The AMD/ATI Bar & Grill

diesel1 · 2016-10-24 23:12:18

Hi ViOLO,

I tried your kernel-git packages but pacman fails with a PGP signature error even though I said yes to import your key.

Any ideas?

Diesel1

Vi0L0 · 2016-10-25 16:28:06

Did you download *sig files and paste it next to packages? Maybe something with read permissions?
I just downloaded it from google, checked signatures and Kleopatra said it's fine, trusted.

Edit: also have in mind that it was compiled before Dirty Cow was known https://dirtycow.ninja/

Last edited by Vi0L0 (2016-10-25 17:56:12)

diesel1 · 2016-10-26 19:57:14

Thanks ViOLO,

I do have the sig packages with the pkg files so I am not sure why it is failing.

Perhaps I will wait until the next compile...

Diesel1.

scobi · 2016-10-29 14:50:01

hey guys!
im just a newbie in Arch and i have readed entire wiki but didnt found answer to my question
i have HP laptop with Radeon R5 as integrated GPU and Radeon R7 M440 as discrete GPU
Should i downgrade xorg to 1.17 and install fglrx? (actually i did it but only visible GPU in this case is R5 (( )
or should i install AMDGPU (Pro, something else...)?
Thank you in advance!

ergya · 2016-10-29 19:53:37

Hi everyone, pacman had been complaining about not being able to upgrade the system because of unsatisfied dependencies (linux<4.8). It happened in the past, no biggies. I read this forum for updates. What surprised me that nobody complained about this. Today I looked into it and I found there might have been some confusion between packer and Vi0L0's repository and AUR packages. I checked AUR and noticed this dependency should not be a problem any more. I Commented out Vi0L0's repository and "forced" to reinstall catalyst-hook (only that!), rebooted and SDDM could not be started anymore. I looked into the logs and found xorg core dump. So here is what I upgraded from:

catalyst-hook-15.12-4-x86_64.pkg.tar.xz

to (yes, from 15.12-4 to 15.12-3, pacman also had been complaining about newer local version then remote):

catalyst-hook-15.12-3-x86_64.pkg.tar.xz

And the xorg error:

okt 29 21:29:11 finchcomputer sddm[359]: Display server starting...
okt 29 21:29:11 finchcomputer sddm[359]: Running: /usr/bin/X -nolisten tcp -auth /var/run/sddm/{b2a46f9c-66fe-4b24-95b7-97426069884e} -background none -noreset -displayfd 21 vt1
okt 29 21:29:11 finchcomputer systemd[1]: Created slice User Slice of sddm.
okt 29 21:29:11 finchcomputer systemd[1]: Starting User Manager for UID 995...
okt 29 21:29:11 finchcomputer systemd-logind[343]: New session c1 of user sddm.
okt 29 21:29:11 finchcomputer systemd[1]: Started Session c1 of user sddm.
okt 29 21:29:11 finchcomputer systemd[400]: pam_unix(systemd-user:session): session opened for user sddm by (uid=0)
okt 29 21:29:11 finchcomputer systemd-coredump[394]: Process 386 (Xorg) of user 0 dumped core.

                                                     Stack trace of thread 386:
                                                     #0  0x00007fa48dcc304f raise (libc.so.6)
                                                     #1  0x00007fa48dcc447a abort (libc.so.6)
                                                     #2  0x00000000005992fe OsAbort (Xorg)
                                                     #3  0x000000000047a5ac ddxGiveUp (Xorg)
                                                     #4  0x000000000059ef62 n/a (Xorg)
                                                     #5  0x000000000059fd9d FatalError (Xorg)
                                                     #6  0x0000000000596c4e n/a (Xorg)
                                                     #7  0x00007fa48dcc30b0 __restore_rt (libc.so.6)
                                                     #8  0x00007fa48a83be72 swlDrmFreeSurfaces (fglrx_drv.so)
                                                     #9  0x00007fa48a800ddd xdl_xs117_atiddxDriCloseScreen (fglrx_drv.so)
                                                     #10 0x00007fa48a800478 xdl_xs117_atiddxDriScreenInit (fglrx_drv.so)
                                                     #11 0x00007fa48a7f9f8b xdl_xs117_atiddxScreenInit (fglrx_drv.so)
                                                     #12 0x00000000004392d1 AddScreen (Xorg)
                                                     #13 0x000000000047bcea InitOutput (Xorg)
                                                     #14 0x000000000043ce6a n/a (Xorg)
                                                     #15 0x00007fa48dcb0291 __libc_start_main (libc.so.6)
                                                     #16 0x0000000000427489 _start (Xorg)
okt 29 21:29:11 finchcomputer systemd[400]: Reached target Paths.
okt 29 21:29:11 finchcomputer systemd[400]: Starting D-Bus User Message Bus Socket.
okt 29 21:29:11 finchcomputer systemd[400]: Reached target Timers.

