(I am guessing what Buddha is talking about, but I'm not sure.)
The idea of subpixel rendering is to triple the resolution by acknowledging that different coloured subpixels are not all in the same location, but are actually a row of 3 seperate pixels. Windows already supports it for fonts (although it doesn't work when I rotate the screen in control panel), but I have never seen anyone implement it in Direct3D.
So instead of 1680 x 1050, your monitor would have a resolution of 5040 x 1050. Maybe iZ3D doesn't need the extra resolution. But the VR920 for example would have a resolution of 1920 x 480.
How would it improve 3D?, you ask. Well the horizontal resolution doesn't just determine the X resolution, it also determines the Z resolution. People can see the difference between a single pixel of separation. The Z resolution might only be a hundred pixels sometimes. Although I think the idea was mostly to improve the graphics as a whole, rather than just the 3D.
You could try rendering to a texture 3 times the width of the screen, and then using a pixel shader to colour the subpixel based on the colour of the "pixel" in the texture that corresponds to that subpixel's location. You might need some form of compensation for the colouring of the edges or fine lines that would result.
Alternatively, you could render geometry at the pixel level, but change all the pixel shaders to work at a subpixel level (so textures would have increased resolution). That would add very little processing cost, although it would only provide half the benefit.
I was hoping to try it on the VR920 to get better than the 640x480 it is limited to. Unfortunately I ran into a few problems... I have no way of knowing the subpixel layout on the VR920, and the VR920 has optics that bend the red, green and blue by different amounts, and strangely some of the red light seems to go straight through while other red light is ghosted lots.
Please read http://www.grc.com/ctwhat.htm
to see what we mean by subpixel rendering. It explains it much better than I could.
EDIT: I just realised, the convergence for anaglyph mode is wrong by half a pixel for LCD monitors. Things rendered at "screen depth" where there is "no separation" actually have a 0.5 pixel separation. A totally useless observation, but I found it interesting. Maybe that's just me though.
Copy this code to clipboard: 0o1rp5zk
then go to http://mtbs3d.com/naw
to register. Use the code for $5000 startup bonus, and to support Gaza.