The AMD/ATI Bar & Grill

Jannis · 2016-12-03 02:18:08

NoSuck wrote:

Jannis wrote:
But yeah, 2D isn't great with catalyst but it also isn't nearly that bad for me with a radeon hd 7870.
Is this statement in comparison to the video I posted?

Yes it was. Sorry I forgot to tag you.

Not sure if my setup is comparable, since I'm using gwenview for image viewing, but I see the full, next image virtually the moment I press the right arrow key. I'm also using an ssd, though, and I have no idea if caching is done by gwenview or some kde service in the background.

But from your video, my wild guess would be that your image viewer is using cpu for decoding instead of making use of gpu hardware acceleration (I might be wrong though; in which case somebody please correct me).

royallthefourth · 2016-12-08 12:57:00

How can I install ocl-icd and upgrade catalyst-test without uninstalling anything? The new version of catalyst-test should remove the conflict, but attempting to install it breaks a libcl dependency on imagemagick. Attempting to install ocl-icd, of course, reports a conflict with catalyst-test, which won't uninstall due to libgl dependencies in other packages.

It seems like I should be able to install both at the same time by saving the package files and using pacman -U, but that just gives me the same problem as trying to install one at a time.
Has anyone found a solution that doesn't involve uninstalling and reinstalling a whole bunch of peripheral packages?

Vi0L0 · 2016-12-08 15:33:04

just like I said on AUR:
pacman -Udd <path_to_catalyst-test_pkg>
then pacman -Syu ocl-icd

Sorry guys for such a long time waiting for icd fix, had some medical treatment and now i got some deadline on the work. Give me some more time, should start updating pkgs tomorrow

royallthefourth · 2016-12-08 21:44:11

Sorry, I missed those instructions somehow.

Thanks for all the work you do providing the packages.

Vi0L0 · 2016-12-11 19:15:23

I've just pushed [catalyst] and [catalyst-stable] repo updates.

New packages in [catalyst] and [catalyst-stable]:
- imagemagick-nocl - because I noted that default imagemagick built with opencl support is making CPU usage spikes, for sure it has negative impact on performance and energy consumption. This package provides imagemagick without opencl support.

New packages in [catalyst]:
- catalyst-control - I took /etc/ati/control file out of the catalyst-utils package and placed it in this package
- catalyst-control-old - it provides /etc/ati/control file from catalyst 15.9, it should help pre-GCN users who had problems with standard catalyst 15.12

also in catalyst-test@AUR pre-GCN users can switch to older control file by changing _old_control variable value to y in PKGBUILD

Last edited by Vi0L0 (2016-12-11 20:02:56)

Arhangel64 · 2016-12-13 21:43:17

Hey everyone. Wanted to aks if there's stil an issue about kwin gl2/3 renderer and catalyst drivers? When i select one of them effects switch off. Everything works fine with Xrender, but i miss wobbling windows)
My card is HD 6850M

Jannis · 2016-12-16 02:09:32

Arhangel64 wrote:

Hey everyone. Wanted to aks if there's stil an issue about kwin gl2/3 renderer and catalyst drivers? When i select one of them effects switch off. Everything works fine with Xrender, but i miss wobbling windows)
My card is HD 6850M

I'm using OpenGL 3.1 rendering backend in kwin_x11 on a radeon hd 7870 with catalyst-hook 15.12-6 from Vi0L0's repo and desktop effects work fine, including the beloved wobbly windows. I only get minor graphical glitches when resizing my konsole window from time to time (quickly worked around by switching focus to another app and back). I've set scale method to "crisp" and vsync to never, just in case that matters.

Edit: I'm using xorg-server 1.17.4-2 from the xorg117 repo.

Last edited by Jannis (2016-12-16 02:12:14)

Arhangel64 · 2016-12-17 10:19:58

Then it looks like I'm doing smth wrong - I've tried every posible configuration of vsync and scale method - but it doesn't help.
Also, I've noticed font corruption on some scrollable widgets, during scrolling, like list of kwin effects in kde settings, for example.
Is there anything else I can try to do about it?

