3D Gaming*
- Thread starter Arwin
- Start date
On halving the resolution:
I play stereoscopic games now with the nVidia kit. Each eye gets a different image when your brain combines them. Textures appear to have more detail. Aliasing is reduced significantly. When viewing in stereo it is similar to a straight doubling of the horizontal resolution, like oversampling, but this time with two viewpoints.
Why not then halve the horizontal resolution to get back to the detail levels we'd see in 2D? Really you should be losing nothing by doing that.
I'd love to test this out, but need to find a 3D compatible game that I can halve the horizontal res of the backbuffer.
I play stereoscopic games now with the nVidia kit. Each eye gets a different image when your brain combines them. Textures appear to have more detail. Aliasing is reduced significantly. When viewing in stereo it is similar to a straight doubling of the horizontal resolution, like oversampling, but this time with two viewpoints.
Why not then halve the horizontal resolution to get back to the detail levels we'd see in 2D? Really you should be losing nothing by doing that.
I'd love to test this out, but need to find a 3D compatible game that I can halve the horizontal res of the backbuffer.
That'll reduce the amount of pixels you have to draw, but triangle setup and draw will be doubled with two eyes, so you should in theory have to halve model density too, unless there's a sneaky workaround that can reuse what's drawn for one eye for another, which is certainly true for distant objects. But there should be other compromises beyond just resolution, at least in an accurate 3D representation rather than a card-board cut-out universe.
If they don't provide a 2D option, they'll be confronted with a very small userbase to sell to!
(I presume you mean properly 2D optimised to make it look as good as it can).Billy Idol
Legend
If they don't provide a 2D option, they'll be confronted with a very small userbase to sell to!
Yeah, this is what I think (and hope) too!!
I've had a longer post about this in the KZ3 thread, but I believe that a fully deferred rendering based engine should be able to add convincing stereoscopy with just a post process, using all the 3D data available in the G-buffer. In this case they could go with a far, far smaller performance hit.
The problem with the card-board effect is that you just shift pixels vertically that seem to belong together and by the same amount. But if you have accurate 3D information from each pixel, you can adjust the shifting on a per-pixel level which would make it work a lot better. And a deferred renderer has both Z-depth and normal positions for each pixel so you have exact 3D data.
It still wouldn't be as good as rendering from two completely different points of view, because it couldn't account for the additional detail visible from the different cameras; although maybe that could be adjusted by a rendering trick as well. Like, sacrifice a little more memory and render dynamic objects into a separate buffer, so that you have pixel data for all the background stuff they'd occlude. Then composite them in only once the pixel shifting is completed.
But it should still be vastly superior to the cardboardish stuff and several times faster then fully rendering two 720p buffers with different geometry.
I'd realy like to hear Sebbi's thoughts on this...
Really though, any object will occlude a different part of something behind it from two different viewpoints. Doesn't matter if the object is dynamic, it's the camera(s) being dynamic that's the problem.
The Viewsonic VX2268WM.
Yes, that's exactly what I meant. It would be such a shame to downgrade whatever graphics they would have achieved if not for the 3d.If they don't provide a 2D option, they'll be confronted with a very small userbase to sell to!
They don't , they reduced rez most probably , but went for reall 3d stereo with dual separated cameras.
Deffered or not you could use the depth buffer for pseudo stereo.But not a really good one.
Yeah, but there are two things that can help.
First is that your brain compensates for this issue in real life by giving you a single "processed" single image that contains data from both eyes, while keeping the depth perception. Accounting for this could help in a 3D game; render one (dominant) eye's point and shift only for the second eye, or render a mid point and shift for both eyes and so on.
Maybe it's even possible to bend the eye vector a little during the G-buffer's rendering to accommodate for a few pixel's worth of extra info and undo it during the compositing, I dunno. There's a lot of research into stereoscopy and image processing that GG can tap into.
The other is that you could render in multiple passes, which most engines are doing to some extent anyway.
1. Render static background into G-buffer, light it, shift and create the left/right buffers.
2. Render dynamic objects, characters vehicles explosion debris etc. into G-buffer, light, shift, comp into left/right views.
3. Render gun (being the biggest occluder in an FPS) into G-buffer, light, shift, comp.
4. Render particles, opaque stuff and so on.
This would already help to preserve a lot of otherwise occluded image detail for a relatively low cost.
Sure it wouldn't be as good as full stereo rendering, it couldn't help with self-occluding stuff like a character's arm in front of his chest and so on. But it sure is a lot better trade off in memory, speed and resolution then rendering everything twice. It can preserve the full 720p res with AA (very important IMHO), the detail, and the speed, for a small extra rendering cost and some extra memory. I think it's a very viable alternative, at least for such kinds of games (maybe not for a racing one).
