On Sunday 16 September 2001 09:33, Dmitry Yu. Bolkhovityanov wrote:
> Hi!
>
> We have received our long-awaited G450PCI cards, and I immediately
> tried to replace with them old Milleniums, which worked together with
> G450AGP (4 heads per computer). This finally worked, but not immediately,
> and some things are plainly strange.
>
> Below is my experience of making the configuration work, which can
> be useful to others. And I'll be very grateful for any comments,
> corrections, or hints.
>
> Hardware: ASUS P3B-F, PIII-800, 1GB RAM;
> G450AGP 32M SDRAM, G450PCI 16M SDRAM in slot 5 and a second unused G450PCI
> in slot 4.
> 2 Nokia 447PRO (30-96KHz, 50-150Hz) + 2 EIZO FlexScan T561 (30-96KHz,
> 50-160Hz).
>
> Software: RedHat 7.1 (kernel 2.4.2) with XFree86 4.0.3, Matrox
> drivers from ftp.matrox.com/..../beta_1_3_0/. Probably in some aspects
> 4.1 will work better, but most bugs seem to be in Matrox drivers. I've
> also tried 4.0.3's mga_drv.o, but this worked worse than Matrox's.
>
> XF86Config was generated with X -configure and later hand-tuned.
> Videomode used was 1152x864@90Hz for all heads.
>
>
> So, the problems:
>
> 0. Non-primary card doesn't work without HAL module (as it should,
> according to documentation). Monitor on its 1st head stays in suspend
> (VFreq=0, HFreq=0), while XFree86.0.log shows that it is being queried and
> DDC data is correctly read; monitor on 2nd head is turned on but shows
> garbage (not even a checkerboard). So, everything below was done with
> mga_hal_drv.o.
>
> 1. When AGP is non-primary, one has to specify VideoRam for it --
> otherwise the driver allocates 1M per head. PCI card doesn't have this
> problem.
>
> 2. Option "MGASDRAM" has to be specified for *both* heads, despite the
> fact that it is one physical device (however, VideoRam has effect only
> when specified in "Screen 0" Device). That's definitely a driver bug.
If you were using a current version of X, you'd see that the problem is fixed.
> 3. Non-primary card doesn't switch off 2nd head when terminating or
> switching to console -- the monitor remains displaying the picture.
>
> 4. The server starts *very* slowly -- as if it spends ~5seconds to boot
> each head (this is a long-known issue, which is absent with XFree driver
> (or is it absent without HAL?)).
It works much faster in 4.1 and cvs versions of X where multiheading is
supported in the X driver without HAL. I suggest getting that.
> 5. When AGP is non-primary and VideoRam >=16384 is specified, if "dri" is
> loaded, machine hangs cold (only Alt+PrtSc+B works). Very strange -- I'd
> expect this behaviour with *little* memory, not with >=...
>
> Of course, not all possible cases were tested (there are too many
> parameters). And option "Int10" wasn't tested (it *was* tested with DRI,
> and X only half-loaded and hanged forever until killed).
>
> #3 and #4 are simply irritating and aren't very serious, #5 would
> probably disappear with 4.1 (AFAIR, some DRI memory management problems
> were fixed there). #1 and #2 are cosmetic (however #1 is bad too, why
> can't memory size be determined?). #0 is serious, since overlay isn't
> supported with HAL.
I believe with matrox, only one card is supported to have DRI at a time.
There is a config option to choose this. Upgrade to 4.1 and lose the HAL
crap.
> To repeat, I hope this info will be useful to other people and
> will help Matrox, and I'll be very thankful for any comments.
>
>
> (Hey, Matrox, wanna hire me as a beta-tester? ;-)
>
> _________________________________________
> Dmitry Yu. Bolkhovityanov
> [EMAIL PROTECTED]
> The Budker Institute of Nuclear Physics
>
>
> _______________________________________________
> Xpert mailing list
> [EMAIL PROTECTED]
> http://XFree86.Org/mailman/listinfo/xpert
_______________________________________________
Xpert mailing list
[EMAIL PROTECTED]
http://XFree86.Org/mailman/listinfo/xpert