I tried to upgrade the whole system, going up to kernel 4.8 from 4.7. No luck. As of now I downgraded (yes, to 15.12-4 from 15.12-3) the relevant packages to this to make it work again:

catalyst-hook-15.12-4-x86_64.pkg.tar.xz
linux-4.7.6-1-x86_64.pkg.tar.xz

Is there anyone who experienced the same?

Last edited by ergya (2016-10-29 20:22:42)

Vi0L0 · 2016-10-29 20:29:04

You probably need to update repo entries since right now I only have access to this one:

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

Current version of catalyst-hook @ [catalyst] repo is 15.12-5 and it does support 4.8 kernel (tho I didn't check 4.8.4)

Last edited by Vi0L0 (2016-10-29 20:32:10)

ergya · 2016-10-29 20:39:58

Vi0L0 wrote:

You probably need to update repo entries since right now I only have access to this one:
[catalyst]
Server = http://mirror.hactar.xyz/Vi0L0/catalyst/$arch
Current version of catalyst-hook @ [catalyst] repo is 15.12-5 and it does support 4.8 kernel (tho I didn't check 4.8.4)

Thank you Vi0L0. After upgrading SDDM started again. I am on 4.8.4. So far so good . May I ask what is the relationship between your "custom" repository packages and AUR packages? AUR seems to be outdated.

Vi0L0 · 2016-10-30 18:53:39

It's because I'm treating 15.12 as unstable because it often has problems on pre-gcn cards. Maybe trick with control file could fix this situation, but even then I'm not sure if it's worth changes - for now it looks like [catalyst-stable] repo and stable AUR packages are bringing 15.9 version, while [catalyst] repo and AUR's catalyst-test are bringing 15.12 version. I'm updating all pkgs and repos, so everyone can find something to pick.

Jannis · 2016-10-31 23:49:00

Could be unrelated to fglrx/catalyst but I noticed on linux-ck 4.8.5 TF2 takes an insane 17 minutes to start up (I think it's compiling shaders during that time) and I get an unplayable sub 20 fps.
On linux-ck 4.8.4 the game starts up in just under 2 minutes and I get a pretty constant 60+ fps with the exact same settings.

This is on a radeon hd 7870 (GCN 1.0) with catalyst-hook 15.12-5 from mirror.hactar.xyz.
glxinfo looks normal; direct rendering working. kwin_x11 with desktop effects working.

Will retry with stock arch kernel when 4.8.5 hits stable repos to see if this might just be a ck kernel issue and not related to fglrx.

Anyone else having similar results with a GCN 1.0 card on a 4.8.5 kernel?

EDIT: Nevermind, it's a problem with linux-ck using 100Hz now. Catalyst/fglrx work fine on stock arch 4.8.6 kernel and on linux-ck 4.8.6 as well ~~when building it with CONFIG_HZ_1000=y and CONFIG_HZ=1000.~~ EDIT²: latest ck also fixes it with the new 100Hz default.

Last edited by Jannis (2016-11-09 15:30:36)

NoSuck · 2016-11-13 13:19:31

I switched back to Catalyst recently. Thank you for doing such a good job with the repos. I probably wouldn't have switched otherwise.

Which of the following is correct?

