RSS Feed

HCW Tech Blog

For the latest info on computer hardware, tech, news, video games, software tips, and Linux, check out our new improved front page: HCW Tech Blog

Reviewed by: Bryan Pizzuti [04.22.02]
Manufactured by: Chaintech


Real-world tests

Dedicated benchmark programs alone do not mean a card is going to perform great, so we also test using actual game engines. We decided to NOT use results from Serious Sam: The Second Encounter, because the results we repeatedly got were very anomalous, and we could not locate an explanation; basically, the Kyro2 ran away with it, but there was no apparent reason why it would do so well, unlike in VillageMark, which uses a huge amount of overdraw to test Z-Culling. So until we are able to reasonably explain test results with Sam2, we will not be using it.

OpenGL - Star Trek Voyager: Elite Force

We simply can NOT let a card get away without being subjected to the Quake3 engine. Out of personal preference, I use Activision's Voyager: Elite Force game, patched to version 1.2, running the internal demo0001, which is a long, involved demo consisting of both closed-in passageways and open-field combat.

These results are EXTREMELY consistent, if you'll notice. At every resolution, the Kyro2 is about 30-35 fps behind the TI200.

Enhanced OpenGL - DroneZ

This is a wonderful benchmark to run, not only because it's visually pleasing, but because it's highly configurable, and will provide a massive amount of data concerning performance. I ran this one in 2 ways for the TI200. The "Normal" mode is where the settings were exactly the same as the the Kyro 2, and the "GF3 Enh" mode turned on all of the enhanced settings that the GeForce3 GPU core offers. I've detailed the specifics below.

Standard Mode (common GeForce3/Kyro2 options activated):

Texture Compression
World specular
Vertex blend
Trilinear filtering
High texture definition
S3TC texture compression
Compiled vertex arrays
Packed pixels
3D textures
Secondary color
Swap control

GeForce3 Enhanced tests turned on the following additional options:

Vertex blend
Texture Cube maps
Register Combiners
Texture shaders
Vertex programs

The reasoning behind this is that when a user buys a video card, they aren't going to play a new game at old-card settings in most cases. They're going to turn on every new feature that their video card might support, and I felt we should also show some results that factor this in. Besides which, the TI200's advanced features are it's main selling point, and it would be somewhat irresponsible to completely ignore them while testing. So here are the max frames-per-second results:

Well, in this case only at 800x600 does the addition of the extra settings impact frame rate to any great degree. That might be due to the game being CPU limited, but we can prove that it isn't with the other data that this benchmark provides, specifically, the T&L polygons per second and GL polygon per second data. Again, we're using the max figures here for consistency.

Well, the first thing we can see through the T&L data is how much less the Kyro2 can process without it's own hardware T&L core, which is interesting. The TI200, when on a level playing field with the Kyro2 can do twice as many T&L calculations, and also render twice as many ordinary GL triangles. But the biggest deal is when turning on the enhanced GeForce3-compatible functions such as pixel and vertex shaders. Notice that the number of T&L triangles nearly doubles, and the GL triangles figure nearly QUADRUPLES, making it pretty much equal to the T&L triangles figure. Basically, what can be gleaned from this is that the nFiniteFX engine is able to generate MANY more triangles when enabling it's new shading features, and more triangles means more detail.

Next Page: (6)