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

splashy · 2013-07-25 20:38:24

I can play to League of Legend with video settings to the max (like in their website videos), the game is smooth no bugs, even the store is working(bnot the rune section thought).
And all of this thanks to the open-source radeon driver on my Hd4850! (with packages from extra).

banks1990 · 2013-07-27 08:24:05

I'm having a problem with my HD5850, where it is running at uvd clockspeed (Core:400Mhz/Mem:900Mhz) eventhough I set the profile to "high".

The max clockspeed should be 725Mhz/1000Mhz.

Currently running kernel 3.10.3 and mesa 9.1.5.

Will kernel 3.11 fix this problem? No flash or video was running when I check the clockspeed.

default engine clock: 400000 kHz
current engine clock: 399990 kHz
default memory clock: 900000 kHz
current memory clock: 900000 kHz
voltage: 1088 mV
PCIE lanes: 16

My apologies if this is not the place to post this question.

lordheavy · 2013-07-27 08:53:14

banks1990 wrote:

I'm having a problem with my HD5850, where it is running at uvd clockspeed (Core:400Mhz/Mem:900Mhz) eventhough I set the profile to "high".
The max clockspeed should be 725Mhz/1000Mhz.
Currently running kernel 3.10.3 and mesa 9.1.5.
Will kernel 3.11 fix this problem? No flash or video was running when I check the clockspeed.
default engine clock: 400000 kHz
current engine clock: 399990 kHz
default memory clock: 900000 kHz
current memory clock: 900000 kHz
voltage: 1088 mV
PCIE lanes: 16
My apologies if this is not the place to post this question.

AFAIK you need at least mesa 9.2 for uvd/vdpau.

banks1990 · 2013-07-27 09:25:55

That's the weird thing. No matter what profile I'm using I can't get it over 400Mhz/900Mhz. I can get it to 157Mhz/300Mhz when using "low". Even when I'm playing games it still stuck at that clockspeed. When I'm using fglrx it doesn't have this problem (though fglrx has other problems too, dat suspend/sleep). Stucked on this clockspeed on kernel 3.9 too. Not sure about 3.8 though, because Dota 2 wasn't released for linux at that time.

Reinstalling the xf86-video-ati also doesn't seemed to fix it at all. Anyone using the same card (HD5850) can comment on this? Or is mine an isolated problem?

agapito · 2013-07-27 11:19:03

agapito wrote:

Vdpau is very unstable in my system (HD 7950). When I resize the window of the video, my system totally crash with a sound loop. Tested in kernel 3.10 and 3.11.

I'm not 100% sure, but i think kernel 3.10.3 solved my problem

punkeroso · 2013-07-28 15:18:57

Hi everyone,
I've a laptop with an Ati HD5650 and, if I'm not wrong, with the last commit now sould be possible to use the open source driver without having high GPU temperature and, therefore, fan spinning like crazy.
What I would like to ask is if I just need to install linux-git (3.11.rc1) and xf86-video-ati-git in order to get what I've wrote above.
Thanks a lot.

Ledti · 2013-07-28 15:47:36

punkeroso wrote:

Hi everyone,
I've a laptop with an Ati HD5650 and, if I'm not wrong, with the last commit now sould be possible to use the open source driver without having high GPU temperature and, therefore, fan spinning like crazy.
What I would like to ask is if I just need to install linux-git (3.11.rc1) and xf86-video-ati-git in order to get what I've wrote above.
Thanks a lot.

Easy Way:

The easiest way would be to just use the [mesa-git] repo, which includes a patched kernel (linux-agd5f) and the latest mesa/ati packages (check the output of pacman -Sl mesa-git). You would also need to download the firmware (*.smc) file pertaining to your card from here, and copy it to the /usr/lib/firmware/radeon/ directory. The radeon page on the Gentoo wiki will tell you which specific firmware file you will need.

Note: the linux-firmware package will include that file in the future and it'll have to be manually removed so as to avoid a file conflict when that upgrade eventually takes place.

Finally, add radeon.dpm=1 to the kernel boot line in your grub/syslinux/gummiboot/etc file to enable the feature.

Manually:

For DPM you would need a 3.11-rc1 kernel at a minimum, though I would use at least 3.11-rc2 which includes more DPM fixes. You would also need xf86-video-ati-git and possibly mesa 9.2 (devel), but I'm not entirely sure. Also, you would need the *.smc file and you'd have to add radeon.dpm=1 to your kernel boot line as mentioned above.

Last edited by Ledti (2013-07-31 22:04:42)

s3kt0r · 2013-07-28 19:05:26

I've encountered a few things, so to start the GPU is radeon 7970 (SI/TAHITI).

There were a few changes on drivers/gpu/drm/radeon/si.c <https://git.kernel.org/cgit/linux/kerne … 5baba90f3b> that linux-firmware-git didn't update to.