Arch Wiki, "Catalyst" wrote:

Compared with the open source driver, Catalyst performs better in both 2D and 3D rendering...

X.org, "ATI Proprietary Driver" wrote:

2D performance is below of what the open source DRI drivers do provide...

In my particular case, the Arch Wiki is wrong. 2D performance is very poor with Catalyst. I can see feh drawing images from top to bottom, as if the image is loading through a web browser with a slow connection.

Are there ways to mitigate this?

Vi0L0 · 2016-11-15 12:46:06

In my case catalyst performance in 2d is not bad. It's way better tho when Tear Free and V-Sync options are disabled (via amdcccle and de) - then performance is equal or even better than open source drivers. Once those options are enabled the perfromance drops but still is acceptable. (my latest problems on catalyst are terrible glitches in chromium when running kde)
Is your catalyst installed correctly? Is glxinfo giving correct info and is there no errors in Xorg.0.log?
Edit: Let me correct wiki, because as for me in generall neither 2d nor 3d performance is better on catalyst. It could be equal, sometimes catalyst is better and sometimes oss is better, but telling that catalyst is definitelly better is a lie. Especially when ie amdgpu oss driver performs much better here on 290x even in demanding games such as Metro Redux or Bioshock Infinite than catalyst...

Last edited by Vi0L0 (2016-11-15 12:52:21)

lo7ev · 2016-11-15 19:47:27

Sup, guys. I'am a "happy" owner of HP m6-1061 w/ AMD Radeon HD 7670M. I want to brut wpa2 handshakes with pyrit, using GPU. I've understood that it takes some time to set everything up. Because pyrit requires catalyst(as I've read) and catalyst requires older version of xorg. Can anybody helps me with this stuff?

NoSuck · 2016-11-16 14:21:49

Vi0L0 wrote:

In my case catalyst performance in 2d is not bad. It's way better tho when Tear Free and V-Sync options are disabled (via amdcccle and de) - then performance is equal or even better than open source drivers.

Good! Then there is hope. I changed V-Sync from "Off, unless application specifies" to "Always off." Image viewing with feh seems to have gotten just a little faster, but that might be my imagination.

Vi0L0 wrote:

Is your catalyst installed correctly? Is glxinfo giving correct info and is there no errors in Xorg.0.log?

I found an error:

Xorg.0.log wrote:

[ 188.441] (EE) AIGLX error: failed to open /usr/X11R6/lib64/modules/dri/fglrx_dri.so, error[/usr/X11R6/lib64/modules/dri/fglrx_dri.so: cannot open shared object file: No such file or directory]

So it looks like I have some troubleshooting to do. I'll report back with results.

Vi0L0 · 2016-11-16 17:13:12

NoSuck wrote:

Xorg.0.log wrote:
[ 188.441] (EE) AIGLX error: failed to open /usr/X11R6/lib64/modules/dri/fglrx_dri.so, error[/usr/X11R6/lib64/modules/dri/fglrx_dri.so: cannot open shared object file: No such file or directory]
So it looks like I have some troubleshooting to do. I'll report back with results.

It's typical and not real error, mentioned path is hardcoded into fglrx but fglrx always finds this library anyway. The problem lays elsewhere, don't waste your time tracking this one. As a workaround for above note you can make a symlink from real lib to the path mentioned above - error will be gone, but behaviour should not change.

lo7ev wrote:

Sup, guys. I'am a "happy" owner of HP m6-1061 w/ AMD Radeon HD 7670M. I want to brut wpa2 handshakes with pyrit, using GPU. I've understood that it takes some time to set everything up. Because pyrit requires catalyst(as I've read) and catalyst requires older version of xorg. Can anybody helps me with this stuff?

