Shimmering

[Ryu Connor]

Since there seems to be some confusion about subpixel data let me post this and see if makes things clearer.

http://www.beyond3d.com/articles/wildcatiii/index.php?page=page3.inc

The submarine model in this demo is designed such that cards with a lower subpixel accuracy will suffer from rounding errors which will appear as gaps in adjacent polygons on the model; in motion these gaps appear as pixel pops, and the more there is the more distracting the rendered image becomes. You can see in the case of the WildcatIII there are still few pixel pops visible, but these are greatly reduced when looking in comparison to the Radeon.

There is a difference between improved texture quality because of finer subpixel data and actually filtering the texture during AA sampling.

[az]

Since when is the GF3's and 4's FSAA edge-only?

AZ

[Ryu Connor]

Since when is the GF3's and 4's FSAA edge-only?

Has been since release. AA is a rather complex subject though and between pixels, subpixels, textures, OGSS, RGSS, Quincunx, and Smoothvision, people have gotten rather confused.

There are some good articles out there that detail the right story, but getting people realize what they had understood before is wrong is rather difficult. And given the relative complexity of the subject it's not a huge surprise that people get confused.

[[Ch]amsalot]

Ryu,

RGSS is a misnomer. The Voodoo5 does not render a "super-sample" or a larger resolution internal image. It works with the same size scenes, renders either two or four of them internally and at different pixel shifts, and then combines them.

GeForce3 does both multisampling and supersampling, depending on the mode selected as I've set forth above.

-[Ch]ams

[Ryu Connor]

http://firingsquad.gamers.com/hardware/multisamp/default.as]

Anyone who has any experience with super-sampling is well aware that it presents a performance issue. In 2x2 (4x) mode, a 75% fill-rate hit is incurred. This, of course, is because we are rendering at what is effectively four times our base resolution (the base resolution is the resolution that we begin with, and also the resolution that we display on screen). In applications where fill-rate is our limitation, this can be a serious problem as it cuts our frame-rate by a theoretical 75%.

http://www.tech-report.com/reviews/2001q2/geforce3/index.x?pg=16

3dfx's approach with the Voodoo 5. This hybrid approach is a modified version of OGSS. Rather than doing its supersampling on an ordered grid, the Voodoo 5 renders its supersample images on a grid with a slight tilt angle to it. In the case of 4X FSAA, the samples are "jittered" from one sample to the next in a diamond-shaped pattern. This slightly "off" pattern doesn't correspond so neatly to the display's grid of pixels, making it harder for the eye to recognize aliasing patterns.

At a given sample size, RGSS is likely to produce results slightly superior to OGSS. However, sample size matters more than the difference between these two techniques. 4X OGSS almost always looks superior to 2X RGSS, no matter what the 3dfx fanboys would have you believe.

The engineers that designed it didn't name it Rotated Grid Super Sampling just to be contradictory.

[[Ch]amsalot]

Ryu,

Damn it I'm on dial up and you have the speed advantage.

I'll either find a link to support what I'm saying about the Voodoo5's MULTIsampling or I'll eat my PCMCIA modem.

Do you agree with what I've said about GF3 using both supersampling and multisampling or should I find a link for that too?

-[Ch]ams

[Ryu Connor]

I'll either find a link to support what I'm saying about the Voodoo5's MULTIsampling or I'll eat my PCMCIA modem.

That's probably not a good idea. We'd have to rush you to the hospital.

Do you agree with what I've said about GF3 using both supersampling and multisampling or should I find a link for that too?

It only does mutlisampling, from a link I posted earlier:

http://www.tech-report.com/reviews/2001q4/r200/index.x?pg=6

Multi-sampling, on the other hand, concentrates mainly on anti-aliasing object edges. Depending on the particular implementation, multi-sampling may or may not filter textures. In the GeForce3, textures are not anti-aliased.

He's even got pictures showing the image impact difference between supersampling and multisampling.

[[Ch]amsalot]

GeForce3 does both, depending on whether you choose 2x (supersampling), 4x (supersampling) or Quincunx (mutisampling).

