most of the time runs smoothly, but cpu usage rises even when a tooltip appears), and 2d performance drops significantly. Without desktop effects, 2d performance in chromium is great.

I'm currently using kde 4.5.1, catalyst 10.9, and I have an ATI Radeon 3200 in my Wind u230 (AMD Athlon Neo MV-40). Of course I have blur effect disabled.

Thanks in advance.

LeCrayonVert · 2010-09-18 17:02:45

I'm trying to downgrade to 10.7, so I've added catalyst-10.7 according to the wiki, but at the end of the installation of catalyst package, I've got an error during compilation of the fglrx module (from /var/log/catalyst-install.log).

--------
2010-09-18 19:00:26
Building fglrx module for 2.6.35-ARCH kernel ...
--------
+ cp /usr/share/ati/build_mod /tmp/catalyst.ogRxWy -R
+ cd /tmp/catalyst.ogRxWy
+ patch_files
+ file_patch=2.6.35.patch
+ [[ -f 2.6.35.patch ]]
++ pwd
+ make -C /lib/modules/2.6.35-ARCH/build SUBDIRS=/tmp/catalyst.ogRxWy ARCH=x86_64 modules
  CC [M]  /tmp/catalyst.ogRxWy/firegl_public.o
In file included from /tmp/catalyst.ogRxWy/firegl_public.c:451:0:
/tmp/catalyst.ogRxWy/drm_proc.h: In function 'FGLDRM__vma_info':
/tmp/catalyst.ogRxWy/drm_proc.h:497:2: warning : format '%08lx' expects type 'long unsigned int', but argument 5 has type 'phys_addr_t'
  CC [M]  /tmp/catalyst.ogRxWy/kcl_acpi.o
  CC [M]  /tmp/catalyst.ogRxWy/kcl_agp.o
  CC [M]  /tmp/catalyst.ogRxWy/kcl_debug.o
  CC [M]  /tmp/catalyst.ogRxWy/kcl_ioctl.o
/tmp/catalyst.ogRxWy/kcl_ioctl.c: In function 'KCL_IOCTL_AllocUserSpace32':
/tmp/catalyst.ogRxWy/kcl_ioctl.c:196:5: error: implicit declaration of function 'compat_alloc_user_space'
/tmp/catalyst.ogRxWy/kcl_ioctl.c:196:5: warning : return makes pointer from integer without a cast
make[1]: *** [/tmp/catalyst.ogRxWy/kcl_ioctl.o] Error 1
make: *** [_module_/tmp/catalyst.ogRxWy] Error 2
make: leaving directory « /usr/src/linux-2.6.35-ARCH »
+ return 1

Vi0L0 · 2010-09-18 17:15:32

@LeCrayonVert:
I was just about to write about it

Stock kernel was patched with very important patches that fix some critical security holes in kernel.

Please look at AUR's catalyst PKGBUILD and arch_compat.patch and implement it to 10.7.

Thomas Bächler at arch-dev-public mailing list wrote about kernel26 2.6.35.4-2 : "This build fixes two local root exploits! I only uploaded x86_64, as i686 is no affected"

I know only about one bug:
http://git.kernel.org/?p=linux/kernel/g … 7a79a81ea6
http://sota.gen.nz/compat2/

Custom x86_64 kernel's users must to patch own kernel with new patches.

Last edited by Vi0L0 (2010-09-18 17:17:22)

LeCrayonVert · 2010-09-18 20:10:52

Thx But how could I get the old PKGBUILD for 10.7 in order to include the patch ? (I guess it is not enought to just change pkgver in the 10.9 PKGBUILD) ?

edit : sorry, I've just forgot this link >> http://catalyst.apocalypsus.net/tarball/

Last edited by LeCrayonVert (2010-09-18 20:15:51)

Arch Linux

#2426 2010-09-17 04:54:26

Re: The AMD/ATI Bar & Grill

#2427 2010-09-17 08:22:51

Re: The AMD/ATI Bar & Grill

#2428 2010-09-17 08:31:47

Re: The AMD/ATI Bar & Grill

#2429 2010-09-17 09:57:43

Re: The AMD/ATI Bar & Grill

#2430 2010-09-17 10:07:19

Re: The AMD/ATI Bar & Grill

#2431 2010-09-17 11:31:22

Re: The AMD/ATI Bar & Grill

#2432 2010-09-17 11:39:50

Re: The AMD/ATI Bar & Grill

#2433 2010-09-17 11:47:29

Re: The AMD/ATI Bar & Grill

#2434 2010-09-17 11:56:12

Re: The AMD/ATI Bar & Grill

#2435 2010-09-17 12:37:28

Re: The AMD/ATI Bar & Grill

#2436 2010-09-17 12:53:22

Re: The AMD/ATI Bar & Grill

#2437 2010-09-17 13:20:33

Re: The AMD/ATI Bar & Grill

#2438 2010-09-17 13:31:16

Re: The AMD/ATI Bar & Grill

#2439 2010-09-17 15:52:51

Re: The AMD/ATI Bar & Grill

#2440 2010-09-17 20:23:11

Re: The AMD/ATI Bar & Grill

#2441 2010-09-17 22:11:09

Re: The AMD/ATI Bar & Grill

#2442 2010-09-18 05:32:35

Re: The AMD/ATI Bar & Grill

#2443 2010-09-18 10:30:57

Re: The AMD/ATI Bar & Grill

#2444 2010-09-18 10:32:58

Re: The AMD/ATI Bar & Grill

#2445 2010-09-18 11:25:36

Re: The AMD/ATI Bar & Grill

#2446 2010-09-18 14:43:23

Re: The AMD/ATI Bar & Grill

#2447 2010-09-18 15:53:10

Re: The AMD/ATI Bar & Grill

#2448 2010-09-18 17:02:45

Re: The AMD/ATI Bar & Grill

#2449 2010-09-18 17:15:32

Re: The AMD/ATI Bar & Grill

#2450 2010-09-18 20:10:52

Re: The AMD/ATI Bar & Grill

Board footer