Transparency & Adaptive AA Explained

Removing the branding...

Transparency Multisampling:
NVIDIA's transparency anti-aliasing technique is a great feature for mid-range video cards and is also a great feature for high end cards in more intensive games like NFS: Most Wanted, where enabling transparency supersampling has massive performance implications that usually mean lowering the resolution in order to achieve smooth gameplay. However, it doesn't offer the same level of quality that ATI's performance adaptive anti-aliasing algorithm can deliver.

There is little reason not to enable this feature, as it does have a positive effect on image quality with a minimal performance hit on the frame rate. It helps to make alpha tested textures within close proximity of the player much smoother and easier on the eye. The quality of distant textures doesn't seem to be improved quite so dramatically - the minimal performance hit explains that, though.

Performance Adaptive AA:
We believe that ATI is still using a supersampled technique for its performance adaptive anti-aliasing technique because there is evidence that supersampling is being applied to the scene. In order to improve performance when performance adaptive AA is enabled, we believe that ATI is using an adaptive anti-aliasing divider to reduce the number of samples taken by the algorithm.

This is actually a pretty intelligent way to do things, because it means that the quality of the output is higher than what can be achieved with a multisampling algorithm. The downside is that it costs more performance than NVIDIA's transparency multisampled anti-aliasing algorithm, which can almost be enabled for free without any hugely noticeable performance deficit.

We need to do some more analysis on how this mode actually works at differing levels of anti-aliasing, but we think that when performance adaptive anti-aliasing is enabled, along with 4xAA, you're only actually getting the equivalent of 2x adaptive anti-aliasing. This means that when you're using 2xAA and performance adaptive anti-aliasing, you're unlikely to see any improvements in image quality over a conventional 2xAA setting.

Quality Adaptive AA and Transparency Supersampling:
The image quality delivered by both Quality adaptive anti-aliasing and transparency supersampling is very similar. We have been really impressed with the results delivered by this image quality enhancer, but you're going to require a high end card to utilise it at high resolutions - the improved image quality comes at a premium.

Essentially, both ATI and NVIDIA are doing the same thing and, not surprisingly, the final result is virtually the same too. In a blind taste test, you're not going to be able to tell the difference between the two supersampled alpha tested texture anti-aliasing techniques. That's good to hear, because we feel fairly confident in saying that if you've got a choice between two video cards and they're both playable with supersampled transparency/adaptive anti-aliasing, you couldn't have a better choice as either card is more than likely to satisfy your needs.

Final Thoughts...

Ultimately, these anti-aliasing enhancements are good for the consumer. We feel that this image quality enhancement is a great addition to both ATI and NVIDIA's never-ending battle to win the hearts and minds of hardware and gaming enthusiasts. This is because it genuinely improves the image quality on current generation hardware from the two graphics giants.

There are advantages for all of the different transparent/adaptive anti-aliasing modes. Transparency multisampling works well in situations where maintaining frame rate is important, while also having a positive impact on image quality, while Performance Adaptive AA provides a higher quality output. Obviously, the higher quality output comes with a larger performance deficit than transparency multisampling.

Both transparency supersampling and quality adaptive anti-aliasing aid in providing the highest anti-aliasing quality available on current hardware. The performance deficit can be quite large, but that is due to the nature of supersampling. That's probably the only major concern, because the image quality output is quite superb.