[ltp] Re: Unstable Radeon output at 1920x1200

Alex Deucher linux-thinkpad@linux-thinkpad.org
Mon, 24 Mar 2008 17:21:32 -0400


------=_Part_11844_21814338.1206393692369
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

On Mon, Mar 24, 2008 at 4:24 PM, Alex Deucher <alexdeucher@gmail.com> wrote:
>
> On Mon, Mar 24, 2008 at 3:01 PM, Shem Multinymous <multinymous@gmail.com> wrote:
>  > Hi,
>  >
>  >  I have a problem driving LCD monitors at 1920x1200 via DVI on my T43
>  >  with Radeon X300. The image shows up, but is unstable: at erratic
>  >  times but generally once per 20-60 seconds, I get two disruptive
>  >  effects: the image disappears for a couple of seconds, or there is a
>  >  momentary image tearing effect. Basically, it looks like the monitor
>  >  is temporarily losing sync with the DVI signal. This occurs with 3
>  >  different LCD monitors and 3 different DVI cables.
>  >  Now, I know 1920x1200 24bpp is pushing the limit of DVI, but here's
>  >  the thing: when dual-booting into Windows, the display is rock-solid
>  >  at 1920x1200 24bpp DVI, using exactly the same hardware.
>  >
>  >  This is reproducible on my setup. I'm running Fedora 8's Xorg 6.8.0,
>  >  and relying on EDID ("xrandr --output DVI-0 --auto", no modelines or
>  >  frequency limits in xorg.conf). Xorg's EDID detection log is attached.
>  >
>  >  There's a related effect with the VGA output: at 1920x1200 with a
>  >  particular Dell monitor, X.org's output causes a weird "pixels
>  >  jiggling around" effect, but again Windows looks stable. This analog
>  >  effect is more subtle and variable, so I'm not 100% sure that the
>  >  difference between Windows and Linux is not a coincidence.
>  >
>  >  Is this just a matter of getting Xorg to use a better modeline, or
>  >  something deeper such as misconfigured Radeon CRTC registers?
>
>  Please attach your full Xorg log, thanks.

Sounds similar to bug 15175.  Does this patch help?
I suspect the monitor may prefer slightly different dividers for the
pll.  Even though they get to the same clock, some monitors seem to
prefer different dividers combinations.

Alex

------=_Part_11844_21814338.1206393692369
Content-Type: text/x-patch; name=154clkfix.diff
Content-Transfer-Encoding: base64
X-Attachment-Id: f_fe7jmm5i0
Content-Disposition: attachment; filename=154clkfix.diff

ZGlmZiAtLWdpdCBhL3NyYy9hdG9tYmlvc19jcnRjLmMgYi9zcmMvYXRvbWJpb3NfY3J0Yy5jCmlu
ZGV4IDZmYmY3ZWQuLjI0MzVlNGMgMTAwNjQ0Ci0tLSBhL3NyYy9hdG9tYmlvc19jcnRjLmMKKysr
IGIvc3JjL2F0b21iaW9zX2NydGMuYwpAQCAtMTg3LDcgKzE4Nyw3IEBAIGF0b21iaW9zX2NydGNf
c2V0X3BsbCh4Zjg2Q3J0Y1B0ciBjcnRjLCBEaXNwbGF5TW9kZVB0ciBtb2RlKQogICAgIGlmIChJ
U19BVklWT19WQVJJQU5UKSB7CiAJQ0FSRDMyIHRlbXA7CiAKLQlpZiAoSVNfRENFM19WQVJJQU5U
KQorCS8vaWYgKElTX0RDRTNfVkFSSUFOVCkKIAkgICAgcGxsX2ZsYWdzIHw9IFJBREVPTl9QTExf
RENFMzsKIAogCVJBREVPTkNvbXB1dGVQTEwoJmluZm8tPnBsbCwgbW9kZS0+Q2xvY2ssICZ0ZW1w
LCAmZmJfZGl2LCAmcmVmX2RpdiwgJnBvc3RfZGl2LCBwbGxfZmxhZ3MpOwo=
------=_Part_11844_21814338.1206393692369--