By volt - April 12, 2003

Synthetic Benchmarks

Although synthetic benchmarks do not represent actual performance in current games, it is none the less a good way to see how your graphics card stacks up against other competitors.

3DMark®2001 SE Pro is the latest installment in the popular 3DMark series. By combining DirectX®8.1 support with completely new graphics, it continues to provide benchmark results that empower you to make informed hardware assessments. The game tests, texturing tests, filtering tests, image quality tests, and others give you an overview of your system’s current performance and show you what kind of performance you could expect if you were to upgrade your PC.

Key 3DMark2001 SE Pro Features

• Support for Microsoft DirectX®8.1.
• A new DirectX8.1 feature test showing Vertex Shaders & Pixel Shaders 1.4. *
• Featuring 3 game tests plus a 4th game test using DirectX®8.0 hardware accelerated features. *
• Directx8.0 feature tests showing Vertex Shaders, Pixel Shaders 1.1 and Point Sprites. *
• Two game tests use Ipion real-time physics by Havok.
• Online services. Submit your results and compare them to 3DMark2001 SE results from all over the world.

*A fully DX8 compliant 3D accelerator is required for running this test.

Both 3DMark scores are very close to each other. The fillrates are a bit larger on the Ti4600, but then again that card is clocked 306MHz core / 661MHz memory against 250MHz core / 513MHz memory for the Ti4200. Our little "Devil" beat his older brother in Game 1 Car Chase - High Detail as well as Game 3 Lobby - High Detail. Let's move down to Environment Bump Mapping. Although the fillrates on our card are lower than on the Ti4600 AND we are using a slower processor, the card somehow managed to win in this test. In the rest of the benchmarks, the Ti4600 wins marginally.

GL Excess is my third attempt at OpenGL, the first being a simple driving simulation I wrote for a classwork at university and the second being Rebirth saver. The idea of making a huge demo was born right after Rebirth v1.1 was released: basically I wanted to develop some techniques I had already included in my past projects and challenge myself learning new OpenGL skills. I immediately decided this would be a multi-scene demo so that I could focalize my attention on different techniques at different times and then put them all together in no particular order.

-Paolo "Bustard" Martella

This is a true OGL based benchmark and it's getting very popular among reviewers and benchmarkers. For our purpose, we have used version 1.1a. Here are some updates from version 1.1:

• Automatic VSYNC management
• Command line arguments added
• XSscript scripting utility added
• Bug fixes
  

P4: Stock Ti4600 + P4 1.6 @ 2.30GHz - GeForce4 Ti 4600/AGP/SSE2
AMD: Ti4200 Stock + AXP 1900+ @ 1.6GHz - GeForce4 Ti 4200 with AGP8X/AGP/SSE/3DNOW!

We ran the tests only at one mode, which will give a good representation of how our card performs against the faster system. The scores vary, but I would like to stress that the Ti4600 is factory clocked at 306MHz core / 661MHz memory.

The Codecreatures engine is far more than a rendering API. The organisations of complex, effect-loaded virtual game worlds and contents with an elaborated AI are the standard in Codecreatures. The Codecreatures engine is based on a totally new architecture, specially designed in the prospect of further extensions and is open for many different areas of usage. Further extensions are planned for the future, so that Codecreatures can be established as a basis for a wide area of genres.

CodeCretures is good demo benchmark that heavily utilizes Vertex and Pixel shaders. You will see low scores because this engine uses high resolution textures (high polygon count) and programmable shaders which aren't widely used yet, ie. not all games use them.

This particular benchmark is not really CPU limited. It's more GPU dependent. Although the Ti4600 (28.2 FPS) is a faster card, our Ti4200 (24.5 FPS) stacked pretty good at 1024x768 resolution. Now let's look at 1280x768 scores. Here we have a different story. That extra speed really helped to push those polygons with an average of 23 FPS.

ChameleonMark is a performance tool for measuring pixel shader performance for a variety of shaders. The test uses three of the Chameleon "skins". These are the "real", "shiny" and "glass" skins.

Note that Pixel and Vertex shaders (PS 1.1, VS 1.1 respectively) were first introduced in GeForce3 line of GPUs. GeForce 4 Ti4x00 series uses improved Pixel shader (PS 1.3) though Vertex shaders are still version 1.1.

P4: Stock Ti4600 + P4 1.6 @ 2.30GHz - GeForce4 Ti 4600/AGP/SSE2
AMD: Ti4200 Stock + AXP 1900+ @ 1.6GHz - GeForce4 Ti 4200 with AGP8X/AGP/SSE/3DNOW!

Since the benchmark measures shader performance the clock speeds are key factors for achieving higher FPS. Again I'm not a fan of synthetic benchmarks. Those usually show how games may look like in future and that's about it. There are huge communities battling with getting the highest score, but is it worth it? That's a matter of preference. In my opinion too many people are obsessed with such benchmarks (*cough* 3DMark *cough*).

Nothing else to see here...move along :).