Jannis · 2016-12-19 02:27:24

@Arhangel64 Only thing coming to mind is check if the kernel module actually got build and you have direct rendering working. If not, you might be missing your kernel's header package or something else needed to build it.

Other than that, I dunno. Maybe someone else has an idea. If all else failed, just use the open source drivers; they should be working as well if not better than catalyst for your card.

iuno · 2016-12-24 03:35:47

Is there anyone here using a 144 Hz WQHD monitor with free amd drivers?
Preferably looking for experience with Hawaii or other DCE8.x Chips (Kaveri, Bonaire, Kabini, Mullins).

Amarildo wrote:

I guess only thing left now is OpenCL on AMDGPU.

There is an AUR package for convenient usage of the (still proprietary) OpenCL driver to run onto the free stack: https://aur.archlinux.org/packages/opencl-amd/ The downside is that you have 2 libdrms hanging around and still a proprietary blob userspace driver.

I'm not sure if a free OCL driver for current and older GCN cards is coming anytime soon. They want to use a new, free implementation on top of ROC for newer cards.

Last edited by iuno (2016-12-24 03:43:22)

Arhangel64 · 2016-12-31 10:35:41

Nopes. Everything seems to be working, opengl version and stuff, looks like the critical part with build is fine. I've found a problem in Xorg logs, something was trying to find something in /usr/X11R6/lib64/modules/dri/fglrx_dri.so, so, i've created a symlink and... it broke the spinning progress on plasma taksbar - it started to spin superfast!))). Everything else looks okay, but i still can't make it work any other way than through XRender. Plasma wayland session also broke - it gives just a black screen after sddm login(

And what's wrong with xorg117 repos? I can't update the system - there's an html page with the message about domain name expiration in /var/lib/pacman/sync/xorg117.db.sig

Jannis · 2017-01-01 00:40:01

Arhangel64 wrote:

And what's wrong with xorg117 repos? I can't update the system - there's an html page with the message about domain name expiration in /var/lib/pacman/sync/xorg117.db.sig

What mirrors are you using? Unfortunately, rts-informatique.fr and wirephire.com are both down.

You're going to have to use:

[xorg117]
Server = http://mirror.hactar.xyz/Vi0L0/xorg117/$arch

[catalyst]
Server = http://mirror.hactar.xyz/Vi0L0/catalyst/$arch

You may have to do a pacman -Syyuu after changing pacman.conf.

Last edited by Jannis (2017-01-01 00:42:26)

mikaelbrun · 2017-01-27 09:22:04

I need some advice.

Now that Linux 4.9 is released, should I convert to AMDGPU, or should I stick with the xf86 driver?
And if I should go for AMDGPU, is this enabled by default? I have a Southern Island card, so it is experimental as I understand, and I am a gamer

nassi · 2017-01-27 16:46:49

mikaelbrun wrote:

I need some advice.
Now that Linux 4.9 is released, should I convert to AMDGPU, or should I stick with the xf86 driver?
And if I should go for AMDGPU, is this enabled by default? I have a Southern Island card, so it is experimental as I understand, and I am a gamer

I'm not sure about the exact performance differences between the different driver versions, but at least Sea Islands support is enabled by default it seems. Currently using an R9 290X and I went from linux-cik 4.9-2 to the current kernel version without issues. Judging by the wiki it seems as if Southern Islands should also work as long as you blacklist the radeon module: https://wiki.archlinux.org/index.php/AM … ands_cards.

Tried out Talos Principle and CS:GO and both seem to work about as well as before, with Talos Principle for example giving me smooth 60 FPS on high settings at 1080p resolution. Both games took a while to start though, and Talos Principle had some weird artifacting while it was starting up.

Also with the combination of the OpenCL parts from the AMDGPU-PRO (https://aur.archlinux.org/packages/opencl-amd/) and the current AMDGPU version my Folding@Home score seems to be at the levels it used to be when I used AMDGPU-PRO and the cik-kernel.

revberaldo · 2017-01-28 00:43:14

What about amdgpu-pro? The AUR package is still on version 16.30, and the latest one has just been released (with support to the R7 260M, three years after I bought my laptop!).

nassi · 2017-01-28 14:53:01

revberaldo wrote:

What about amdgpu-pro? The AUR package is still on version 16.30, and the latest one has just been released (with support to the R7 260M, three years after I bought my laptop!).

