Radeon HD 6370D - GPU from hell checkerboard problem

marko2010 · 2016-10-01 06:23:02

I have a computer with a Radeon HD 6370D integrated GPU, and am using xf86-video-ati package with the linux-lts kernel. Every time I start X I get an unusable screen with a checker-board pattern (see below).

monitor

This problem not only shows up in Arch other but distros too -- Debian, Mint, & Opensuse. I can get the card to work partially in Debian Jessie, but only if I do not install the firmware-linux-nonfree package thereby severely limiting the performance of the card.

Any ideas for making this card useable? I already swapped out the monitor to eliminate that as a cause. And I tried creating a simple /etc/X11/xorg.conf.d/20-radeon.conf file as suggested in the troubleshooting section of the Arch Wiki.

olive · 2016-10-01 09:22:24

Have you checked the Xorg log file? What does glxinfo gives? To have the output of this command use some tricks, for example launch a terminal and type glxinfo > test.txt (this should be doable even if you do not see what you do). Or go to a virtual console (Ctrl-Alt-F2, for example); then

export DISPLAY=:0.0
xhost +
sleep 5 ; glxinfo

Then switch back to the non functional graphic display (typically Ctrl-Alt-F7), wait 5 seconds, and come back to the virtual console. Or ssh to the machine and do something similar. Notice that it is important that the non functional X display is enabled while running glxinfo (hence the sleep 5 trick).

I have a desktop AMD HD6570 (seems similar to yours) that work fine. What you describe seems to be a bug in the driver. This driver need the proprietary firmware to work properly. There is also the proprietary catalyst driver, but it is more difficult to install (see https://wiki.archlinux.org/index.php/AMD_Catalyst)

Last edited by olive (2016-10-01 09:29:56)

R00KIE · 2016-10-01 11:27:23

You could try to remove the xf86-video-ati package and try again, that will use a different xorg driver which might work. When trying this make sure you don't have custom configuration files that might be getting in the way.

olive · 2016-10-01 12:21:15

R00KIE wrote:

You could try to remove the xf86-video-ati package and try again, that will use a different xorg driver which might work. When trying this make sure you don't have custom configuration files that might be getting in the way.

That will at best fall back to the generic vesa driver with no hardware acceleration and possibly not the correct resolution. This driver is usually only useful to launch an installation program or to have a basic graphic display to see something when you try to debug problems; but not as a definitive driver.

R00KIE · 2016-10-01 14:05:40

olive wrote:

That will at best fall back to the generic vesa driver with no hardware acceleration and possibly not the correct resolution. This driver is usually only useful to launch an installation program or to have a basic graphic display to see something when you try to debug problems; but not as a definitive driver.

I don't know what makes you say that, the modesetting driver, which uses glamor for acceleration, is part of the X server and will be used if no hardware specific driver is found. It is perfectly usable and since it is completely different code it might work well.

olive · 2016-10-02 07:26:28

@ROOKIE You are right and you learn me something. I have about the same performance with and without the ati xorg driver. But then what's the point os these particular drivers? Is there any plan to abandon them to have a single good universal Xorg driver? It would be worthwhile to update the wiki. At least of AMD card the wiki says that we have the choice between catalyst and xf86-video-ati, but there is this third default universal driver too.

Last edited by olive (2016-10-02 07:30:52)

R00KIE · 2016-10-02 11:32:52

The hardware specific drivers came a long time before the modesetting driver and use 2D hardware acceleration blocks while the modesetting driver uses the 3D engine to do 2D acceleration. I believe the plan is to stop using hardware specific drivers once glamor acceleration and the modesatting driver are more mature and most bugs and speed problems have been worked out.

In this case the OP could also have toggled the acceleration method of xf86-vide-ati, but given that the defaults don't work at all it's probably not a bad idea to try with a different driver.

olive · 2016-10-02 19:22:01

According to ome benchmarks (Unigene Valley, glxgears, gtkperf). I have consistently slightly better performance with the modesetting driver. That's interesting.

R00KIE · 2016-10-02 19:54:28

Please don't hijack/derail the thread, lets keep it on topic. If you want to discuss benchmarks and do comparisons between the two xorg drivers please open a new thread specifically for that.

Arch Linux

#1 2016-10-01 06:23:02

Radeon HD 6370D - GPU from hell checkerboard problem

#2 2016-10-01 09:22:24

Re: Radeon HD 6370D - GPU from hell checkerboard problem

#3 2016-10-01 11:27:23

Re: Radeon HD 6370D - GPU from hell checkerboard problem

#4 2016-10-01 12:21:15

Re: Radeon HD 6370D - GPU from hell checkerboard problem

#5 2016-10-01 14:05:40

Re: Radeon HD 6370D - GPU from hell checkerboard problem

#6 2016-10-02 07:26:28

Re: Radeon HD 6370D - GPU from hell checkerboard problem

#7 2016-10-02 11:32:52

Re: Radeon HD 6370D - GPU from hell checkerboard problem

#8 2016-10-02 19:22:01

Re: Radeon HD 6370D - GPU from hell checkerboard problem

#9 2016-10-02 19:54:28

Re: Radeon HD 6370D - GPU from hell checkerboard problem

Board footer