The AMD/ATI Bar & Grill

flamelab · 2010-02-01 08:37:43

OF course there is Turbo, and it IS enabled. It just isn't reported on cpufreq-info. If you use i7z you can see that not only Turbo is enabled, but that the i5/i7 CPUs don't require any software in order to have their internal (BIOS) cpufreq features enabled in the operating system.

hunterthomson · 2010-02-01 09:30:05

Ya, I was unsure. Something regarding it is fixed in 2.6.33. I guess it must be the reporting of it. Ya, the Intel guy as saying that it Turbo Boosts independent of the OS. Maybe I should start a new thread about the i7 and how to take full advantage of it. This thread was a bad place to state it I guess... Off topic.

But supposedly the Nehalem chips have a built in function to do the ZFS redundancy checking. The Solaris dispatcher will put info needed for each thread in the memory closest to the core that will compute the thread. Solaris is suppose to organizes the threads to try and put them all on as few cores as posable so as to enable Turbo Mode as often as posable. bla bla you can Google OpenSolaris Nehalem for it.

However, I can't find anything saying Linux is making use of the Nehalem. If Gentoo is saying to compile for the Core2 then I guess nothing is making use of the new instructions in the Nehalem CPU's. Benchmarks are showing Linux running like crap on the i7's. Like Linux is running faster in a VM ! <However.. when I say Linux I mean Ubuntu with it's old kernel and Old everything>

----------
OK I started a thread about this so if anyone is interested look here
http://bbs.archlinux.org/viewtopic.php?pid=699533

Last edited by hunterthomson (2010-02-01 10:30:39)

timong · 2010-02-01 15:20:58

timong wrote:

==
EDIT3:
Interestingly, if I start a fresh terminal and do 'su - username' the flickering is not present in the applications started from the given terminal.

After trying different settings and restarts and stuff, I concluded that this could be just as much a compiz problem as an fglrx related thing. Starting with a fresh terminal it works great, so I have my workaround. The compiz desktop seems to work okay after resume regardless of the other problems. I'm suspecting it's related to something like a vsync setting because when I have this trouble with flickering glxgears it writes that it's going to use vsync. When I start a new terminal and do a 'su - user' it does not output the vsync related message to the console when running glxgears.
I will try to dig up some info how libGL vsync can be switched...
glxgears output when flickering
Running synchronized to the vertical refresh.  The framerate should be
approximately the same as the monitor refresh rate.

Well, tested it over and over again. Flickering issues are mentioned at a lot of google hits. So it's not so weird I run to it.

What I find rare is that it can be fixed with doing a 'su - user' and start the 3D app after that...everyone else with this thing? I hope this may help some compiz-fglrx users.

Vi0L0 · 2010-02-03 23:35:03

The good news is that i was just talking to Anssi Hannula - mainainer of mandriva's catalyst package and hes got some pretty 2.6.33.patch right here http://svn.mandriva.com/cgi-bin/viewvc. … 6.33.patch

Now...
Can someone please help me with this patch? Patch looks reasonable. I've diged in this for some time, but i cant figure out whats wrong... Patch and patching output can be found here: http://hpaste.org/fastcgi/hpaste.fcgi/v … 054#a17054

i have used this command inside PKGBUILD to patch: patch -Np0 -i ${srcdir}/fglrx-2.6.33.patch || return 1

Edit: I can sure you that this patch is workin - i have manualy fixed not patched by patch_command files and catalyst 10.1 is compiled on 2.6.33-rc6-rc succesfully! I cannot test it now, but it must to work. So... it would be very nice if someone can help me with patch_command otherwise i will have to upload all manualy patched src files with some install instructions Anyhow i will post prebuilded packages for 2.6.33-ARCH as soon as it will be in [core] ([testing] maybe)
Cheers!

Last edited by Vi0L0 (2010-02-04 00:32:26)

Vi0L0 · 2010-02-04 12:43:04

AUR's catalyst 'I will do math' 10.1 should now support 2.6.33 kernel

Edit: tested on 2.6.33-rc6 hd4850 x86_64

Last edited by Vi0L0 (2010-02-04 13:40:26)

timong · 2010-02-05 18:37:08

Vi0L0 wrote:

AUR's catalyst 'I will do math' 10.1 should now support 2.6.33 kernel
Edit: tested on 2.6.33-rc6 hd4850 x86_64

This is really great news! I was wondering how it is affected by the KMS thing. Looks like it will work out fine when arch gets its 2.6.33!

ChemBro · 2010-02-05 21:26:25

timong wrote:

This is really great news! I was wondering how it is affected by the KMS thing. Looks like it will work out fine when arch gets its 2.6.33!

KMS? With closed-source-drivers?

timong · 2010-02-06 08:34:04

ChemBro wrote:

timong wrote:
This is really great news! I was wondering how it is affected by the KMS thing. Looks like it will work out fine when arch gets its 2.6.33!
KMS? With closed-source-drivers?

No, i was just wondering how the code changes related to KMS and UMS in the kernel source will affect the catalyst released for the 2.6.22. And I'm happy that the kernel didnt change so much that the catalyst became unusable.

