Johannes Erdfelt wrote:
On Wed, Oct 25, 2006, Tom Horsley <tomhorsley@xxxxxxxxxxxx> wrote:
On Wed, 25 Oct 2006 14:02:48 -0700
Johannes Erdfelt <johannes@xxxxxxxxxxx> wrote:
What monitor do you use?
It's kinda weird - I use a Westinghouse LVM-42w2 HD monitor, but I have both
FC5 and FC6 supposedly using the refresh rates it gets from the monitor
for the 1920x1080 display mode, so the fact that it is much more frequent
on FC6 (and I've never seen it happen on Windows) makes me think it is more
like a random bug of some kind than a refresh rate problem. (The monitor
will tell me if the signal is out of range, and it doesn't do that, it
just turns off).
It also shows some tendency to happen more frequently when running things like
opengl screensavers than when running a simple 2d desktop.
It's not exactly a refresh rate problem. I'm not a graphics technology
guru, but I'll try to explain.
The speed of the dotclock is causing either the ATI card to produce a
slightly dirty signal or the monitor is having trouble handling it. I
never found out which is the problem (and from reading around, it might
be either or both depending on your setup), but I did find out that
reducing the dotclock helps the problem.
You can reduce the speed of the dotclock by using reduced blanking
interval timings. Since you're using an LCD, you don't need all of the
extra blanking intervals that CRTs need. As a result, you can keep your
high resolution without sacrificing refresh rate and hopefully solving
your problem.
The problem with the radeon driver is that it rejects timings with a
reduced blanking interval. You can hack the driver to fix that.
It shouldn't be doing so anymore, at least in FC6.
The other problem is that your card only has a fixed amount of memory
bandwidth. The DAC or TMDS controller has to compete with the rendering
engine for access to pixels; if it loses, the frame will just go black.
1920x1080 at 60 Hz is 124 megapixels per second, and at 32bpp that's
500MB/sec. VRAM is fast but not always that fast.
- ajax
--
fedora-test-list mailing list
fedora-test-list@xxxxxxxxxx
To unsubscribe:
https://www.redhat.com/mailman/listinfo/fedora-test-list