everything should be in the wiki, use catalyst-stable and xorg117 repo (to downgrade xorg-server packages use pacman -Syuu) add those repos above even [core] in /etc/pacman.conf, install needed catalyst packages like :
pacman -S catalyst-dkms catalyst-utils catalyst-libgl opencl-catalyst lib32-catalyst-utils lib32-catalyst-libgl lib32-opencl-catalyst
run aticonfig --initial to create xorg.conf, modify boot options and add nomodeset to it, you can additionally create /etc/modprobe.d/blacklist.conf file with something like:
blacklist radeon
blacklist amdgpu
blacklist kms
blacklist drm_kms_helper
then reboot and hopefully you should be good. Run clinfo to make sure opencl is supported well
PS. if your hardware supports sume kind of hybrid graphics/powerXpress then don't add nomodeset and don't blacklist

Edit: TearFree is now available in fresh xf86-video-amdgpu-git, phoronix said it also works on fresh radeon , you only have to add
Option "TearFree" "1"
line to your device section of xorg configuration file
Log assured me that it was enabled:

cat /var/log/Xorg.0.log | grep -i tear
[     6.521] (**) AMDGPU(0): Option "TearFree" "1"
[     6.680] (**) AMDGPU(0): TearFree enabled
[     6.680] (WW) AMDGPU(0): KMS Pageflipping: disabled because of ShadowPrimary/TearFree

After disabling vsync in kde and games i can see no tearing. It does have some negative impact on performance but it can be clearly seen mainly in multiplayer fpses, like cs:go. In DE and many titles you won't see its impact.
For sure it performs better than catalyst' version.

Last edited by Vi0L0 (2016-11-19 13:23:54)

diesel1 · 2016-11-20 21:49:32

Hi Vi0L0,

Am I correct in thinking that I can use the current 4.8 Arch kernel and headers to reconfigure and recompile to allow amdgpu-pro on my 290?

I am beginning to look for GTX 970/980 cards with a view to replace the R9 290 for an easy way to enjoy my 4K Dell monitor even though the gains would be minimal!

I am currently using Catalyst on Manjaro (4.4.33-1) as I could not achieve an operational configuration using Arch.

I really do want to be using Arch.

Diesel1.

*EDIT* Okay, having read a couple of articles on Phoronix, it would make sense to hang on until 4.9 comes online in the repos.

*EDIT 2* Having read some more it seems that there will be no HDMI audio in AMDGPU for the foreseeable future until kernel 4.11 comes online. If this is true then I am afraid my dealings with AMD graphics cards will be over as soon as I can buy a GTX 1060.

Last edited by diesel1 (2016-11-22 20:08:50)

Vi0L0 · 2016-11-23 20:48:39

diesel1 wrote:

Am I correct in thinking that I can use the current 4.8 Arch kernel and headers to reconfigure and recompile to allow amdgpu-pro on my 290?

sorry for late answer, i was hoping to check it on my own but still can't find time for this. Last stable kernel I did check with amdgpu-pro was 4.7, but like on other kernels I was catching lock ups, complete freezes. And i'm almost 100% sure that new kernel won't fix it, that problem lays within libraries. Latest amdgpu-pro i checked is one available on AUR so in version 16.30.3.315407-2. Comments are saying that dkms build is crashing on 4.8 arch kernel. You can always take arch kernel, compile it with cik enabled and use amdgpu-pro simply without dkms package, it will work with amdgpu kernel module and -pro libs, just as good if not better than with dkms module.
Some time ago 16.40 was released but AUR still isn't updated, tho people are offering some patches in comments. Could be worth checking, but im not giving it that many hope after reading phoronix comments which were mentioning that theres not many changes applied.

I'm personally charmed by amdgpu drivers and it's easy to see that amd is focusing on it, not on the -pro version.

diesel1 wrote:

*EDIT* Okay, having read a couple of articles on Phoronix, it would make sense to hang on until 4.9 comes online in the repos.

AMD didn't precise when amdgpu for cik and si cards will be flagged as stable, but it won't happen in 4.9 and afaik not in near future. So you won't be able to use it out of the box via arch official repos.

diesel1 wrote:

*EDIT 2* Having read some more it seems that there will be no HDMI audio in AMDGPU for the foreseeable future until kernel 4.11 comes online. If this is true then I am afraid my dealings with AMD graphics cards will be over as soon as I can buy a GTX 1060.