Perry3D · 2010-02-06 09:33:22

timong wrote:

No, i was just wondering how the code changes related to KMS and UMS in the kernel source will affect the catalyst released for the 2.6.22. And I'm happy that the kernel didnt change so much that the catalyst became unusable.

There are no change as you have to disable the radeon module (!radeon in the modules array).

Cory · 2010-02-06 19:11:31

Yesterday probe XvBA in vlc-git-va 20100125-1 and it worked great I am very impressed with the result

timong · 2010-02-06 21:02:13

Cory wrote:

Yesterday probe XvBA in vlc-git-va 20100125-1 and it worked great I am very impressed with the result

Is this with catalyst? I cannot seem to install that with yaourt, ati-dri is needed, which needs libgl which conflicts with catalyst. Are you installing it like this: install libgl, install ati-dri, force-remove libgl and install catalyst again, and after build the vlc-git-va?

Cory · 2010-02-06 23:52:55

Yes it works with catalyst. I only do
yaourt -S libva-sds
yaourt -S vdpau-video
yaourt -S ffmpeg-svn (with the option --enable-vdpau)
yaourt -S vlc-git-va

Conflicts vlc-git-va
conflicts=('vlc' 'vlc-svn' 'vlc-nightly')

i don't know where is your problem. Before doing that already XvBA had working (catalyst 9.10)

Last edited by Cory (2010-02-06 23:55:05)

timong · 2010-02-07 09:18:39

Thanks. Seems like i had ati-dri installed and yaourt told me its missing libgl. Apparently ati-dri is not needed anyway, so removing it helped...installing it now, if i succeed.

flamelab · 2010-02-07 09:32:56

You use ati-dri only with xf86-video-ati, not catalyst.

timong · 2010-02-07 12:28:07

Cory wrote:

Yesterday probe XvBA in vlc-git-va 20100125-1 and it worked great I am very impressed with the result

I've tried it but I'm not sure how to configure this thing. Installed things as you suggested, things went right.
Is there any configuration I need to do in xorg.conf? Also which video output should I select in vlc? (Tried XV but it seems to eat CPU with 1080p just as with SDL)

Screen capture of what i tried:
http://imagebin.org/83715

Okay, think i found it, experimental support for codec accel...trying it

EDIT:
Tried the advanced settings mode and fiddled with it, but still cannot get it GPU accelerated. Any hints about where to look for settings info?

Last edited by timong (2010-02-07 12:47:31)

timong · 2010-02-07 16:10:29

Okay, got it "working" with mplayer-vaapi-pulse and installing xvba-video package. vainfo tells nicely that things are okay.

mplayer -vo vaapi -va vaapi xy.mp4 gives my the following garbled result (1080p) on my HD5650 Mobility:

http://imagebin.org/83730

I have disable composite extension and edited my device section in xorg.conf:

Section "Device"
        Identifier  "Card0"
    Option "AccelMethod" "EXA"
    Option "OpenGLOverlay" "1"
    Option "OpenGLOverlay" "on"
    Option "XAANoOffscreenPixmaps" "true"
        Driver      "fglrx"
        VendorName  "ATI Technologies Inc"
        BoardName   "Radeon Mobility"
        BusID       "PCI:2:0:0"
EndSection

Any suggestions what could be causing the trouble? I will try 720p too now.
BTW, CPU idles while playing it, so it's theoretically working now, and the garbled output is related to the movie too, so not just random.

720p h264: http://imagebin.org/83734
garbled...

hw-accel demos (vaa_h264 --glx) result garbled: http://imagebin.org/83738

Last edited by timong (2010-02-07 16:33:22)

timong · 2010-02-07 17:13:55

okay, ignore it. Other folks with 5xxx (evergreen) have similar problems (@phoronix: http://www.phoronix.com/forums/showthre … 83&page=50 , http://www.phoronix.com/forums/showpost … tcount=498 .) Looks like i will have to wait more, but somehow it might be near future, if i'm optimistic.

By any chance, Cory, are you using an 5xxx successfully with vlc vaapi??

Last edited by timong (2010-02-07 17:15:52)

MAroco · 2010-02-07 17:22:08

@ Cory sorry if its slightly off topic but I was under the impression that vdpau is for NVidia only and Atis equivalent would be XvBA? So why install vdpau-video and enable vdau for ffmpeg for the catalyst driver? Am I missing something? Just curious...

Cory · 2010-02-07 20:15:24

@ timong No my card is a HD4670
@ MAroco Because that is the only way i can get work vlc-git-va. If don't active vdpau don't compile ok.

idx · 2010-02-07 20:24:22

timong wrote:

okay, ignore it. Other folks with 5xxx (evergreen) have similar problems (@phoronix: http://www.phoronix.com/forums/showthre … 83&page=50 , http://www.phoronix.com/forums/showpost … tcount=498 .) Looks like i will have to wait more, but somehow it might be near future, if i'm optimistic.

