Hi folks, I've already posted most of this the the Linux support forum at Matrox; I'm hoping perhaps some part of this problem may be attackable from the source driver and that the problem isn't in the black recesses of the HAL library. I'm happy to put a shoulder into the hacking myself if folks can lend initial guidance.
The basic problem is: I have a G450 PCI with a DVI port that happens to be living in a box with a G550 Dual DVI. The G550 dual DVI is functioning very well with a single minor niggle I'll mention later. The G450 is working nicely-- once it's up. The major problem is that it doesn't always manage to come up (there are other more minor issues as well). Essentially anything involving a mode change seems to go wrong. When it doesn't come up stably, a quick flick to text VT and then back to X causes the garbage G450 screen to pop to a useful display mode and eveything looks fine. It does not matter if the G550 is also in the box, the G450 behavior is identical. A few points up front: My production setup is all DVI, all three monitors. The analog port on the G450 comes up fine all the time (OTOH, it disappears from X on the first VT switch or screensaver blank and never returns. An analog monitor reports 'no signal'). Even if I boot the machine with the PCI G450 as primary or only card, the text VTs only work on DVI until I start X. After that, text mode becomes a garbage mode (monitor says 'out of sync range: H=60kHz, V=38.8Hz') Screensaver screen blanking also goes wrong; the screens are not shut off but rather all go to undisplayable garbage modes; all monitors are continuously trying to sync to crazy modes with static, flashing and the occasional disturbing effect that looks like a piece of movie film melting in a projector.. This happens on the G550 and G450 both. Server startup also takes bloody forever with lots of flashing, scrolling garbage, pieces of previously dead X server framebuffers with shifted colors and cute dancing colored text characters. It looks like a reall bad crash until after about a minute, 'ping!', it's up. Well, it's up about 2 of three times. Odd thing is that this does not plague the kernel framebuffers. When I start up / drive / modify the G450 video mode through the mga.o fb interface, mode changes are crisp, immediate and require no settling time... just like in Windows. Oh, and I've tested both cards in Windows. They work flawlessly (as they do in Linux once they're up). Last problem which is minor and may be a stright up bug: The mga driver offers several modes on the DVI port of the G450 card that are available only via the analog interface. The problem is that XFree86 always chooses the highest refresh rate of a collection of same-resolution built-in modes, and in this case, that will both not display correctly and eventually fry the card. Specifically, the MGA driver insists on using 1280x1024 @ 75Hz (which is valid on the analog VGA port) rather than 1280x1024 @ 60Hz which is the highest the DVI port will go. I have to either force a modeline or force-reduce the monitor specs to trick it into not using the 75Hz mode. A complete dump of all config files, logs, a screenshot and system setup are at: http://forum.matrox.com/mgaforum/Forum2/HTML/002427.html Better there than cluttering up the mail here :-) Thanks folks. If it makes any difference, I'm trying to get this machine going as the main Ogg Theora hacking box. Assistance rendered will go stright toward reducing our release delay on the codec :-) Monty _______________________________________________ XFree86 mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xfree86