- Date:
- Wednesday, October 15, 2003
- Author:
- Brent Justice
- Editor:
- Kyle Bennett
- Google +1

ATI Radeon 9600XT Review
We have ATI’s new mainstream video card in the house, the 9600XT. We run it through performance tests and look at the gaming experience delivered with this brand new video card and compare it with the competition.
Halo: Combat Evolved
(DirectX 9)
Halo, developed by Gearbox Software and published by Microsoft, has been out for a while now on the X-Box. It was ported to the PC and was released recently. Along with it comes much better graphics compared to the X-Box version. Halo is very demanding graphically and has the ability to bring any of today's hardware to a crawl should you wish to.
Halo for the PC includes four different rendering paths:
1.) Pixel Shader 2.0 (DX9) – At this code path everything is rendered as best as possible with as complex of a calculation that is necessary to generate the best visual result possible. However, there are some simple effects that automatically use a 1.4 or 1.1 shader because the visual result is exactly the same.
2.) Pixel Shader 1.4 (DX8.1) – This code path reduces some effects, specifically there are no bumped mirrored surfaces and some video effects are done in two passes.
3.) Pixel Shader 1.1 (DX8) – There are reduced visual effects such as No model self-illumination, No animated lightmaps, Fog calculations are triangle based not pixel based and there are no specular lights.
4.) Fixed Function (DX7) – Many compromises are made such as no shadows, simple active camouflage effect, no glows or flares and very basic fog, water and lighting.
The game automatically chooses which code path it needs to use based on what your video card supports. You can also force what code path you want the game to run in. In the shortcut for Halo you can add the command line –use20 for Pixel Shader 2.0, -use14 for Pixel Shader 1.4, use11 for Pixel Shader 1.1 and –useff for Fixed Function. In our testing we went ahead and made sure all the cards were using Pixel Shader 2.0 since they support it and specified the –use20 command.
We are also using the latest 1.02 patch for this game.
This game does have a built in timedemo mode. If you use the command line –timedemo in the shortcut to the game it will run through a sequence of benchmarks using some of the cut-scenes in the game. However, cut-scenes are scripted events that run through a pre-determined path. They do not represent actual gameplay because they do not allow full free 3D movement.
In order to represent actual gameplay we relied on FRAPS and a manual run-through. Using FRAPS we ran through the “Assault on the Control Room” mission. This level was the hardest in terms of performance on our cards. There are lots of mirrored bumped surfaces over this entire level that put the stress on these video cards.
As shown above, we set our visual qualities to what we thought most gamers would to enjoy the best visual quality while maintaining adequate performance. We had to turn off dynamic shadows as well as the flash-light visual feature. We benchmarked this level at 1024x768 with no anti-aliasing and no anisotropic filtering. We made one histogram showing performance through our manual run-through. Obviously you will see why we had to play with no AA and no AF and at such a "low" resolution on these cards.
We think that this information is much more important to the gamer than what we have been sharing in the past. If you spend $500 on a new video card, we feel as though you need to know if your actual gaming experience is going to suffer due to low frame rates or sub par Image Quality...or both. Still, if you spend $100 on a video card, we feel as though you should have the same information. Obviously you are not going to hold the two cards to the same standard, but certainly we all want to be doing as little "chugging" as possible and don't want to be looking at a rendered image that seems as if it was drawn with chalk and Crayons.
Bottom line it is all about playing games. So we will be finding just how high we can be turning up IQ and resolution, while still maintaining a playable frame rate. Our exact settings will not represent everyone's on the planet, but we do feel as though we will be able to drop a good beacon that will allow you to figure out if the card showcased is right for you and your gaming needs.
We have not gotten perfect at our new techniques yet so you will certainly see our graphic comparisons improve. Do rest assured that these frame rates you see in our graphs do represent real-world gameplay, and that is something that our video card reviews have been lacking. Also feel confident that our IQ comparisons and screen shots are realistic and represent actual gameplay. Relating what your gaming experience will be with the video card is now our priority. By the time of our next review, we should be fully geared up for the new format. My apologies for not being able to do that fully in this 9600XT review.

The Radeon 9600XT maintains a more consistent higher than 30FPS frame rate compared to the Radeon 9600Pro. However, there are still some instances where performance dips down to 20FPS. But that is no comparison to the horrible framerates achieved on the GFFX 5600Ultra. Performance in this level on the 5600Ultra was nothing short of a slide show. With performance differences like this jumping out at us it isn’t hard to make a call on which mainstream card is better for playing Halo. (And by the way, it is a great game to play, we highly suggest it if you have not experienced it on XBox already.)
Screenshots:
Above are three shots from Halo taken at 1024x768 with No AA or AF. The levels the shots were taken in are as follows: Shot 1 is from "Halo", Shot 2 is from "The Silent Cartographer", and Shot 3 is from "Assault on the Control Room".
Similarly we also took the same shots on the 5600 Ultra. While comparing these images plus in-game gameplay we noticed nothing major in difference comparing image quality. The only major difference between these cards in this game is pure performance.
