Motion blur and Depth of field in NV30?

[Sage]

ok you want to talk about TNT, the Voodoo3 had edge antialaising so but seriously, the performance of the GF2 and 4xfsaa is almost identical to that which it runs at twice the resolution. its not that hard to downsample, i believe Metabyte (Wicked3d folks) came up with software to do that back in 97 / 98?

 

[randycat99]

Excuse my newbie-ness on the topic, but is the accumulation buffer somehow related to a certain AA technique? That's the impression I got from some of the posts, but maybe I'm not understanding what exactly was said.

Also, are some depth-of-field blurring techniques more vulnerable than others to noisy or grainy artifacts in the final image?

(Edit done above- just noticed I wasn't asking the question I originally intended)

 

[Tagrineth]

No, the accumulation buffer is not used for AA...

3dfx's FSAA was done via their T-Buffer, which is a more powerful variation of an accumulation buffer, but it isn't an accu. buffer per se anyway.

 

[Vince]

No, the accumulation buffer is not used for AA...

Actually the two guys from, IIRC as it's been awhile, Silicon Graphics who propsed the A-Buffer in a paper called The Acumulation Buffer: Hardware Suport for High-Quality Renderings proposed it as a form of spatial anti-aliasing.

The other 'effects' of the A-Buffer, as thus by extention the T-Buffer, are nothing more than variations and manipulations/weightings of the samples - just as spatial AA.

3dfx's FSAA was done via their T-Buffer, which is a more powerful variation of an accumulation buffer, but it isn't an accu. buffer per se anyway.

The T-Buffer is definatly a derivative of the Accumulation Buffer in that it renders multiple scenes into the back buffer(s) and then accumulates them in some way for the final swap/output. The idea is fundimentally the same, the T-Buffer just does a hacked down version thats cheaper preformance wise.

I also wouldn't say it's a more powerful variation - although it has some advantages such that you can apply effects to a specific buffer or areas of the final output and thus re-render less. Where as on the A-Buffer you'd have to re-render the entire output each time.

 

[Bambers]

The gf2s FSAA wasnt thrown in at the end.

My gf256 (Creative) that I bought in nov99 did actually state Hardware FSAA as a feature and it even had the slider in the control panel under D3d on the creative drivers. It was only when the gf2 came out that they enabled forcing AA.

 

[Humus]

No, the accumulation buffer is not used for AA...

Actually the two guys from, IIRC as it's been awhile, Silicon Graphics who propsed the A-Buffer in a paper called The Acumulation Buffer: Hardware Suport for High-Quality Renderings proposed it as a form of spatial anti-aliasing.

Sure the accumulation buffer can be used for antialiasing, but it requires application support.

 

[Tagrineth]

The gf2s FSAA wasnt thrown in at the end.

My gf256 (Creative) that I bought in nov99 did actually state Hardware FSAA as a feature and it even had the slider in the control panel under D3d on the creative drivers. It was only when the gf2 came out that they enabled forcing AA.

Dude. Voodoo2 (and I think also Voodoo Graphics), and TNT both have the old D3D 'Edge AA' application-controlled AA. It's useless, there are maybe three games in existance which use it. It shouldn't even be factored into the discussion.

 

[Bambers]

I said FSAA ie Full Scene Antialiasing, not edge AA.

 

[Randell]

DOnt forget the Savage 4!

Name a game with it? Expendable I beleive plus Tomb Raider 1 or 2, cant remember.

What else had it?

 

[Nappe1]

Glaze3D supposedly had full Accumulation Buffer implementation on Hardware and it would have been using it for FSAA.

don't know much about it because something that never materialized to silicon isn't something I would like to look for.

 

[arjan de lumens]

The way I understand it, A-buffer, Accumulation buffer and T-buffer are three very different beasts indeed:

• The A-buffer essentially stores a coverage bitmap for each pixel, using the bitmap to determine which subpixels of a pixel are covered; the result is (a rather sort-dependent) edge anti-aliasing.
• The Accumulation buffer contains an RGBA color value for each pixel, much like a framebuffer; it is used to hold an intermediate result when blending together multiple frames. Can be used for application-controlled supersampling; it requires passing all geometry data multiple times for this use. Supported by OpenGL for ages, but hardware support is still not very common.
• The T-buffer is fundamentally a multisample buffer, storing multiple RGBA color values per pixel and blending them together when displaying them on screen. Included in the T-buffer functionality is multisample masking too.

Also, having an old Geforce256 lying around, I can indeed verify that the card is capable of just the same kind of OGSS as any Geforce2 (which is easily forced by the driver).