MSAA in NV2x actually is an enhanced SSAA!?

misae said:
You mean iXBT and all you need to do is a change of resolution within windows for the change to be applied... which OpenGL Guy mentioned.

So not some websites.. just the one.

The article published in English (Digit-Life) already contained the right results.
 
alexsok said:
Why haven't the p/reviews of the R300 thus far talked more about the R300's 6xAA?

Well, as far as I know, some websites were having problems with 6xAA modes, where the quality was indistinctive from the 4xAA modes (the problem is in the drivers, where u need to restart the maching when switching things for them to take effect).
Well, if that is the case, it appears to be doing a big injustice and disservice to the R300 coz based on Zephyr's screenies, the R300's 6xAA looks really good. Regardless of whether it was ixbt or any other website who knows about this "mode restart" bug, I haven't read enough about the R300's 6xAA... I suppose that's cause websites do not know how to "compare" (i.e. include in shootout benchmarks) the R300's 6xAA to others in terms of benchmarks, which is silly IMO.
 
Reverend said:
Xmas, from where did you get those NV AA "codes"?
I did not get them from anywhere. I drew my own conclusions based on information about the known modes and the mode descriptions.
 
Reverend said:
Question re Dave's Wildcat AA shots - why do the "whites" on the numpad on the AA shot look "less white" in the non-AA shot?

Because the pixels in that area comprise multiple subsamples, some of which are white and some of which were black. The non AA shot was showing the aliased 'either white or black' so because there aren't any extra subsamples you are only getting pure white (with aliasing); the AA shots are built up from both black and white samples, so when the pixels are averaged the net result is grey.
 
DaveBaumann said:
Zephyr said:
it can explain why MSAA can't resolve alpha texture problem, but why can MSAA resolve texture edge? Here are some examples, note the ceil lamps on the left:

Actually, here's a slightly clearer example taken from my look at WildcatIII's Super Scene Multisampling AA:

http://www.beyond3d.com/articles/wildcatiii/index.php?page=page8.inc

Specifically, take a look at the numbers on and next to the radio:

When you look at the geometry in that area it become easy to see why this occurs:

Hi, Dave, I have read ur "3Dlabs Wildcat III & SuperScene AntiAliasing", a nice article, and I understand ur pictures. However, i think ur pics are different with my pics. As U can see, lamps in Parhelia's 16x FAA picture without anti-aliasing prove that there are no triangles' edges crossing those lamps.
 
Zephyr said:
As U can see, lamps in Parhelia's 16x FAA picture without anti-aliasing prove that there are no triangles' edges crossing those lamps.
No, it only proves that Parhelia's 16xFAA is not able to cope with these polygon edges correctly.

btw, that R9700 6x shot is *really* impressive.
 
Xmas said:
Zephyr said:
As U can see, lamps in Parhelia's 16x FAA picture without anti-aliasing prove that there are no triangles' edges crossing those lamps.
No, it only proves that Parhelia's 16xFAA is not able to cope with these polygon edges correctly.

btw, that R9700 6x shot is *really* impressive.

Any proof?

I know 16x FAA in Parhelia has difficulties with scene using stencil buffer or using edge rendered in the form of texture, but it seems that JK2 does not use stencil buffer for those lamps.
 
As far as I know, 16xFAA also has issues with Alpha Blended edges. Haven't seen anyone investigate it though. The initial 3dmark screenshots from Matrox suggested it did.

ATi's 6x looks so much like 8x on the V56K, except on the 9700 it's actually usable. I think i *need* a 9700 :)
 
Colourless said:
As far as I know, 16xFAA also has issues with Alpha Blended edges. Haven't seen anyone investigate it though. The initial 3dmark screenshots from Matrox suggested it did.

ATi's 6x looks so much like 8x on the V56K, except on the 9700 it's actually usable. I think i *need* a 9700 :)

Of course, lamps in JK2 do not use "alpha blended edges", because if they do so, both G4Ti's 4x MSAA and R300's 4x MSAA can't deal with them correctly too.

As far as i know, R300's current driver cannot cope with alpha texture correctly. So be careful if u pay more attention to the quality of FSAA.
:D
 
Parhelia FAA also fails on random edges where it should work. So it can'y be used as a proof.

Actually, the fact that the lamps are aliased shows that they are done with separate polys. Since the texture filtering would have handled it far better than that. I can't see any indication from other parts of the screen that they had such extreme MIP LOD bias to cause that aliasing. Or are you implying that they use point sampling on that texture?
 
Alpha blended edges would be smooth even without any antialiasing (as long as you don't use point sampling).

Any proof?
Well, multisampling is a well-known algorithm. We know what it does and what it does not, so we can conclude that these edges must be either polygon (or intersection) edges, or stencil edges, because they're jagged without AA but smooth with AA. Other types of edges would be either jagged (alpha test) or smooth (alpha blended, edge inside a texture) in both cases. And since I find it very unlikely that JK2 uses stencil for these, I conclude that these edges are polygon edges.

Parhelia also totally fails to smooth polygon intersection edges.

Oh, and WRT to alpha textures on parhelia... well, these screenshots below show something really strange.
http://www.3dcenter.de/artikel/parhelia_firstlook/screenshot_geforce4ti_aa04_ultima9.php
http://www.3dcenter.de/artikel/parhelia_firstlook/screenshot_parhelia_aa16_ultima9.php
Apart from the fact that Parelia fails to antialias virtually all polygon edges here, it does smooth the edges of alpha textures (HUD, leaves) where GF4 doesn't. However, it blends these edges with the background (sky). So it would seem that FAA isn't really order independent and only keeps one depth value per edge pixel.
 
Zephyr said:
Of course, lamps in JK2 do not use "alpha blended edges", because if they do so, both G4Ti's 4x MSAA and R300's 4x MSAA can't deal with them correctly too.

I said Alpha BLENDED, not Alpha TESTED.

Also, I do not mean the edge formed within a triangle by an Alpha blended texture. I am refering to the edge of an Alpha blended triangle, whether that be the actual edge, or an edge created because of an intersection or overlap.

-Colourless
 
Now I know the key reason why I cannot understand the lamps aliasing smoothed by MSAA is my misunderstanding the alpha blended edges.

Alpha blended edges would be smooth even without any antialiasing (as long as you don't use point sampling).

That's to say alpha blended edges will be smoothed at least by bilinear filtering, right? (sure, we dont consider point sampling :) )

BTW, why the edge inside a texture is named by alpha blended edge?

I am refering to the edge of an Alpha blended triangle, whether that be the actual edge, or an edge created because of an intersection or overlap.

Here, could u give more details about the edge of an alpha blended triangle?
 
Zephyr said:
Here, could u give more details about the edge of an alpha blended triangle?

It's just any standard triangle, you know 3 edges, but is being alpha blended. Nothing special about that.
 
Back
Top