3DMark05 First Look

Written By : Jeff Nettleton and Trevor Haag
September 2004

Introduction:::...

The buzz on the net was that Futuremark would release 3DMark05 soon, many had speculated its arrival and now it's here.

I must say I didn't hear too much about it until some sites reported release dates for the benchmark and there were some leaked images on the web, some 2 weeks in advance of the release date.
So I figured the release would be a long way from now, I mean they wouldn't call it 3DMark05 for nothing, but here it is.

Futuremark started off with the release of XL-R8R, back when they were called the mad onion.
Each year another benchmark was released, 3DMark99 in October of 1998, then came 3DMark 2000, followed by 3DMark 2001 and then 3DMark03.
3DMark03 and 3DMark05 weren't released the next year, but instead after 2 years.
This shows that the benchmarks grew and matured in both size and shape.

Of course the benchmarks also represent the close following of Microsoft's DirectX API.
Each version of 3DMark is able to benchmark the then current Microsoft API to the fullest capacity.
Here is a list of each version of 3DMark and its intended designed DirectX version:

- 3DMark99 designed for DirectX 6
- 3DMark 2000 designed for DirectX 7
- 3DMark 2001 SE designed for DirectX 8.1
- 3DMark03 designed for DirectX 9.0a
- 3DMark05 designed for DirectX 9.0c

Some of the benchmarks still support and test older DirectX versions.

Remarkable is the fact that this benchmark uses an entire new engine rather than the Max-FX 3d engine used previously. This is a more 'game like' engine using the CPU for preparing and optimizing the content to be rendered for the graphics hardware.
There are only rendering related tasks for the CPU in the Game Tests, no other CPU tasks typical for games are utilised though as Futuremark go to great lengths to point out, this is how the vast majority of benchmarks in games work too.

 
Website Design and Graphics Copyright Wayne Brooker 2004
All images Copyright 3DVelocity.com unless otherwise stated