Now this is strange because I'm using amdgpu (since like 4.7 maybe earlier) and pulling sound to my tv via HDMI. TV is then pulling sound to my stereo amplifier and I can swear that I can hear movie music comming from my stereo speakers. The quality is also fine. I'm unable to test more channels tho.

Last edited by Vi0L0 (2016-11-23 20:51:55)

mirh · 2016-11-24 20:08:57

diesel1 wrote:

*EDIT 2* Having read some more it seems that there will be no HDMI audio in AMDGPU for the foreseeable future until kernel 4.11 comes online. If this is true then I am afraid my dealings with AMD graphics cards will be over as soon as I can buy a GTX 1060.

First "DAL won't be mainlined at least until 4.11" doesn't mean 4.11 will see it.
Second it's not like only torvalds branch exists.

Third, nvidia cards are usable only with closed source driver. Which for amd is called AMDGPU-PRO and has everthing working.
So I really don't see the shadow of a point there.

diesel1 · 2016-11-25 00:28:19

Vi0L0, thanks for your advice, packages and determined support regarding all things AMD. It is highly commendable that you put so much effort into providing packages for Arch.

mirh, true 4.11 is far from being set in stone regards features. I cannot disagree that there are many branches of the Linux Kernel tree. Finally, you are quite correct in saying nVidia cards seem less useful than AMD cards (when using the free/libre drivers on an equivalent) unless you use the proprietary drivers (I wish nVidia would free their drivers).

Ultimately this has all come too late for me as I purchased a GTX 1060 6Gb MSI GAMING X card this afternoon. I wish AMD the greatest of success in their CPU (I am still a staunch AMD CPU supporter) and GPU businesses.

I am currently setting up Arch on my PC for maximum use and pleasure!

Thank you Vi0L0.

Kind regards,

Diesel1.

Last edited by diesel1 (2016-11-25 00:29:42)

NoSuck · 2016-11-26 07:29:22

This video shows how slow 2D Catalyst is on my system. Please confirm that this is slow. I only called it slow, because it is slower than xf86-video-ati, but who knows? Maybe this is normal?

■ sudo lshw -sanitize -class display
  *-display
       description: VGA compatible controller
       product: Pitcairn PRO [Radeon HD 7850 / R7 265 / R9 270 1024SP]
       vendor: Advanced Micro Devices, Inc. [AMD/ATI]
       physical id: 0
       bus info: pci@0000:01:00.0
       version: 00
       width: 64 bits
       clock: 33MHz
       capabilities: pm pciexpress msi vga_controller bus_master cap_list rom
       configuration: driver=fglrx_pci latency=0
       resources: irq:48 memory:e0000000-efffffff memory:f7e00000-f7e3ffff ioport:e000(size=256) memory:c0000-dffff

Drehstromlampe · 2016-11-27 22:52:41

NoSuck wrote:

This video shows how slow 2D Catalyst is on my system. Please confirm that this is slow. I only called it slow, because it is slower than xf86-video-ati, but who knows? Maybe this is normal?

■ sudo lshw -sanitize -class display
  *-display
       description: VGA compatible controller
       product: Pitcairn PRO [Radeon HD 7850 / R7 265 / R9 270 1024SP]
       vendor: Advanced Micro Devices, Inc. [AMD/ATI]
       physical id: 0
       bus info: pci@0000:01:00.0
       version: 00
       width: 64 bits
       clock: 33MHz
       capabilities: pm pciexpress msi vga_controller bus_master cap_list rom
       configuration: driver=fglrx_pci latency=0
       resources: irq:48 memory:e0000000-efffffff memory:f7e00000-f7e3ffff ioport:e000(size=256) memory:c0000-dffff

Yes, the 2d-performance is bad. I'm not sure it should be that bad.

But no sane person would use fglrx drivers.

Gaming on linux works fine with open source drivers. They deliver ~80% of the performance of the windows drivers. (confirmed that, had the r7 265 myself)

For everything else there's dual boot with win7

NoSuck · 2016-11-28 12:56:54

Thank you for confirming. I'll figure out what I need to do.

