With the advent of the Voodoo5, anti-aliasing became an important new feature - drivers from both NVIDIA and ATI were quick to materialize with FSAA support, however shoddy. The method that was employed in these drivers was known as super-sampling. Known as a "brute force" technique, super-sampling renders each frame at an extremely high resolution before scaling and filtering it down to the chosen display resolution. This technique results in much slower performance since the many of the pixels rendered are never even seen.

Using a new technique dubbed HRAA, NVIDIA claims that they can achieve better performance than super-sampling with comparable quality. A "reconstruction filter" uses data from neighboring pixels to compute the final output of any given pixel. The sampling pattern used by the reconstruction filter is called Quincunx and, according to NVIDA, offers quality rivaling traditional four-sample super-sampling while incurring the performance hit of two-sample super-sampling.

The benefit of Quincunx comes from using more input data to construct the final pixel while minimizing the memory footprint that is left by rendering scenes at high resolutions. As you can see, Quincunx requires the same amount of frame buffer storage as two-sample super-sampling, while taking nearly as many input points as four-sample.

Although we don't have a GeForce3 in the lab, NVIDIA has projected performance numbers for the part. If these numbers prove to be inline with real-world performance, the main benefit of the GeForce3 for today's applications will be the frame rate gains with anti-aliasing enabled.

Although we did not receive any specifications regarding the RAMDAC speed of the GeForce3, we were told that the card would offer " the same great video and 2D functionality as the GeForce2 family." Generally we have found the 2D quality of the GeForce2 to be lacking in 1600x1200 and above, so we can only hope that the GeForce3 will not suffer the same fate.

On the video acceleration front, the GeForce3 will be offering the same motion compensation engine found on the GeForce2 GTS along with alpha sub-picture blending to ensure those menus and sub-titles are decoded in hardware. NVIDIA's 5-tap horizontal/3-tap vertical filter is in place for the purpose of scaling accuracy. As far as quality is concerned, ATI looks to maintain the crown in this department with their adaptive de-interlacing support.