For middle-distance 3D I can see that working. Say for a top-down racer. For a first-person title, I'm not so sure. I'm thinking of typical up-too-close-to-an-NPC views, where I'm not sure if the face would resolve convincingly or just look a bit warped, but it shouldn't be too hard to knock up a demo and give it a whirl.
Well it will never be powerful enough. This is the same as the 60 FPS argument. We could have all games 60 FPS on any generation of hardware. Just a matter of sacrifices.
At best I think we mean to say something like, if we start next gen off with 3D games as the baseline, it'll be a lot easier to swallow, since we'll be getting the generation upgrade anyway (even if it could be bigger without 3D), rather than actually taking a step backward.
Whats the effect if one eye is looking into relatively perfectly rendered face for the relative technology level and the other is staring into a distorted looking face due to an artifact?
If this makes a person feel uneasy then it could be a good way to introduce a horror aspect back into a very jaded and cynical audience.
If this makes a person feel uneasy then it could be a good way to introduce a horror aspect back into a very jaded and cynical audience.
TheWretched
Regular
Just ask a person, who needs extremely different glasses for each eye...
Some time ago, I always got extreme headaches, when working on my computer... then I went to the optician (?), which deduced, that I needed new glasses... One eye was nearly correct, while the other was extremely off... after I got my new glasses, my depth perception rose INCREDIBLY. I don't know if this is placebo or anything else... but I guess this is comparable.
But... if the geometry is correct, whilst the shaders and stuff isn't, I guess this shouldn't be much of a problem for headaches or anything. It's more what computer games always did... fake stuff.
Couldn't it be possible to hack your geometry buffer into 3 layers... two near buffers for each eye, and a third for the far plane (which should be the same for each eye anyways)? And then overlay the near and far planes together for each eye? I don't know if this is possible though (still learning graphics programming, and I am by no means very advanced^^).
Some time ago, I always got extreme headaches, when working on my computer... then I went to the optician (?), which deduced, that I needed new glasses... One eye was nearly correct, while the other was extremely off... after I got my new glasses, my depth perception rose INCREDIBLY. I don't know if this is placebo or anything else... but I guess this is comparable.
But... if the geometry is correct, whilst the shaders and stuff isn't, I guess this shouldn't be much of a problem for headaches or anything. It's more what computer games always did... fake stuff.
Couldn't it be possible to hack your geometry buffer into 3 layers... two near buffers for each eye, and a third for the far plane (which should be the same for each eye anyways)? And then overlay the near and far planes together for each eye? I don't know if this is possible though (still learning graphics programming, and I am by no means very advanced^^).
If 3d TVs go mainstream like HDTVs now, then it would be natural to code and design 3d games. Currently it just doesn't seem to make sense for game developers to go that route and expect a reasonable amount of sale. The 3d install base is simply laughable and god knows if it would be even adopted, if GG doesn't handle this right then Sony would simply shoot themselves in the foot and loose dearly. 3D is truly a make or break decision, with possible reduced effects or resolution, a ridiculously small installbase, why would Sony risk that much? If they're doing it because they can compared to the competition then it could be a very ignorant and potentially disastrous move. Of course I hope all turns out well, but it's a dangerous game they're playing here, no pun.
Actually, my concern would be rendered ineffective IF developers can make two separate versions with the 2d's effects and res being pushed to heaven's end while the 3d version being optimized for whatever it takes to be 3d. That way we all win and I can stop whining.
Last edited by a moderator:
ShadowRunner
Veteran
It seems quite obvious to me that enabling 3D in games is not to the detriment of the 2D version. All the games we have seen in 3D so far are existing 2D games, proof that the 2D version isnt being sacrificed.
Making a game 3D should capable should be relatively cheep so it becomes a why not sort of situation.
Making a game 3D should capable should be relatively cheep so it becomes a why not sort of situation.
Exactly, games and movies can both release supporting 3D and 2D from the same disc or whatever distribution media.
Also, 3D TVs won't sell without content, and Sony's highly motivated in a boost, so supporting the effort with 1st party games on the PS3 is a pretty straightforward move. I'd even say it would be insane not to add 3D support to every game they can.
Also, 3D TVs won't sell without content, and Sony's highly motivated in a boost, so supporting the effort with 1st party games on the PS3 is a pretty straightforward move. I'd even say it would be insane not to add 3D support to every game they can.
