Certain benefits of the GeForce3 will not be realized until game developers begin to exploit new features through DirectX 8 or NVIDIA's new OpenGL extensions. In particular, the use of the programmable graphics processing unit and vertex and pixel shaders are necessary components to fully utilize the GeForce3. However, it's expected that it will take a minimum of six months before these types of games hit the market.
In the interim, we have a DirectX 8 based syntehtic benchmark that uses the capabilities of the GeForce3 - MadOnion's 3DMark2001.
MadOnion has been a leader in developing benchmarking software that contains the latest features of Microsoft's DirectX applications programming interface. The overall 3DMark2001 score is derived from the four game tests which simulate processing that may be used in future games. Since the game tests are based on a custom 3D graphics engine, they also provide a measurement of 3D graphics performance.
With 3DMark2001 and Remedy Entertainment's MAX-FX Technology, features of the GeForce3 can be put to the test which include transform and lighting, vertex and pixel shaders, environmental bump mapping, and even antialiasing.
3DMark2001 supports three rendering options which consist of:
- Pure Direct3D Hardware T&L - A new rendering pipeline in DirectX 8. When used, the graphics accelerator performs all transform, lighting, and vertex and pixel operations.
- Direct 3D Hardware T&L - This pipeline uses the 3D hardware to do transform and lighting. Vertex Shaders may be calculated on the CPU if they are not supported in hardware.
- D3D Software T&L - This pipeline uses the transform and lighting engine that's inside Microsoft DirectX 8. It is the default setting for all 3D accelerators that do not support Hardware T&L.
The following image lists all the 3DMark2001 test results based on the default benchmark settings from the GeForce3 and GeForce2 Ultra. Click to enlarge.
All Test Results - 1024x768@32bpp
In the non-game tests, 3DMark2001 gives the GeForce3 a significant edge in the single and multi-texturing fill rate tests as well as the DOT3 bumpmapping, vertex shader speed, and point sprite speed tests. It's also interesting that transform and lighting performance is faster on the GeForce2 Ultra than the GeForce3.
Note that Game 4 and pixel shader test results are not available on the GeForce2 Ultra as pixel shaders are not supported by the graphics hardware. According to the 3DMark2001 Testing & Test Related Questions FAQ, the Game 4 Test does affect the overall 3DMark2001 score:
You need a fully DirectX 8 compliant 3D accelerator in order to run Game Test 4. It will affect your 3DMark2001 score as much as one of the high-detail game tests.
Processor speed is a major factor in the performance of 3DMark2001 - especially in the Game 1 and 3 Tests. Therefore, performance results from a 1.3GHz Athlon processor have been included which are based on Sharky Extreme's GeForce3 preview.
The Game 1 Test uses Ipion real-time physics which was developed by Havok for the truck movements. Artificial intelligence is also used by the flying robots when chasing and firing at the truck.
The high detail mode of the Game 1 Test adds a variety of special effects and more high-detailed objects to the scene. The special effects include dynamic shadows, spotlights, and additional texture layers (up to three), for added realsim.
The dynamic shadows are done by rendering into textures. If rendering to textures is not supported by hardware, shadows are rendered with the CPU. Transformation of the dynamic shadow polygons is done using a vertex shader.
Game 1 Results - 1024x768@32bpp
With a slower processor, there's not much difference in performance between the GeForce3 and GeForce2 Ultra due to the heavily reliance on the CPU for the real-time physics and dynamic shadows. A fast processor provides a considerable performance boost in this test.
The Game 2 Test utilizes the Limigraph Global Illumination (GI) system which was developed by Remedy and Hybrid . Lighting is accomplished using multiplicative light maps which are not hardware accelerated by the GeForce. However the morphing effects, which include animations of the dragon, the bowmen, and the running people, are as they are accomplished using vertex shaders.
As with the Game 1 Test, the high detail scene in the Game 2 Test contains additional graphics-rich features such as rendering the entire dragon, a much larger village, additive light maps, and dynamic shadows (whose polygons are transformed using a vertex shader).
Game 2 Results - - 1024x768@32bpp
Unlike the Game 1 Test, the Game 2 Test is less dependent on the central processor. With a slower CPU, the GeForce3 shines and brings both systems up to the same performance level which is not the case with the GeForce2 Ultra.
All the characters in the Game 3 Test use vertex shader skinning except when a system has a DirectX 7 compliant 3D graphics accelerator, capable of hardware transformation and lighting. In this case, skinning is done using a custom skinning technique, which does skinning on the CPU, but transforms and illuminates the skinned vertices in the graphics hardware. The empty shells and discarded guns are controlled by Ipion real-time physics.
The high detail scene has the following additions compared to the low detail scene: Everything is reflected in the marble floor. All characters have dynamic shadows. Some shots that hit the walls shatter a part of the wall surface material.
Game 3 Results - 1024x768@32bpp
As with the Game 1 Test, processor speed plays a major role in this test. Also note that in the high detail test the 1.3GHz Athlon and GeForce2 Ultra (41.0 fps) easily outperforms the 735MHz Pentium 3 and GeForce3 (30.2 fps).
The 3DMark2001 results are a huge eye-opener as none of the high detail tests with either processor were able to achieve a frame rate of over 50 frames per second using the default benchmark settings. Compared to the Game 2 Test, the Game 1 and 3 Tests were developed to represent actual gameplay as a real-time physics engine is being utilized along with artificial intelligence.
There's no doubt that the graphics under high quality settings are top-notch, but it remains to be seen if 3DMark2001 and Remedy's MAX-FX technology will represent the anticipated performance of future DirectX 8 games. Only time will tell.
Next: AquaNox Performance