Time to begin the testing phase. Our test bed is an Intel P4 2.4C, 1GB OCZ PC3200 on an ABIT IC7 motherboard. The ATI Radeon 9600 XT used Catalyst 3.8 Drivers and the Chaintech FX-71 used the Nvidia Detonator 53.03's.

Test Software will be:

Unreal Tournament 2003
Halo
Splinter Cell
Max Payne 2
Call Of Duty

The comparison video card will be the Chaintech 5600 Ultra I reviewed before which is aimed at the same segment of the market, although it has been superceded by the newer 5700 Ultra chipset. We'll also be covering image quality and overclocking. was used to measure in-game performance where applicable.

For those of you who don't know, there are a lot of problems with high levels of AA and some Direct 3D games, which happens to be the majority of those tested today. We'll present some games where we didn't experience abnormalities, but in many cases, 6xAA was faster than 4xAA, which cannot be, but was the case during benchmarks.

Unreal Tournament 2003

For UT 2003 we use Test on maps Antilus and Inferno. This will automatically configure and run the system through its test, hopefully leaving Human Error behind.

Antalus - 1024x768, Maximum Quality

Antalus - 1280x1048, Maximum Quality

The only thing that can be said here is that the 9600 XT runs UT2003 flawlessly at High Quality 1024x768 or 1280x1024 screen resolution. The demo, as well as playing the game, is smooth with no noticeable hiccups or stutters. When comparing the results of the benchmarks we can see that on Antalus, there is a significant difference between the 9600 XT and the 5600 Ultra, it is even more evident on Inferno.

Inferno - 1024x768, Maximum Quality

Antalus - 1280x1024, Maximum Quality

Turning on AA and AF drives home that point, a 10% difference in performance on Antalus without AA or AF turns into a near 25% difference with 2xAA and 8xAF. Anything beyond 2xAA / 8xAF becomes a little to much for the cards to handle, although, the 9600 XT makes a good run at it. As I have stated before, if you are going to tweak your image so you can run at 6xAA / 8xAF, you are tweaking away what you are trying to gain, best to leave it off or at 2xAA / 8xAF, the image is quite stunning at that level.

Halo

Halo is a fairly new with some very intense graphics, this should be a good test of what a current card needs to be at today as well as a look into what the card may hold for future games. Halo has yet to fix it's AA so we will not be testing that phase of it. It is unfortunate that the only test we have is the imbedded timedemo which has no actual gameplay to show.

Halo @ 1024x768 w/Timedemo

From this benchmark you can see that the cards are fairly evenly matched, lets take into real game play and see how it goes. For this we are going to use FRAPS and run / look around the initial control room once while looking up and down, not an exact science, but should show us a better level of comparison between the cards.

Halo @ 1024x768 w/Fraps

Both cards are run @ 1024x768, you can see they are evenly matched here with a slight edge going to the 9600 XT, not enough to warrant a clear winner stamp however.

Halo @ 1024x768, % time

Looking back at the Halo timedemo, we also gathered some information on the percentage of time it spent at various framerates

What we learn from this graph is the both cards spend most of their time below 60 frames per second, but more specifically, more than 95% of the time was spent under 50fps. Overall, the majority of the time was spent between 20 to 50fps, at 1024x768, which works out right based on our average timedemo benchmark score.

NEXT