When trying to boot 3.11-rc1 and above, <http://ix.io/6UI> happens and reboot is needed with 'radeon.modeset=0' to work with tty.

Jul 28 18:36:02 valhalla kernel: si_cp: Failed to load firmware "radeon/TAHITI_smc.bin"
Jul 28 18:36:02 valhalla kernel: [drm:si_startup] *ERROR* Failed to load firmware!
Jul 28 18:36:02 valhalla kernel: radeon 0000:01:00.0: disabling GPU acceleration
Jul 28 18:36:02 valhalla kernel: radeon 0000:01:00.0: ffff880128ac6400 unpin not necessary
Jul 28 18:36:02 valhalla kernel: radeon 0000:01:00.0: ffff880128ac6400 unpin not necessary
Jul 28 18:36:02 valhalla kernel: [drm:si_init] *ERROR* radeon: MC ucode required for NI+.
Jul 28 18:36:02 valhalla kernel: radeon 0000:01:00.0: Fatal error during GPU init

On stable, that didn't exist <https://git.kernel.org/cgit/linux/kerne … c?id=v3.10>

I searched around a bit and I found a LKML message that mentions it <https://lkml.org/lkml/2013/7/21/40>

So I guess we're still waiting?

edit: I did find this though <http://people.freedesktop.org/~agd5f/radeon_ucode/>, it has that TAHITI_smc.bin. Would it be wrong for me to...

Last edited by s3kt0r (2013-07-28 19:09:18)

lordheavy · 2013-07-28 19:12:14

s3kt0r wrote:

I've encountered a few things, so to start the GPU is radeon 7970 (SI/TAHITI).
There were a few changes on drivers/gpu/drm/radeon/si.c <https://git.kernel.org/cgit/linux/kerne … 5baba90f3b> that linux-firmware-git didn't update to.
When trying to boot 3.11-rc1 and above, <http://ix.io/6UI> happens and reboot is needed with 'radeon.modeset=0' to work with tty.
Jul 28 18:36:02 valhalla kernel: si_cp: Failed to load firmware "radeon/TAHITI_smc.bin"
Jul 28 18:36:02 valhalla kernel: [drm:si_startup] *ERROR* Failed to load firmware!
Jul 28 18:36:02 valhalla kernel: radeon 0000:01:00.0: disabling GPU acceleration
Jul 28 18:36:02 valhalla kernel: radeon 0000:01:00.0: ffff880128ac6400 unpin not necessary
Jul 28 18:36:02 valhalla kernel: radeon 0000:01:00.0: ffff880128ac6400 unpin not necessary
Jul 28 18:36:02 valhalla kernel: [drm:si_init] *ERROR* radeon: MC ucode required for NI+.
Jul 28 18:36:02 valhalla kernel: radeon 0000:01:00.0: Fatal error during GPU init
On stable, that didn't exist <https://git.kernel.org/cgit/linux/kerne … c?id=v3.10>
I searched around a bit and I found a LKML message that mentions it <https://lkml.org/lkml/2013/7/21/40>
So I guess we're still waiting?
edit: I did find this though <http://people.freedesktop.org/~agd5f/radeon_ucode/>, it has that TAHITI_smc.bin. Would it be wrong for me to...

You can find the firmware here: http://people.freedesktop.org/~agd5f/radeon_ucode/

s3kt0r · 2013-07-28 19:15:05

lordheavy wrote:

s3kt0r wrote:
I've encountered a few things, so to start the GPU is radeon 7970 (SI/TAHITI).
There were a few changes on drivers/gpu/drm/radeon/si.c <https://git.kernel.org/cgit/linux/kerne … 5baba90f3b> that linux-firmware-git didn't update to.
When trying to boot 3.11-rc1 and above, <http://ix.io/6UI> happens and reboot is needed with 'radeon.modeset=0' to work with tty.
Jul 28 18:36:02 valhalla kernel: si_cp: Failed to load firmware "radeon/TAHITI_smc.bin"
Jul 28 18:36:02 valhalla kernel: [drm:si_startup] *ERROR* Failed to load firmware!
Jul 28 18:36:02 valhalla kernel: radeon 0000:01:00.0: disabling GPU acceleration
Jul 28 18:36:02 valhalla kernel: radeon 0000:01:00.0: ffff880128ac6400 unpin not necessary
Jul 28 18:36:02 valhalla kernel: radeon 0000:01:00.0: ffff880128ac6400 unpin not necessary
Jul 28 18:36:02 valhalla kernel: [drm:si_init] *ERROR* radeon: MC ucode required for NI+.
Jul 28 18:36:02 valhalla kernel: radeon 0000:01:00.0: Fatal error during GPU init
On stable, that didn't exist <https://git.kernel.org/cgit/linux/kerne … c?id=v3.10>
I searched around a bit and I found a LKML message that mentions it <https://lkml.org/lkml/2013/7/21/40>
So I guess we're still waiting?
edit: I did find this though <http://people.freedesktop.org/~agd5f/radeon_ucode/>, it has that TAHITI_smc.bin. Would it be wrong for me to...
You can find the firmware here: http://people.freedesktop.org/~agd5f/radeon_ucode/

