Futuremark's 3DMark05

Introduction

If there is one benchmark that has stood tall over the years, it is the 3DMark series from Futuremark. Ever since the release of 3DMark99 back in 1998, Futuremark has continued to provide us users (and reviewers) with a way to compare our video cards in a reproducible way.  True, the benchmarks have been surrounded with some controversy the last few years, and many have questioned the use of synthetic benchmarks. The controversy hasn’t changed the fact that the 3DMark programs have always been the benchmark to use when you want to brag: “I got 11,000 in 3DMark03, what did you get? 8000? Hah, that’s … pretty good …. (I rule!)”.

Today, we will be looking at Futuremark's latest in video benchmarking: 3DMark05.

Bringing Your Video Card to Its Knees for the Last 6 Years

3DMark benchmarks have always been designed to get the most out of your video card.

From Futuremark’s whitepaper:

“3DMark99 concentrated on measuring fixed function vertex transformation and lighting, and multitexturing.  3DMark2000 added support for graphics hardware supporting transformation and lighting, and the complexity of the game tests was increased. 3DMark2001 increased the complexity of the fixed function game scenes to tens of thousands polygons per frame on average and also introduced shader technology. The scenes mainly used fixed function vertex and pixel processing, while shaders were used for special effects. There was skinning, morphing and massive amount of animated grass and leaves, all using 1.1 vertex shaders. Game test 4 presented the first higher level material using a 1.1 pixel shader.  3DMark03 concentrated on testing the 1.x and 2.0 shader model. Only one game test, meant for legacy systems, offered fixed function multitexturing, while the other three used pixel shaders for all materials. All vertex processing used vertex shaders, mainly of the 1.x model. The last game test presented the first vertex and pixel shaders of shader model 2, while the majority of the shaders in that test still were of the 1.x model. The scene complexity was raised to several hundred thousand polygons per frame on average.”

Note that it wouldn't have been a smart move by Futuremark to require a DirectX 9 video card for 3DMark03 since those cards still weren’t that common at that time. Today, things look different though. While there still are a lot of DX8.1 video cards out there, DX9 cards have been available for quite a long time. With 3DMark05, Futuremark has brought us the first benchmark that requires a DX9 based video card with support for at least Shader Model 2.0.

3DMark05 extensively uses Shader Model 2 and 3 for all vertex and pixel processing. You can also run all the compatible shaders with SM 2.0a and SM 2.0b.