Diamond HD6450 Strangeness

I didn't know what noveau was, but it appears to be a driver for nVidia graphics cards, per

. So I am a little bit confused about how you are using it with ATI/AMD graphics cards. Do you also have an nVidia card in this machine?

You might boot with the MSI card and capture the dmesg output (kernel messages) at different points. It might also be interesting to get the list of kernel modules that are loaded. Grab a copy of the X log too - it will probably be named /var/log/Xorg.0.log or something close.

# boot up into text mode here, then dmesg >msi-before-x.txt lsmod >msi-modules.txt # start X # start an xterm, then dmesg >msi-in-x.txt # shut down X, then dmesg >msi-after-x.txt cp /var/log/Xorg.0.log msi-xorg.txt

Can you switch to a text console while X is running? (control-alt-F1)

# boot up into text mode here # log in blind dmesg >diamond-before-x.txt lsmod >diamond-modules.txt # start X # switch to text console, or ssh in from another machine, then dmesg >diamond-in-x.txt # shut down X, then dmesg >diamond-after-x.txt cp /var/log/Xorg.0.log diamond-xorg.txt # possibly start X again and then dmesg >diamond-in-x-again.txt cp /var/log/Xorg.0.log diamond-xorg-again.txt

Then... compare all your dmesg outputs. There may be some warning or information message that points to the problem. It may also be interesting to compare which modules are loaded with the different cards. Compare both of the X logs for the Diamond card to see if there are different messages at startup.

X drivers are getting better and better at this all the time, but one common case (IMHO) of a bad picture is that the driver isn't auto- detecting something correctly. If your video card driver reports anything in dmesg or the X log when it starts, compare that to the known specs of your video card to see if they match. There may be a command- line option to the video card driver to force (for example) a certain memory size or GPU type.

You already know about control-alt-backspace; when playing X games it can also sometimes be helpful to ssh into the X machine from another machine, so you can shut things down or restart if you go blind.

It may also be helpful to boot your machine with a live CD (Knoppix, SystemRescueCD, or one from your favorite distribution) to see if it behaves any differently. The live CD will almost certainly have different versions of the video drivers than what you have, which can help tell you if your problem is fixed in a newer driver, or maybe that it used to work and then broke.

Before X even starts, a lot of distributions try to load a custom VGA font in text mode to make the console look "better". Perhaps this process is screwing up for some reason on your Diamond card. If you can disable this custom font loading, it may help. I would look in /etc/init.d and /etc/rc[0-6S].d to figure this out, but I know that some distributions have reinvented the startup wheel and you'd need to look elsewhere.

Bugs like this have existed before. On the way, way back, for a few video cards, sometimes you had to boot to DOS or Windows first to get the card set up, and then reboot into Linux. You probably don't need to start carving out space for a DOS partition, though. :)

Matt Roberds

I'm sorry, I should have said Radeon instead of nouveau. It's definitely the Radeon driver. Thanks for your help.

I tried to follow your suggestions as closely as I can, but it's very difficult to switch the cards in and out of the machine to run these tests. As nearly as I can see there is no change to the modules loaded for the two graphics cards, either at the original boot into the console, before starting X or after shutting X down.

The dmesg lines for the two cards are substantially different. For the card that does not come up readable I don't see any error message, except for this possibility: "Radeon 0000:01:00.0: registered panic notifier." That's the only thing that looks like it doesn't belong.

When I start X there are numerous repeated sequences of a ~40-line block that begins with 'GPU lockup CP stall for more than 10000 ms'. In the last block it ultimately says 'GPU reset succeeded trying to resume'. There are no lines like this when X starts normally with the working graphics card.

I tried the card in another machine with an AMD processor and a different motherboard, but running the same linux kernel. It didn't work -- I couldn't ever get it to display readable text, and trying to start x resulted in nothing at all.

I'm really hoping that somebody else has run into this problem -- and solved it. The card really isn't worth any more effort than we've already put into it. In its defense, it doesn't say 'linux compatible' on the box.

Thanks again for your suggestions.

--
Cheers, Bev 
==================================================================================== 
"The problem with socialism is that you eventually run out of other  
people's money."

mint 15 light weight desktop, not petra 16

OK.

I don't know exactly what that is, but it doesn't *sound* like a big deal.

Hey... you found a difference! I don't know how to *fix* that particular error, but I think it is a symptom of your problem.

The X.org page for the Radeon driver

says bugs can be filed at . It might be worth a search there to see if anyone else has reported it. If not, report it yourself.

The X.org page also lists the kernel module parameters at

and shows how to check the current values. Perhaps changing one of those will help, although it may mean a lot of rebooting. (I'm not sure if you can remove and reinsert the radeon module when you're only running on a text console, or not.)

I usually ignore that, because it often means "we tested it on one specific 4-year-old distro with propietary drivers and it kind of worked". Googling is a much better plan, but it does kind of fall down with brand new hardware.

Matt Roberds

Reminds me of an old story about a patient telling the doctor that it hurts when I do this...