The last time I used AMDGPU-PRO it worked pretty much out of the box with the then current version of linux-cik from AUR. This was quite a few months back though, and since the AMDGPU-PRO AUR-package has been neglected for a while, it might require some tinkering. Only reason I used the AMDGPU-PRO version back then was to get Folding@Home working on my R9 290X, but nowadays even that is not needed thanks to the opencl-amd package.

The only noticeable difference between AMDGPU-PRO and AMDGPU for me currently is the fact that I can't use four monitors on the AMDGPU driver (http://pastebin.com/2Jz0BwGp, maybe I should report this somewhere?).

iuno · 2017-01-29 11:59:56

nassi wrote:

The only noticeable difference between AMDGPU-PRO and AMDGPU for me currently is the fact that I can't use four monitors on the AMDGPU driver (http://pastebin.com/2Jz0BwGp, maybe I should report this somewhere?).

Are you using xf86-video-amdgpu? Have you tried the generic modesetting driver?
You should definitely report bugs in general.

mikaelbrun wrote:

Now that Linux 4.9 is released, should I convert to AMDGPU, or should I stick with the xf86 driver?

My personal opinion:
If you rely on OpenCL or Vulkan: go for amdgpu, else: don't bother (stay with radeon).

Last edited by iuno (2017-01-29 12:11:51)

nassi · 2017-01-29 12:51:07

iuno wrote:

nassi wrote:
The only noticeable difference between AMDGPU-PRO and AMDGPU for me currently is the fact that I can't use four monitors on the AMDGPU driver (http://pastebin.com/2Jz0BwGp, maybe I should report this somewhere?).
Are you using xf86-video-amdgpu? Have you tried the generic modesetting driver?
You should definitely report bugs in general.

Currently using xf86-video-amdgpu. By the "generic modesetting driver" you mean I should see if the situation is different with xf86-video-amdgpu removed or..?

I guess this is where I should be reporting the bug: https://bugs.freedesktop.org/buglist.cg … esolution=--- ?

iuno · 2017-01-29 13:28:55

You don't have to remove the package, you can also add a x config file, e.g. my /etc/X11/xorg.conf.d/20-modesetting.conf

Section "Device"
        Identifier      "radeon"
        Driver          "modesetting"
        Option          "DRI"           "3"
        Option          "AccelMethod"   "glamor"
        Option          "PageFlip"      "true"
EndSection

But removing xf86-video-amdgpu works as well.

Yeah, look if someone has already opened a bug with the same problem: https://bugs.freedesktop.org/query.cgi? … ver/AMDgpu
If not, report using this form: https://bugs.freedesktop.org/enter_bug. … ver/AMDgpu

nassi · 2017-01-29 15:11:45

I did as you suggested, but apart from different naming scheme between drivers (DP-1, HDMI-1, DVI-D-1, DVI-D-2 on the radeon driver and DisplayPort-0, HDMI-A-0, DVI-D-0, DVI-D-1 on the AMDGPU driver), the observable effects were identical. Also seems that xrandr reports more possible resolutions on the radeon driver.

Upon booting both drivers organize the screens in a row in order of the ports mentioned above using the native resolutions of the screens, and in this situation the monitor on DVI-D-2 (DVI-D-1 on AMDGPU) remains blank. If I however change the resolution of DP-1 (DisplayPort-0) from it's native resolution of 1280x720 to 1920x1080 (it's a Sanyo PLV-Z5 projector) all of the displays work! But now if I try to disable the DP-1 (DisplayPort-0) connection in arandr all of the displays go blank. I can interact with the system though and perform a blind reboot. If I disable the projector while it's using the 1280x720 resolution the rest of the screens (including DVI-D-2 (DVI-D-1)) stay/turn on.

I'm fairly certain that even using the projector on its native resolution in combination with the rest of the screens worked on AMDGPU-PRO, but maybe it was for some reason automatically setting the projector resolution to 1920x1080. I don't know if it matters, but the projector is connected to the displayport via a passive DP-HDMI -adapter. In light of the new evidence, I'm no longer certain if this is a software bug, but a hardware limitation instead.

mikaelbrun · 2017-01-30 15:24:50

Thanks for your feedbacks, nassi and iuno!