Yeah, i did find it. Have a pretty good idea about how, but I shall search on how to incorporate it to aur/linux-firmware-git.

agapito · 2013-07-30 23:59:24

linux-agd5f 3.10.0-14 fixed the BSOD when radeon module is loaded.

cedeel · 2013-07-31 18:43:59

agapito wrote:

linux-agd5f 3.10.0-14 fixed the BSOD when radeon module is loaded.

Since when does Arch have BSOD?

marazmista · 2013-07-31 23:06:02

I made little app for display radeon basic parameters (clocks, volts, temp). Also can change power profile. Works on my system, and I wonder if it will work on somebody else's. Maybe you find it useful.

Last edited by marazmista (2013-08-15 17:22:31)

crshbndct · 2013-08-02 01:23:19

marazmista wrote:

I made little app for display radeon basic parameters (clocks, volts, temp). Also can change power profile. Works on my system, and I wonder if it will work on somebody else's. Maybe you find it useful.
screenshot
AUR

Does this work with the new DPM stuff? Any chance you might be interested in getting it merged into driconf somehow?

ky · 2013-08-02 07:35:45

marazmista wrote:

I made little app for display radeon basic parameters (clocks, volts, temp). Also can change power profile. Works on my system, and I wonder if it will work on somebody else's. Maybe you find it useful.
screenshot
AUR

It didnt really work on mine, im running mesa-git and 3.10 from core, and it crashed and killed everything

Stunts · 2013-08-02 08:19:52

DPM is now working for me (Mobility Radeon HD3650) with the latest patches from agd5f tree (drm-fixes-3.11 branch). Cheers to that.
I will soon update my Radeon-tray (https://github.com/StuntsPT/Radeon-tray) program to work with DPM.
It's sort of like marazmista's program, but instead of a window you control everything from the systray.
I'll post here when the changes are ready.

agapito · 2013-08-02 10:04:46

cedeel wrote:

agapito wrote:
linux-agd5f 3.10.0-14 fixed the BSOD when radeon module is loaded.
Since when does Arch have BSOD?

Black Screen of Death. When radeon module was loaded, screen went totally black and system was totally dead.

Latest kernels from mesa-git repo are OK.

marazmista · 2013-08-02 13:52:42

crshbndct wrote:

Does this work with the new DPM stuff? Any chance you might be interested in getting it merged into driconf somehow?

Yes, it should work with DPM. I'm running 3.11-rc3 and it's ok.
Program still need some debugging (for different system files), and it is radeon-only. Maybe in the future if driconf maintainer also would be interested.

ky wrote:

It didnt really work on mine, im running mesa-git and 3.10 from core, and it crashed and killed everything

Hmm, could you check if you have these files in system? Also knowing what is inside could be helpful (especially last one).

/sys/class/drm/card0/device/power_method
/sys/class/drm/card0/device/power_profile
/sys/kernel/debug/dri/0/radeon_pm_info

You can also try run it from console and check if app throw something.

xorgx3 · 2013-08-02 14:35:11

marazmista wrote:

I made little app for display radeon basic parameters (clocks, volts, temp). Also can change power profile. Works on my system, and I wonder if it will work on somebody else's. Maybe you find it useful.
screenshot
AUR

I think you need to change PKGBUILD: qmake-qt4 instead of qmake.

Last edited by xorgx3 (2013-08-02 14:35:25)

marazmista · 2013-08-03 08:40:34

xorgx3 wrote:

I think you need to change PKGBUILD: qmake-qt4 instead of qmake.

Thanks, fixed.

I pushed some changes in git, now it should work better.

siavashserver · 2013-08-03 22:02:14

It works great for me on RadeonHD 4890. Now it's smoother and even works cooler than old dynpm forced at low power profile.

Thank you very much for the [mesa-git] repo!

Aldur · 2013-08-05 09:14:16

With my Radeon HD 6870 I got a BSOD with linux-a5gdf, I had to remove the radeon module from mkinitcpio.conf to make it work (if somebody else is out there who still had that line in there). When I use radeon.dpm=1 I get massive screen corruptions (flickering, short freezes, frozen pixels, blinking pixels, etc...).

I guess it's still a bit buggy for the BARTS series? Because I've heard of similar issues with the Radeon HD 68xx series before and that they got fixed in a later RC (2 or 3)?

