Test setup:

Graphics card: Club3D Radeon X1900 XTX
CPU: Athlon 64 3200+ at 2.7GHz
Motherboard: DFI LanParty CFX3200-DR
Chipset: ATI Radeon Xpress 3200
RAM: 2x1GB Geil DDR 3200 (6-2-2-2.5)
Hard Disk: 320GB Western Digital SATA 8MB Cache
Operating System: Windows XP SP2
Other software: .NET Framework 2.0
GPU drivers: Catalyst 06.5

Here is the software we used to benchmark this card (all games/benchmark software updated to their latest versions):

3D Mark 2005
Quake 4
F.E.A.R
NBA Live 2006
Half Life 2 – Lost coast
Need for Speed – Most wanted

A clean system was created for this test. First the operating system was installed, fully patched, installed .NET Framework 2.0 and then the appropriate drivers for the chipset and the graphics card. For the games we used the appropriate in-game settings for Anti-Aliasing and anisotropic filtering, where they exist – otherwise we forced the settings from the Catalyst Control Centre. All games were benchmarked using FRAPS and in-built timedemo functions, unless otherwise noted. To make sure we got it right, each benchmarking session was run three times, to make sure we get consistent results. Onboard sound was enabled for all tests, to better simulate the kind of performance an ordinary user would experience.

For this review we decided to try something different as far as our settings: we didn't bother at all with No AA / No AF situations. These cards are true monsters, and we bet you don't go out paying more than 350 pounds just to disable Anti-aliasing and anisotropic filtering. So, all of our measurements have some degree of AA and AF enabled. We should also note that, using the Catalyst Control Centre, we turned the graphics quality all the way up, meaning that at all times Adaptive Anti-Aliasing was set to Quality and High Quality Anisotropic Filtering was enabled. As far as resolutions are concerned, we chose to benchmark in three resolutions: 1024x768, 1280x960 and 1600x1200, simply because the games we used limited us to 1600x1200 as the highest resolution we could use. For Lost Coast, we also benchmarked at 1920x1200 since the option was available. In-game, all image quality options were set to their highest values.

Overclocking

The CPU we had in our disposal was the Athlon 64 3200+ overclocked to 2.7 GHz (from 2.0 GHz originally). That allowed us to have a very good performance gain and alleviate much of the CPU bottleneck we would have at stock speeds. Of course having one of those AthlonFX-60s would make the problem go away even more, but we felt that we needed a system that was fast, but not top of the line, in order to get a better real-world performance picture from the card.

In order to overclock the X1900XTX we used the overclocking tool hosted at DriverHeaven, and which you can find here. Since the new cards allow us to alter not only the clocks, but the board voltages as well, that gave us the opportunity to play some more with the values in order to get the most stable clocks for our particular board. Here were our results:

As you can see, the card we had in our disposal had very little headroom for overclocking, giving us a rather modest boost in clock rates, especially for the memory. Whether this boost translates in an equally modest performance increase is something we will see in the following paragraphs; since what we are about to do is to find out the performance the card gives us in two scenarios: at stock clocks and at overclocked speeds.