Thanks for those links. XvBa does the same thing on my evergreen card, and I hadn't worked out whether I'd messed something up or if it was driver-related. Guess we're just going to have to wait a little longer...

MAroco · 2010-02-07 21:34:53

@Cory: Ah ok, I see so vdpau = nvidia still applies
Thanks for clarifying.

Infin1ty · 2010-02-09 14:20:08

Well, i'm using the ready package with the maximized catalyst fix (the fedora dont back fill patch) but this package has a version of xorg that causes memory leak (i remember it from ubuntu, the other version 1.6.4+ is better, but it seems as i have problems compiling 1.6.4 myself (i changed the PKGBUILd from abs), anyone knows what's the best way of compiling the old xorg-server myself? anyone can post the PKGBUILD of the already compiled old package in here?
thanks.

Vi0L0 · 2010-02-09 17:34:50

Infin1ty wrote:

Well, i'm using the ready package with the maximized catalyst fix (the fedora dont back fill patch) but this package has a version of xorg that causes memory leak (i remember it from ubuntu, the other version 1.6.4+ is better, but it seems as i have problems compiling 1.6.4 myself (i changed the PKGBUILd from abs), anyone knows what's the best way of compiling the old xorg-server myself? anyone can post the PKGBUILD of the already compiled old package in here?
thanks.

Have you red my note on http://aur.archlinux.org/packages.php?ID=26687 ? which tells:

@ moimael: yes, xserver will work with libs from xorg 7.5, but it's impossible to compile xserver 1.6.x if you are using newest Xorg packages. libxext is first lib that you must to downgrade if you want to compile xserver 1.6.x, but not last, you must to downgrade a whole bunch of libs. Also the newest xorg packages got different files inside, for example file a was in lib package called b, now file a is in lib package c, and in lib package b there are some other files. Thats why it's preety hard to tell how much packages you must to downgrade, and which ones exactly.
But I could be wrong, maybe theres some simple way

Have you downgraded "tons" of packages before compiling?

Last edited by Vi0L0 (2010-02-09 17:36:07)

Warper4 · 2010-02-09 18:25:36

Might someone provide the tarball of the package to me catalyst-opencl-9.12 or catalyst-9.12 and patch for kernel 2.6.32?

I need these drivers, since the 10.1 me are not OK with X-Plane

Thank's in advance

Vi0L0 · 2010-02-09 20:37:30

Warper4 wrote:

Might someone provide the tarball of the package to me catalyst-opencl-9.12 or catalyst-9.12 and patch for kernel 2.6.32?
I need these drivers, since the 10.1 me are not OK with X-Plane
Thank's in advance

http://dl.dropbox.com/u/4321633/catalyst-opencl.tar.gz

http://dl.dropbox.com/u/4321633/catalyst_9.12.tar.gz

Arch Linux

#1226 2010-02-01 08:37:43

Re: The AMD/ATI Bar & Grill

#1227 2010-02-01 09:30:05

Re: The AMD/ATI Bar & Grill

#1228 2010-02-01 15:20:58

Re: The AMD/ATI Bar & Grill

#1229 2010-02-03 23:35:03

Re: The AMD/ATI Bar & Grill

#1230 2010-02-04 12:43:04

Re: The AMD/ATI Bar & Grill

#1231 2010-02-05 18:37:08

Re: The AMD/ATI Bar & Grill

#1232 2010-02-05 21:26:25

Re: The AMD/ATI Bar & Grill

#1233 2010-02-06 08:34:04

Re: The AMD/ATI Bar & Grill

#1234 2010-02-06 09:33:22

Re: The AMD/ATI Bar & Grill

#1235 2010-02-06 19:11:31

Re: The AMD/ATI Bar & Grill

#1236 2010-02-06 21:02:13

Re: The AMD/ATI Bar & Grill

#1237 2010-02-06 23:52:55

Re: The AMD/ATI Bar & Grill

#1238 2010-02-07 09:18:39

Re: The AMD/ATI Bar & Grill

#1239 2010-02-07 09:32:56

Re: The AMD/ATI Bar & Grill

#1240 2010-02-07 12:28:07

Re: The AMD/ATI Bar & Grill

#1241 2010-02-07 16:10:29

Re: The AMD/ATI Bar & Grill

#1242 2010-02-07 17:13:55

Re: The AMD/ATI Bar & Grill

#1243 2010-02-07 17:22:08

Re: The AMD/ATI Bar & Grill

#1244 2010-02-07 20:15:24

Re: The AMD/ATI Bar & Grill

#1245 2010-02-07 20:24:22

Re: The AMD/ATI Bar & Grill

#1246 2010-02-07 21:34:53

Re: The AMD/ATI Bar & Grill

#1247 2010-02-09 14:20:08

Re: The AMD/ATI Bar & Grill

#1248 2010-02-09 17:34:50

Re: The AMD/ATI Bar & Grill

#1249 2010-02-09 18:25:36

Re: The AMD/ATI Bar & Grill

#1250 2010-02-09 20:37:30

Re: The AMD/ATI Bar & Grill

Board footer