[rescue] Video Creator

[Big Endian]

><snip>
>So, we've established that the MCO reads from the frame buffer and splits
>the framebuffer for additional images.  Already knew that.

and unfortunately without some SERIOUS custom circuitry its the only 
way to do it.

>You are not saying anything terribly new about the Video Creator.  What
>does commenting about it's VLAN port have to do with anything?  The
>question was, is SCSI used only for controlling the settings, and I guess
>also the frame advance, stop, etc, VLAN features.  If so, any ideas why
>SCSI instead of serial.  Also, if Matrox cards can be used to drive old
>SGI monitors, why couldn't a matrox card with a few video splitters and a
>few of these boxes be used to get the same end effect of the MCO.  IE,
>split the matrox output out to each box.  Then, use the cropping features
>of the Video Creator boxes to pick out a different part of the 1280x1024
>video signal from the matrox for video output.

Why use the a matrox at all then?  If you're changing the waveform 
that much you will HAVE to generate new sync signals anyway.  Also 
this cropping/scaling only goes down to NTSC resolution so the most 
you can do is split 1280x1024 down to something you could display on 
a TV.  It wouldn't get you anything that could drive multiple 
monitors at any resolution.  If you want something like this you 
would have to go with a multi head AGP card or AGP + multiple PCI.

>Now, why do this instead of PCI video cards?  Because PCI sucks.  And
>having multiple video cards sucks.  If you need 64megs for textures, then
>each PCI card needs 64megs for textures, but the chances are that these
>textures are the same (assuming that the multiple video streams are
>something like different views of the same scene), so why not have the
>textures all on one card?  Further, to get the same bandwidth for moving
>stuff from memory or the CPU to the video cards, you would need 4
>different PCI buses to be able to drive 4 different PCI cards as well as
>one AGP bus can drive one video card rendering to four displays (which is
>why a multihead AGP card is far better than a single AGP and a single
>PCI).

PCI is perfectly capable of using AGP Apeture (sp?) memory.  This is 
main system ram mapped into the PCI address space for use as 
additional AGP/PCI texture storage.  I agree that it would be ugly 
and a large duplication of effort with the PCI cards.  Why would you 
need an MCO style thing?  If you want to take multiple lower 
resolution views on the same high rez pipe and display them on 
multiple monitors for something of a wrap around display type thing 
then wouldn't the G200/G450 do?   I don't think these will do what 
you want.  Also an option is multiple cards on an octane style box. 
Get multiple SI+tex cards together and do it that way.   Problem with 
that is that the minimum resolution of an octane is 1024x768 I think 
so smaller displays wouldn't work as well.

>OK, so maybe it would be over kill to use these boxes.  After all, you can
>buy quad head g200 cards and run it in low resolution.  But, quad g200s
>cost a furtune, and G200s don't have much texture memory, and I don't
>think that a G200 can drive a 640x480 display as well as an NVidea
>Geforce2 can drive a 1280x1024 display.

I agree.  I love my Geforce2.  I also like running it with my TNT 
16mb PCI card as a second head.  It works well for me on my games 
that support dual heads.  I've never done much work with writing 
things on PCs... I do my OpenGL work on SGIs and my powerbook G3. 
I'm still not sure why you need 64mb of texture memory, AGP andd 
multiple heads though... it seems to me that such things could be 
better done with SGI or sun equipment.

daniel
--