Mailinglist Archive: radeonhd (427 mails)

< Previous Next >
Re: [radeonhd] HD 3650/ fglrx wants 4GB of VideoRAM/ radeonhd v 1.2.5 crashes
  • From: Urigeller <Urigeller@xxxxxx>
  • Date: Wed, 06 May 2009 13:15:11 +0200
  • Message-id: <4A01713F.4040708@xxxxxx>
Matthias Hopf wrote:
On May 06, 09 10:34:53 +0200, Christoph Piefke wrote:
Does removing one of the ram chips in your system help?

Yes, it works fine, except that this procedure creates a rather useless
paperweight worth of 2GB ram.
But screen resolution is just fine.

Xorg.0.log can be found here:

You might have some luck with this description:

Given the complexity of the issue, you're pretty much on your own right
now, until somebody who *really* understands the issue (which is, for
instance, not me) sits down and creates a great mtrr workaround patch
for the kernel :-(


I think I can update some of the informations on the issue which I
described in the post that Matthias linked to. I recently chose to look
again at the mentioned kernel option CONFIG_MTRR_SANITIZER. It's located
in the "Processor type and features" section, down at the option for
MTRR support.

I enabled MTRR cleanup support (CONFIG_MTRR_SANITIZER), set MTRR cleanup
cleanup spare reg num (CONFIG_MTRR_SANITIZER_SPARE_REG_NR_DEFAULT) to 0.
After that, my dmesg looks like this (excerpt):

[ 0.000000] original variable MTRRs

[ 0.000000] reg 0, base: 3328MB, range: 256MB, type UC

[ 0.000000] reg 1, base: 3584MB, range: 512MB, type UC

[ 0.000000] reg 2, base: 0GB, range: 4GB, type WB

[ 0.000000] reg 3, base: 4GB, range: 512MB, type WB

[ 0.000000] reg 4, base: 4608MB, range: 256MB, type WB

[ 0.000000] total RAM coverred: 4096M

[ 0.000000] Found optimal setting for mtrr clean up

[ 0.000000] gran_size: 64K chunk_size: 64K num_reg: 5
lose cover RAM: 0G
[ 0.000000] New variable MTRRs

[ 0.000000] reg 0, base: 0GB, range: 2GB, type WB

[ 0.000000] reg 1, base: 2GB, range: 1GB, type WB

[ 0.000000] reg 2, base: 3GB, range: 256MB, type WB

[ 0.000000] reg 3, base: 4GB, range: 512MB, type WB

[ 0.000000] reg 4, base: 4608MB, range: 256MB, type WB

This corresponds to the values in my original post.
So I disabled my little hack and I'm having no problems so far (running
this configuration since about a month). This workaround inside the
kernel is probably much better than mine because it corrects the MTRRs
as one of its first actions. I had to rely on a boot script (which of
course kicks in much later) to do this cleanup.

Also, this probably implicates that the kernel has a viable method to
detect and corret erroneous MTRR tables without any further user action.
If you want to take a shot at this I recommend to read the help pages of
the mentioned kernel options, this behaviour can be modified through
boot parameters (to save you reboots and reconfigurations of your kernel).

To unsubscribe, e-mail: radeonhd+unsubscribe@xxxxxxxxxxxx
For additional commands, e-mail: radeonhd+help@xxxxxxxxxxxx

< Previous Next >
Follow Ups