My reason for switching to Catalyst was daisy-chaining support (Crossfire). Another perk of Catalyst has been finer response of my graphics cards' on-board fans while outputting 3D. (The open-source drivers just spin at 100%.) I appreciate all the hard work that goes into making these drivers available.

Jannis · 2016-11-28 21:54:48

Drehstromlampe wrote:

[...] But no sane person would use fglrx drivers.
Gaming on linux works fine with open source drivers. [...]

Unless you want to play TF2 on Linux; see bugreports with valve: #1943, #1991, #2013 and mesa: #93649, #95308.

Until somebody has an epiphany and can fix that, fglrx it is

But yeah, 2D isn't great with catalyst but it also isn't nearly that bad for me with a radeon hd 7870. I'd say with vsync disabled, I get a pretty usable plasma desktop and even firefox with hundreds of tabs is bearable (until I hit a cpu bottleneck with too many youtube tabs open).

NoSuck · 2016-11-29 01:25:57

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?

ElNinijo · 2016-12-02 05:03:26

Vi0L0 i have the same problem like Moggele there

I Have ATI 3550 when I install the newest kernel (any from 4.8) and start XServer the whole system freeze, when I downgrade kernel to 4.7 everything is perfect. I have no error in Xorg or loading kernel modules.

Vi0L0 · 2016-12-02 09:41:42

That's bad because I don't have hardware to help myself tracing it. Can you help me? At first let's get back to AUR as it seems better place to write about it.
Can you install linux in earlier 4.8.x version like 4.8.1? I don't know when the bug occured, so at first I would like to know if it's falling on whole 4.8 line or if it was some recent change that brake it. If you can check it can you give me answer on the AUR?

Arch Linux

#7476 2016-10-24 23:12:18

Re: The AMD/ATI Bar & Grill

#7477 2016-10-25 16:28:06

Re: The AMD/ATI Bar & Grill

#7478 2016-10-26 19:57:14

Re: The AMD/ATI Bar & Grill

#7479 2016-10-29 14:50:01

Re: The AMD/ATI Bar & Grill

#7480 2016-10-29 19:53:37

Re: The AMD/ATI Bar & Grill

#7481 2016-10-29 20:29:04

Re: The AMD/ATI Bar & Grill

#7482 2016-10-29 20:39:58

Re: The AMD/ATI Bar & Grill

#7483 2016-10-30 18:53:39

Re: The AMD/ATI Bar & Grill

#7484 2016-10-31 23:49:00

Re: The AMD/ATI Bar & Grill

#7485 2016-11-13 13:19:31

Re: The AMD/ATI Bar & Grill

#7486 2016-11-15 12:46:06

Re: The AMD/ATI Bar & Grill

#7487 2016-11-15 19:47:27

Re: The AMD/ATI Bar & Grill

#7488 2016-11-16 14:21:49

Re: The AMD/ATI Bar & Grill

#7489 2016-11-16 17:13:12

Re: The AMD/ATI Bar & Grill

#7490 2016-11-20 21:49:32

Re: The AMD/ATI Bar & Grill

#7491 2016-11-23 20:48:39

Re: The AMD/ATI Bar & Grill

#7492 2016-11-24 20:08:57

Re: The AMD/ATI Bar & Grill

#7493 2016-11-25 00:28:19

Re: The AMD/ATI Bar & Grill

#7494 2016-11-26 07:29:22

Re: The AMD/ATI Bar & Grill

#7495 2016-11-27 22:52:41

Re: The AMD/ATI Bar & Grill

#7496 2016-11-28 12:56:54

Re: The AMD/ATI Bar & Grill

#7497 2016-11-28 21:54:48

Re: The AMD/ATI Bar & Grill

#7498 2016-11-29 01:25:57

Re: The AMD/ATI Bar & Grill

#7499 2016-12-02 05:03:26

Re: The AMD/ATI Bar & Grill

#7500 2016-12-02 09:41:42

Re: The AMD/ATI Bar & Grill

Board footer