1. 3D Scanners so real objects can be inserted into the game.

2. Ray tracing so you get most optical effects (shadows, refraction) without programming them specifically.

3. General physics APIs so everything can be treated using the same basic mechanics theory.

I don't believe that Forza 3 really uses 1 million polygons per car.

It supports up to 8 cars at 60 fps, which means rendering at 480 million polygons per second for the cars alone... X360's peak polygon rate is 500 million per second, which doesn't leave almost any capacity left for rendering the environment, track and the pixel shaders.

^I don't believe it either, but even if each Forza 3 car is just 100,000 polys, I still prove that a whole scene consists of a much higher poly count than the the total amunt of 900,000 pixels in a 720p screen res.

Or you actully did it better by citing the specific poly count crunching ability of the X360.

http://www.xbox.com/en-AU/support/xbox360/manuals/xbox360specs.htm

I think HS's point is that if you don't increase the resolution, increasing the number of polygons by a wild amount isn't going to be worth it... so raytracing would be the logical step.

Yes, that's his point, but as much as i like HS, he's totally wrong.

Look at the minimal 120x67 pixel screen shots in my sig - if a scene like that was deep enough (view distance) it's not hard to imagine that it would need a massively higher poly count than just 120x67= 8040 polygons to properly and fluidly render that scene with a dynamic camera (such as in normal game play).

Really there's a magnitude of difference between 2D and 3D - that extra dimension can increase the polys/pixels that need to be calculated by a magnitude of at least 100 or even 1000s.

If you want to render a massive open space, you should use a less detailed model for the far away objects... you don't need to calculate thousands of polygons for an object which will take up 100 pixels on the screen.

I suspect that's what games like GTA4 do.

Well, that's true of course. That's why devs use LODs and stuff.

So maybe a magnitude of 1000 is an over-exaggeration, but the number crunchin power of the X360 - 500 million polys per second already hints that it indeed is really important to be able to "render (dramatically) more polygons than you have pixels" (the citation was HappySquirrels original statement).

@Slimebeast: Remember that the 360's GPU uses unified shaders, so the 500 million polygons per second are only possible if the game isn't rendering any pixels.

Divide the capacity between vertex and pixel shaders, and you get a quite smaller amount (dependant on the game of course).

At 60 fps, it's a theoretical maximum of 8.3 million polygons per frame... and then reduce that further to account for the pixel shaders. In the end you aren't that much above the number of pixels.

True, but the limit is not the resolution. For example, a normal SD TV broadcast is a lot less pixels than a 720p rendered Call of Duty or any other HD game, but the TV movie broadcast looks a lot better because it's photorealistic. To reach that quality you would need a much higher polygon count on your graphic models than in current gen games.

Why else would a dev like Turn 10 and others mention the huge poly counts in character models in games if they have no use (even if 1 million polys in a Forza 3 car is an exaggeration it still gotta be many)? I know it's mostly marketing talk, but all serious devs talk poly counts too in interviews.

Meanwhile, we know that screen res will not dramatically go up in the coming decade. 2500x1600 res screens will still be rare in 2015.

The amount of polygons is important, but if you look at Forza 3 and GT5 it seems clear that the amount of polygons is good enough. I think they would improve much more by using better pixel shaders and textures, than more polygons.