Anti-aliasing Evaluation, and Overclocking
This content was originally featured on Amdmb.com and has been converted to PC Perspective’s website. Some color changes and flaws may appear.
In this section, we examine some anti-aliasing features exclusive to NVIDIA, specifically Quincunx anti-aliasing, 4XS and 6XS anti-aliasing. We will use Unreal Tournament 2003 to benchmark these features and compare image quality.
2xAA vs. Quincunx Anti-aliasing
Click thumbnails for larger image.
Comparing Quincunx and 2x anti-aliasing, we can see that there is no perceptual difference between the two. In fact by quickly alternating between the two visual samples, I could not even distinguish a difference. Comparing the frame rates for these two anti-aliasing methods, you can see that there is no performance difference either. So for all intents and purposes, Quincunx anti-aliasing is the same as 2x anti-aliasing.
4xAA vs. 4XS vs. 6XS vs. 8xAA
Click thumbnails for larger image.
Here is where we begin to see a major difference. By examining the 4xAA and 4XS samples, we can see that the 4XS anti-aliasing method is clearly superior in terms if visual quality. There are less noticeable jagged edges with 4XS, but at what cost? Comparing the performance of the two methods, there is a huge difference. We see that 4xAA is by far faster than 4XS, even with anisotropic filtering enabled.
Click thumbnails for larger image.
Comparing the images in 6XS to 4XS, we see something surprising – it appears that 6XS anti-aliasing is actually worse than 4XS! There are more noticeable jagged edges in 6XS than in 4XS. Alao, if you compare the 4xAA sample to the 6XS sample, they look nearly identical. As far as I can tell, 6XS anti-aliasing is similar in quality to 4xAA but with a greater performance hit. However to the credit of 6XS, foreground texture quality appears to be better than in 4XS or 4xAA.
Click thumbnails for larger image.
Just for curiosity’s sake, let’s compare 8xAA to 6XS. Comparing the image samples, I cannot see a real difference in anti-aliasing. In fact, there are still slight jagged edges at 8xAA which is the highest anti-aliasing level you can achieve on this video card. Foreground textures look sharpest at 8xAA, but background textures look the same as in 2xAA.
3DMark 2001SE, 3DMark 2003, and Overclocking
Here are the only theoretical benchmarks in the entire review of the MSI FX5600-VTDR128. 3DMark has become a good way of seeing the relative performance of a particular card when compared to another. We will use 3DMark 2001SE (v320) and 3DMark 2003 (v3.1.3) to demonstrate the effects of overclocking. We will then use Unreal Tournament 2003 to see if any of the overclocking makes a real difference in performance.
At 341/652 overclock, I began to see noticeable tearing and artifacting on the screen. Unlike the results seen in the Ti4800SE, the tearing and artifacting did not decrease the scores in any way, which I find very curious. You would figure that when a scene gets improperly rendered frame rates and 3DMark scores would go down, but this is not the case. The maximum overclock I was able to achieve without artifacts was 331/652 which is a 22% / 18% (core / mem) MHz increase.
We check our overclocking results with Unreal Tournament 2003 to see the real-world effect of overclocking. We can see that overclocking has minimal effect with only an increase of about 5 extra frames per second.
