Since doing my last article, I had mixed feeling about it. On the one hand, I think I did a pretty good job of exposing
edge anti-aliasing. Obviously, this is just one aspect of the overall feature. You could literally do about 5 good articles,
covering all the different aspects of FSAA between the GF3 and 8500. Unfortunately, one only has so much time.
Today, I'm not going to concentrate on a comparison between these 2 products, but a rather interesting development. One
of my close partners, Alexey Nicolaychuk, author of the very excellent RivaTuner,
approached me about an interesting find: making use of Super-Sampling on the GeForce3. Alexey and
I have had numerous exchanges over the last several months,
and he had allowed me access to Beta versions of his utility. We thought this would make for an interesting article,
so he agreed to let me prepare this article before releasing his version which exposes this feature.
There are a couple of drawbacks in using this feature, however. Namely, no GeForce3 features. What you basically are doing is
instructing the driver to treat the GeForce3 as a NV10/15 chip. So, you won't have the robust GF3 features, such as
Lightspeed Memory, occlusion detection, etc. At the end of the day, you basically have a REAL expensive, high-octane
In addition, this can only be applied to OpenGL. Despite the drawbacks, the nice thing about this unsupported-feature
is the fact that it can do wonders for some of the older-generation games, particularly those that make heavy use
of Alpha-Textures. In case you aren't aware of this, the GeForce3 uses Multi-Sampling to perform Full-Scene Anti-Aliasing,
and one thing that MSAA hardware cannot handle are Alpha-Textures. When you see GF3 FSAA screenshots that don't seem to
anti-alias certain textures of a game, you can be pretty well assured that these are, indeed, Alpha-Textures.
The goal of this article is twofold: Compare Super-Sampling to Multi-Sampling on the GeForce3, and to show the benefits/
drawbacks of these 2 implementations.