First Look: NVIDIA GeForce 3 (4/6) Programmable Vertex Shaders The vertex shaders allow the GeForce 3 to calculate transformations - changes to a 3D object - completely on-chip, becoming independent of the CPU. This allows for a whole range of effects, like a furrowed forehead or a flag flying in the wind or even motion blur effects.
When the Vertex Shader is used, it completely replaces the T&L unit. This is reported to be one of the major drawbacks of the GeForce 3 in its current form, as only either / or scenarios work for any given scene. This means that Vertex Shader and T&L unit can't work in conjuntion, only exclusively. Programmable Pixel Shaders Pixel Shaders were already a feature of the GeForce 2 family and aren't really anything new. They can create some nice effects like environment mapping or bump mapping. The GeForce 2's Pixel Shaders were hardwired, so to speak, making them rather inflexible. The GeForce 3's Pixel Shaders can be programmed in a language similar to an assembler. The technology, dubbed Pixel Forge, allows a number of interesting effects, from comic rendering to reflections on rough (bumped) surfaces. The GeForce 3 is able to process 4 textures per clockcycle or pass (single pass multitexturing) - the GeForce 2 can only handle 2.
Id software's John Carmack, gaming god and coder supreme, took a relatively sceptic stance towards these features in his .plan, since the number of available commands is quite limited:
Meanwhile, other developers seem rather taken with this feature. Again, it remains to be seen what the developers choose to create with the tools available to them. Some examples:
Fisheye Lens: The view is distorted using a vertex.
Wave: A Vertex Shader effect is used on a flat surface to create a typically wave-like rippling effect.
Bump Reflection: This effect applies bump mapping to a surface or plane and a real(time) reflection is added. More examples will follow in our upcoming Shader Special. Anti Aliasing NVIDIA never fully realized the marketing potential of this feature. Now, with the GeForce 3, NVIDIA is hopping onto 3dfx's FSAA bandwagon of hype. Instead of the more memry intensive SuperSampling, the GeForce 3 comes equipped with the more "intelligent" MultiSampling method. This is supposed to guarantee acceptable performance even at high resolutions. NVIDIA claims framerates of around 70fps at 1024x786-32 at 2x FSAA. Another new development is the unpronouncable Quincunx technology, which is supposed to allow for 4x FSAA at 2x FSAA performance. Well, the only way to really judge that would be with a testing sample, though...*sigh* Erstwhile Impressions It's hard to come to any substantial comclusions at this point. At the official European GeForce 3 launch on February 27th in Paris, NVIDIA did show some very impressive demos. These consisted of some gorgeous upcoming DX8 gaming titles (Aquanox and Grint), the visually spectacular 3DMark2001 and some NVIDIA techdemos, just to name a few. This time around, the wait for real world applications and games that take advantage of GeForce 3 technology won't be long, since the Xbox, for which literally hundreds of titles are in the works, will use an almost similar GeForce 3 chip as well. But of course: You do not need DX8 to use the new features! Every new effect does also work in OpenGL! Nonetheless, with initial prices of around 1200 DM ($600-700), the GeForce 3 will remain a luxury ware. But enough of whitepapers, bland theory and unproven speculation! Now for something more concrete: let me treat you to shots and specs of some of the upcoming GeForce 3 boards by ASUS, ELSA, and Hercules as well as some pictures of the presentation in Paris! |
|
||||
Copyright: 27.02.2001 -
RIVA
Station 2001 - Lars Weinand Translation by Benjamin Kraft URL of this Article: www.rivastation.com/gf3_e.htm - If you want to link to it, please use this URL! :-) |
 |
