[help index]

Results

3DMark outputs several different results. Here is a small explanation of the terms.

Test type:
3DMark Result

Synthetic CPU 3D Speed
Rasterizer Score

Game Scenes
Fill Rate
Texture Rendering Speed
Texture Filtering Speed
Bump Mapping
N Pixel Polygons

Unit:
3DMarks
CPU 3DMarks
3DRasterMarks

FPS
Mtexels/s
FPS
% of Bi-Linear
FPS
Kpolygons/s

The tests which rely on how long each frame took to render, the values are calculated by using standard Harmonic Mean formula:

harmonic-formula.gif (2690 bytes)

For a more detailed description of the testing process, please see 3DMark 99 Help.

[help index] [top of page]
3DMark Result [ 3DMarks ]

Is calculated from the frame rates of the two Game Scenes. The result is in 3DMarks.

The 3DMark Result shows how the 3D accelerator and the system will generally perform in games.

Next...

 

[help index] [top of page]
Synthetic CPU 3D Speed [ CPU 3DMarks ]

Is calculated from the "frame rate" of Synthetic CPU 3D Speed test data (described better in 3DMark Test Descriptions). The test itself uses a "null driver" that simulates running synthetic data with very heavy polygon and light counts but doesn't actually send anything to the 3D Accelerator.

The Synthetic CPU 3D Speed test shows only the pure CPU speed and is not affected by the 3D Accelerator at all.

Next...

[help index] [top of page]
Rasterizer Score [ 3DRasterMarks ]

Is calculated from the Texture Rendering Speed tests and the Fill Rate Tests and shows the theoretical 3D accelerator performance. The Rasterizer Score is affected very little by CPU Speed.

Next...

[help index] [top of page]
Game Scenes [ frames per second ]

Game Scene 1 - The Race uses single texturing (only one bitmap layer for each polygon) and shows how the card will perform in games using only one texture pass, such as Incoming, Turok or Forsaken.

Game Scene 2 - First Person uses dual texturing (two bitmaps for each polygon). In this case the dual texturing is used in the walls, which have a base texture and a shadow map on top of it. The weapon has a shiny reflection texture on top of the normal weapon texture. Multi-texturing has not been possible with DirectX until version 6.0. This test shows how the card will perform in games that use several texture passes, such as Quake II and Unreal.

Next...

[help index] [top of page]
Fill Rate Tests [ millions of texels per second ]

Fill Rate tests how quicky the 3D Accelerator is able to draw (fill) the screen. The higher the fill rate, the more depth complexity and better resolutions can the card handle.

Fill Rate with Multitexturing uses the new DirectX6 multitexturing features. If the hardware is capable in drawing more than one texture in one pass, this test will show it. Many modern games take advantage of Multitexturing, for example in creating darkening shadow maps or brightening reflection maps. If the hardware tested does not support multitexturing, the test will revert to rendering the polygons as seperate layers..

Next...

[help index] [top of page]
Texture Rendering Speed [ frames per second ]

These tests show what kind of a framerate the 3D Accelerator is able to maintain when the scene has different amounts of textures visible. The geometry of the scene is kept to the minimum.

Next...

[help index] [top of page]
Bump Mapping [ frames per second ]

These tests show how fast the 3D Accelerator is capable of doing emboss bump mapping.

Next...

[help index] [top of page]
Texture Filtering Speed [ % of the speed of bi-linear filtering ]

The values of these tests are compared to the speed of mip-mapped bi-linear filtering. These tests show how much the 3D Accelerator gets slowed down by using tri-linear or anisotrophic filtering. The point sample texture filtering shows if the the accelerator would be any faster without any filtering.

Next...

[help index] [top of page]
n Pixel Polygons  [ thousands of polygons per second ]

The n Pixel Polygon test results show what is the 3D Accelerator's theoretical speed in drawing polygons of a certain size.

Next...