I enabled AMDGPU just now, and now I have pink lines on the left side on both monitors. Anyone knows how to get rid of thoser?
They are 1-2px wide.

Last edited by mikaelbrun (2017-01-30 15:25:12)

nassi · 2017-01-30 15:34:45

mikaelbrun wrote:

Thanks for your feedbacks, nassi and iuno!
I enabled AMDGPU just now, and now I have pink lines on the left side on both monitors. Anyone knows how to get rid of thoser?
They are 1-2px wide.

Most likely this problem (fix included in the link): https://www.reddit.com/r/archlinux/comm … ource=link. For some reason it doesn't affect me, even though I'm using HDMI for my main display.

forlorn · 2017-01-30 20:30:01

Is there a way to disable the power saving with AMDGPU? My 470 seems to be stuck in mode 2 and I cannot figure out a way to get it above mode 2 even when I try to push the graphic limits.

radeonjet get core table
0: 300Mhz 
1: 466Mhz 
2: 751Mhz *
3: 1019Mhz 
4: 1074Mhz 
5: 1126Mhz 
6: 1169Mhz 
7: 1230Mhz

iuno · 2017-01-30 22:10:26

try (as root)

echo "high" > /sys/class/drm/card0/device/power_dpm_force_performance_level

Last edited by iuno (2017-01-30 22:10:43)

forlorn · 2017-02-01 17:48:16

iuno wrote:

try (as root)

echo "high" > /sys/class/drm/card0/device/power_dpm_force_performance_level

Works, thanks! I also figured out it was one application run through wine, fixed it by installing wine-staging-nine. There no manual input was needed, RX470 would go max MHz when needed.

Arch Linux

#7501 2016-12-03 02:18:08

Re: The AMD/ATI Bar & Grill

#7502 2016-12-08 12:57:00

Re: The AMD/ATI Bar & Grill

#7503 2016-12-08 15:33:04

Re: The AMD/ATI Bar & Grill

#7504 2016-12-08 21:44:11

Re: The AMD/ATI Bar & Grill

#7505 2016-12-11 19:15:23

Re: The AMD/ATI Bar & Grill

#7506 2016-12-13 21:43:17

Re: The AMD/ATI Bar & Grill

#7507 2016-12-16 02:09:32

Re: The AMD/ATI Bar & Grill

#7508 2016-12-17 10:19:58

Re: The AMD/ATI Bar & Grill

#7509 2016-12-19 02:27:24

Re: The AMD/ATI Bar & Grill

#7510 2016-12-24 03:35:47

Re: The AMD/ATI Bar & Grill

#7511 2016-12-31 10:35:41

Re: The AMD/ATI Bar & Grill

#7512 2017-01-01 00:40:01

Re: The AMD/ATI Bar & Grill

#7513 2017-01-27 09:22:04

Re: The AMD/ATI Bar & Grill

#7514 2017-01-27 16:46:49

Re: The AMD/ATI Bar & Grill

#7515 2017-01-28 00:43:14

Re: The AMD/ATI Bar & Grill

#7516 2017-01-28 14:53:01

Re: The AMD/ATI Bar & Grill

#7517 2017-01-29 11:59:56

Re: The AMD/ATI Bar & Grill

#7518 2017-01-29 12:51:07

Re: The AMD/ATI Bar & Grill

#7519 2017-01-29 13:28:55

Re: The AMD/ATI Bar & Grill

#7520 2017-01-29 15:11:45

Re: The AMD/ATI Bar & Grill

#7521 2017-01-30 15:24:50

Re: The AMD/ATI Bar & Grill

#7522 2017-01-30 15:34:45

Re: The AMD/ATI Bar & Grill

#7523 2017-01-30 20:30:01

Re: The AMD/ATI Bar & Grill

#7524 2017-01-30 22:10:26

Re: The AMD/ATI Bar & Grill

#7525 2017-02-01 17:48:16

Re: The AMD/ATI Bar & Grill

Board footer