I’m trying to find a good graphics accelerator for use in a cave style enviroment. Can anyone give me some advice on this? I’d like to get away with consumer hardware, but it’s not a must. What I basically want is:
-Solid drivers
-Stereo view
-Good fillrate and lots of triangles. Preferably relatively fast render-to-texture too. Not so interested in programmability.
-Lots of video memory
Also, is the zbuffer based stereo trick any good for professional applications, or should I use quad buffer stereo?
Forgive my ignorance, but what is the “zbuffer based stereo trick” ?
If you can afford it, I would suggest an Nvidia Quadro card with dual monitor output, to project with two DLP projectors, with polarized filters, and a special screen.
You may find this site very useful : http://www.geowall.org/
PS: instead of using quad bufferd stereo, you can get away with any Geforce able to do dual monitor output. Simply span your gl window across the two screens. Then draw the left eye in the left half and the right part in the right part with glViewport(). I hacked slightly a version of Rollercoaster 2000 and it worked great.
[This message has been edited by ZbuffeR (edited 02-01-2004).]
Thanks, the link is very informative and the double monitor trick sounds nice. I have to count how the prices add up, though, two projectors per view (there will be three) might get expensive. And the positioning of the six projectors might get tricky.
I can’t find any information on the trick I mentioned, but they used to somehow generate the stereo images from a single zbuffered image, making it work with most of the existing games. But I only read about it a long time ago, it might be buried by now.
If you’re using an active stereo (shuttered) system then you’ll need genlock across graphics cards and probably also some pretty nice projectors to handle the interleaved stereo fields at a decent rate with no persistence issues, only passive stereo (polarized filters) will work across multiple systems without a video level lock (line lock at a minimum). Genlock is still a very high end feature for graphics cards even if you can get a shutter sync from a card that won’t necessarily mean you have a genlock sync.
Passive stereo while it takes more projectors is a lot simpler to implement and you’ll probably get away with cheaper projectors (if you take left & right video channels from a single card as suggested), in addition to being able to go with cheap LCD projectors with no worries.
P.S. Real stereo rendering where you render two eyes is better than some bogus z buffer scheme that would require post processing anyway. Ultimately any scheme may need the equivalent display technology anyway to present to two eyes unless it’s truly horrific.
[This message has been edited by dorbie (edited 02-03-2004).]
You’re right. I really have to count the prices, but using two projectors per view might be the best alternative after all.
Only… The user won’t be looking directly at the side screens, doesn’t that mess up the polarization? Perhaps it’s tolerable and when you turn your head it’ll be allright.
And can I project the images through a mirror (to save some space) and still maintain polarization?
Dorbie, thanks for the information on genlock systems.
…you’ll probably get away with cheaper projectors […], in addition to being able to go with cheap LCD projectors with no worries.
Well, almost no worries.
To be precise, standard LCD projectors are often already polarized, i.e. blue & red are polarized horizontally, and green is polarized vertically.
I had to put my polarizing filters at 45°, one to the left, the other to the right, for the colors to appear somewhat correct… Plus the additional lost in brightness. Finally, it is just a tradeoff between brightness vs. cost.
And if you need info on how to get HW acceleration in multi-cards display, see http://www.realtimesoft.com/multimon/ogl/ , and the whole website, they have many interesting info. And a commercial product, of course