Texture Filtering Performance
Here we'll take a look the the rendering performance of the Radeon X800's under a few different texturing scenarios.
| RightMark - Lighting (Phong) - PS2.0 | 1X | 2X | 4X | 8X | 16X |
| X800 XT PE | 277.9 | 277.8 | 277.7 | 277.6 | 277.6 |
| X800 PRO | 190.8 | 190.5 | 190.6 | 190.5 | 190.5 |
| % Drop from 1X | 2X | 4X | 8X | 16X |
| X800 XT PE | 0.0% | -0.1% | -0.1% | -0.1% |
| X800 PRO | -0.2% | -0.1% | -0.2% | -0.2% |
| % Drop from Previous | 2X | 4X | 8X | 16X |
| X800 XT PE | 0.0% | 0.0% | 0.0% | 0.0% |
| X800 PRO | -0.2% | 0.1% | -0.1% | 0.0% |
Looking at a theoretical case of just a pure pixel shader test, we can see that with a separate texture address processor and effective instruction scheduling, enabling Anisotropic filtering on the X800 boards needn’t incur any performance penalty at all.
| 3DMark03 - GT2, 1600x1200 (FPS) | 1X | 2X | 4X | 8X | 16X |
| X800 XT PE | 48.8 | 47.2 | 44.8 | 41.8 | 40.5 |
| X800 PRO | 35.0 | 33.5 | 31.6 | 29.8 | 28.6 |
| % Drop from 1X | 2X | 4X | 8X | 16X |
| X800 XT PE | -3.3% | -8.2% | -14.3% | -17.0% |
| X800 PRO | -4.3% | -9.7% | -14.9% | -18.3% |
| % Drop from Previous | 2X | 4X | 8X | 16X |
| X800 XT PE | -3.3% | -5.1% | -6.7% | -3.1% |
| X800 PRO | -4.3% | -5.7% | -5.7% | -4.0% |
With the more varied environment of 3DMark03’s GT2 test we can see that there is more of a performance drop from 1X Anisotropic Filtering to 16X, up to 17% for the X800 XT PE and 18% for the PRO.