To provide more information I'm currently compiling linux-mainline from aur and linux-firmware-git which seems to be more recent. I'll report back if things have improved.

Edit:
I can report back that dpm is still broken for me, even with linux-mainline (rc3). The kernel boots just fine without dpm but as soon as I add radeon.dpm=1 I get a BSOD, while with the patched 3.10 kernel I get the above corruptions.

Edit 2:
Phoronix has tested DPM on the 6870 already and it seemed to work flawlessly. So it must be something in my setup, I just can't figure out what it is.

Edit 3:
From what I've gathered this might indeed be a regression. I'll report it on the Phoronix forums, too.

Last edited by Aldur (2013-08-05 11:05:04)

Ledti · 2013-08-05 18:23:37

Aldur wrote:

With my Radeon HD 6870 I got a BSOD with linux-a5gdf, I had to remove the radeon module from mkinitcpio.conf to make it work (if somebody else is out there who still had that line in there). When I use radeon.dpm=1 I get massive screen corruptions (flickering, short freezes, frozen pixels, blinking pixels, etc...).
I guess it's still a bit buggy for the BARTS series? Because I've heard of similar issues with the Radeon HD 68xx series before and that they got fixed in a later RC (2 or 3)?
To provide more information I'm currently compiling linux-mainline from aur and linux-firmware-git which seems to be more recent. I'll report back if things have improved.
Edit:
I can report back that dpm is still broken for me, even with linux-mainline (rc3). The kernel boots just fine without dpm but as soon as I add radeon.dpm=1 I get a BSOD, while with the patched 3.10 kernel I get the above corruptions.
Edit 2:
Phoronix has tested DPM on the 6870 already and it seemed to work flawlessly. So it must be something in my setup, I just can't figure out what it is.
Edit 3:
From what I've gathered this might indeed be a regression. I'll report it on the Phoronix forums, too.

I'm running 3.11-rc4 with the same card and DPM is working fine for me. Though there is an occasional hard lock. As in... everything completely freezes and I have to push the reset button to forcibly restart my computer. I'm not sure if it's related to DPM or the fact that I'm using a prepatch kernel with the latest mesa-git and ATI packages...

In any case, have you tried 3.11-rc4 yet? There were some additional radeon fixes included with it that might resolve your issue. Though in my case 3.11-rc3 works just as well.

Aldur · 2013-08-05 19:10:24

Thanks for your reply!

That it works for you means that it must really be something on my end. The fixes in RC4 seem to be related to Southern Island but I will give it a try anyways, it can't hurt.
I really can't think of anything that I've forgotten: I removed Plymouth (just for the sake of getting anything out of the way that might - even remotely - interfere), the radeon module from mkinitcpio, installed latest mesa/ati-dri from mesa-git, downloaded the firmware (even tried linux-firmware-git) and tried the patched 3.10 kernel from mesa-git and 3.11rc3 from linux-mainline.
Without dpm everything works great, even video decoding/rendering in Firefox which I could never get to work before.

As soon as I activate dpm I get massive screen corruption, occasional freezes which recover some time and after a while a total lock up. In the meantime I can find several "panicked" messages in dmesg.

Did you - maybe - take any additional steps?

Last edited by Aldur (2013-08-05 19:10:53)

Ledti · 2013-08-05 19:58:19

Nope, but there was a previous issue exactly like yours that was fixed a few weeks ago. Here's a link. Those patches were already incorporated into the kernel (rc2+), but you should consider updating to that bug report with your issue. The ATI developers don't always follow those Phoronix threads, so you might be forced to wait a while for a reply otherwise.

Arch Linux

#2501 2013-07-25 20:38:24

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

#2502 2013-07-27 08:24:05

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

#2503 2013-07-27 08:53:14

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

#2504 2013-07-27 09:25:55

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

#2505 2013-07-27 11:19:03

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

#2506 2013-07-28 15:18:57

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

#2507 2013-07-28 15:47:36

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

#2508 2013-07-28 19:05:26

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

#2509 2013-07-28 19:12:14

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

#2510 2013-07-28 19:15:05

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

#2511 2013-07-30 23:59:24

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

#2512 2013-07-31 18:43:59

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

#2513 2013-07-31 23:06:02

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

#2514 2013-08-02 01:23:19

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

#2515 2013-08-02 07:35:45

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

#2516 2013-08-02 08:19:52

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

#2517 2013-08-02 10:04:46

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

#2518 2013-08-02 13:52:42

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

#2519 2013-08-02 14:35:11

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

#2520 2013-08-03 08:40:34

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

#2521 2013-08-03 22:02:14

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

#2522 2013-08-05 09:14:16

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

#2523 2013-08-05 18:23:37

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

#2524 2013-08-05 19:10:24

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

#2525 2013-08-05 19:58:19

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

Board footer