Why not "blown up grid AA" instead of RGSS

[Squeak]

I know this technique is currently not used (RGSS that is), but I was wondering why instead of meticulously rendering all the polygons first a little to the left, then right, then up and so on... why can't the renderer just blow the whole scene up one pixel or less, and then alphablend that on top of the scene to be AAed?
Why isn’t that a viable way to do it?

 

[Grall]

That's not AAing the scene as much as simply blurring it. Would just look kinda messy.

*G*

 

[DeathKnight]

Yeah, I just tried it out in PSP7 and the end results were fuzzy in some areas, but the jaggies were still very present.

 

[Tagrineth]

I know this technique is currently not used (RGSS that is), but I was wondering why instead of meticulously rendering all the polygons first a little to the left, then right, then up and so on... why can't the renderer just blow the whole scene up one pixel or less, and then alphablend that on top of the scene to be AAed?
Why isn’t that a viable way to do it?

Uh... look up ordered-grid supersampling.

That's a lot like what GeForce256 and GF2 do - render the scene with increased dimensions and filter down to a lower res (so 800x600 with 4x AA runs internally at 1600x1200 then gets filtered back to an 800x600 front buffer).

It isn't viable because it looks like crap compared to a good rotated grid, and a programmable rotated grid can perform BETTER too (3dfx, anyone? 4x AA didn't quite drop performance to 1/4 in many situations).