Test Descriptions... Overall Performance: These tests are used to measure your system's overall performance and must be run in order for the score to be calculated. The emphasis is to stress the factors that are required for good game performance. To be able to measure correctly the amount of time that it really takes to render one frame, we will lock the frame buffer immediately when it's possible, to be sure the rendering of that frame has ended. By using this method we will ensure that the monitor's refresh rate or disabling vertical sync has no effect on the results. The drawback of this method is that the frame rates of the game scenes are lower than with most of the inaccurate benchmarks that are built into games. Wherever possible, the tests take full advantage of DirectX 6 features such as Vertex Buffers and Multitexturing. Currently 3DMark 99 has no direct support for AMD K6-2 3DNow! instructions. Game 1 - The Race These test simulates a game with single rendering pass. Game 2 - First Person These test simulates a game with two rendering passes. CPU Geometry Speed This test determines your system processors' maximum 3D performance, independently of your 3D accelerator.It shows the potential performance of your system in 3D graphics. Fill Rate This test renders the screen four times each frame. At first the single texture fill rate is measured by drawing four times overlapping polygons. The second fill rate test - "with multitexturing" creates a material that uses just one polygon layer with four texture layers on it. Texture Rendering Speed This test shows what are the maximum frame rates on your accelerator when the scene has different (2, 4, 8, 16, 32 MB) amounts of textures. Theoretical Speed Tests: (only available in 3DMark 99 Pro) These tests do not affect the Overall Scores directly but are good indicators of certain areas of your 3D hardware performance. Texture Filtering Speed n Pixel Polygons Image Quality Tests: These tests help you to analyse the image quality of your 3D accelerator and compare it to other accelerators Image Quality Test 1 - Engineering Image Quality Image Quality Test 2 - Game Image Quality

 

Game 1 - The Race The first game test simulates the graphics in a racing game. The landscape is drawn using single pass rendering (one texture only). Transparency is heavily utilized for different effects. Scene parameters: Duration: 18 seconds Vertices: 4134 Triangles: 6648 Textures: 2600 KB in 16bit format with mipmaps. (Maximum size 256x256) Features: - Trilinear filtered texture mapping - Gouraud shading - Vertex Fog - Alpha Blend - Alpha Blend + Vertex Alpha used to fog the alpha blended textures - Additive Alpha + gouraud to fog the additive alpha textures The result is a harmonic mean of the frame rate (FPS - Frames Per Second).

Game 2 - First Person The second game test simulates the graphics in a typical first person shooter. Every surface in the corridors has two textures on top of each other, one for the bitmap, one for lighting (light map). Scene parameters: Duration: 20 seconds Vertices: 6104 Triangles: 4033 Textures: 3146 KB in 16bit format with mip-maps. (Maximum size 256x256) Features: - Trilinear filtering - Texture + Multiplicative alpha blending (scene shadow maps) - Texture + Additive alpha blending (weapon reflection) - Alpha Blending (particles, on-screen displays) Uses exit technology for visibility optimization. The result is the harmonic mean of the frame rate (FPS - Frames Per Second).

CPU Geometry Speed

This test measures your CPU's maximum performance in real-time 3D. It runs both of the game tests described above without sending anything to the 3D accelerator and gives a performance score based on the "frame rates". As the CPU's ability to process 3D geometry quickly enough affects everything, the CPU score is quite important factor in the system's overall performance. The CPU Geometry Speed score and 3DMark Score are not directly comparable.

 

Texture Rendering Speed The texture rendering speed test measures capability of your 3D accelerator to handle different amounts of textures simultaneously. There are five separate tests with different texture amounts: 2MB, 4MB, 8MB, 16MB and 32MB. The test generates as many 256x256 textures as are required for each texture amount, in the desired bit depth (16 bit textures if the screen is in 16 bit color, 32 bit textures if the screen is in 32 bit color ). Mip-mapping is disabled (no mipmaps are generated). The test output might vary according to the 3D Accelerator. This is because the test always tries to generate the closest possible amount of textures for the currently selected MegaByte amount. Eg, if the benchmark is run in 32bit color (using 32-bit textures), the number of textures that fill up 16MB is smaller than if the test was using 16-bit textures. If the system you are benchmarking on has 64MB or less system memory, the 32MB texture rendering speed test is skipped. The result is the harmonic mean of the frame rate (FPS).

Fill Rate The test consists of 4 geometry layers. Each geometry layer consists of 2 polygons, filling the screen completely. Every geometry layer has the same texture. The test uses separated geometry so that the "Multi-texturing" features could not be used. --------------- layer 4 additive --------------- layer 3 additive --------------- layer 2 additive --------------- layer 1 copy   Fill Rate using Multi-Texturing This test is similar to the above, except that it attempts to use "multi-texturing". There is only one geometry layer having four textures mapped onto it. The textures are constructed as a single material and optimized to use as few passes as possible on different hardware. If the hardware supports single pass dual texturing, this test uses it. In theory if there would be hardware that could do four textures in single pass, this test would will take advantage of it. If no single pass multitexturing capabilities were found, the test reverts to doing multiple passes. --------------- layer 1 additive + additive + additive + copy The Fill Rate tests use Z-buffer, no mipmaps, bilinear filtering. The texel / pixel ratio varies when resolution is changed, just as in games when the games are run in higher resolution. A texel is a "texturized pixel". In the Fill Rate tests, each physical screen pixel gets four textures on it, so each individual pixel has four texels on it. For example: A 800 * 600 Resolution screen has 480,000 individual pixels on it. This test fills each of them 4 times. So the screen has: 4 * 800 * 600 = 1,920,000 texels on it. The result is given in Mega texels per second.

Texture Filtering Speed This test compares the rendering speed difference between point sampling, bilinear filtering, trilinear filtering and anisotropic filtering. Results are given as speed difference percentages compared with bilinear filtering. If the accelerator (or the driver) does not support bilinear filtering, these tests are skipped.