Support:

http://www.hardcoreware.net/reviews/video_cards/chaintech_ti200se/5.php

and

http://www.geocities.com/tnaw_xtennis/NV20-11.htm#16

Still looking for support on my position on Voodoo5's FSAA. Again, my position is if it's not rendering the sample at a larger scale, it's not supersampling. My understanding is that it's rendering the same sized scene (either twice or four times) at different shifts. That's not super-sampling in the true meaning of the word.

Supportive links OTW.

-[Ch]ams

[[Ch]amsalot]

To add to the confusion:

Here is an excerpt from The Tech Report discussing FSAA. He calls the GeForce3's FSAA "MULTISAMPLING" but then talks about it using a Rotated Grid

[editing in process- damn toshiba notebook buttons!]
[damn sticky eraserhead]

THIS IS TORTURE

[Ryu Connor]

They are wrong. This just isn't going to work. I'm going to have to dig up a white paper.

Here is an excerpt from The Tech Report discussing FSAA. He calls the GeForce3's FSAA "MULTISAMPLING" but then talks about it using a Rotated Grid

That's because multi-sampling can apply its data in either rotated grid of ordered grid. Not because it's doing super sampling.

Read here:

http://www.tech-report.com/reviews/2001q2/geforce3/index.x?pg=16

[[Ch]amsalot]

The white paper is a good idea.

The more I search the more conflicting info I obtain:

http://www.xbitlabs.com/video/geforce4/

http://www.tech-report.com/etc/2002q1/geforce4/index.x?pg=3

The tech guide review clearly says rotated grid sample pattern but just about in the same breath he is calling it multisampling. That would support my position that Voodoo5's rotated grid sample pattern is multisampling.

The review from x-bit is just giving me a headache at 800x600. Why do I bother using this notebook?

Still searching for more links... but maybe going to NVIDIA's white pages is the way to go...

[Ryu Connor]

http://firingsquad.gamers.com/hardware/multisamp/page5.asp

Notice that we have made distinctions between this new multi-sampling anti-aliasing and the multi-sampling that is found on the Voodoo5. This issue has been a hotly discussed, online topic.
The Voodoo5 renders multiple samples (hence it is sometimes also referred to as multi-sampling) to achieve a form of super-sampling. Super-sampling, in its simplest of definitions, is combining multiple samples to achieve each pixel. The Voodoo5 does this. On a Voodoo5, you will render your scene multiple times and buffer these into four (or however many are supported in hardware) separate buffers.

As the scene is sent to the screen there is a combine unit that averages each of the four buffers at a sub-pixel level to produce the final, displayed image. As we can see, this is considerably different than the form of multi-sampling that we are discussing here. The terms themselves can certainly be confusing, but everything makes a lot more sense once we understand exactly what we kind of multi-sampling we're talking about.

Notice. The form of mutlisampling discussed in the above article is not the same as the implementation used in the Geforce 3 and 4 line.

Again to reiterate:

V5 does super sampling, which is a full scene AA as it filters the textures.
GeForce 3 and 4 do multisampling, which is a edge based AA as it does not filter textures.

[[Ch]amsalot]

%#(% so-called "TECH" sites:

From Anandtech:
"Although the GeForce3 core supports both supersampling and multisampling, the currently available drivers only support multisampling which is by far the most efficient method of AA on the GeForce3. We incorrectly assumed that the 4X AA mode on the GeForce3 was using the supersampling capabilities of the core and thus labeled it as so in our last article. "

http://www.anandtech.com/showdoc.html?i=1562&p=2

See also nice chart at:

http://www.3dconcept.ch/chips/chiptable6th.htm

If I convince you that the Voodoo5 is really multisampling, can I just eat half my PCMCIA modem? I have some Arctic SilverIII left over so it should go down smoooth..

-[Ch]ams

[Allankyoto]

So now that you guys are done arguing the technicals could someone please clarify this for me. Does the Parhelia get rid of texture shimmering with it's form of AA? That's all I want to know. Until a card does that I am not interested in upgrading from my Geforce 3.

[Ryu Connor]

Does the Parhelia get rid of texture shimmering with it's form of AA?

Yes, but only with it's 4x FSAA option. The 16x FAA option will not.