NVIDIA GeForce4 - NV17 and NV25 Come to Lifeby Anand Lal Shimpi on February 6, 2002 8:51 AM EST
- Posted in
We tested using a AMD Athlon XP 2000+ on an ASUS A7V266-E under Windows XP. All NVIDIA cards used the latest 27.30 drivers with GeForce4 support while all ATI cards used the latest 7.66 beta drivers released on ATI's website.
Due to time constraints we were not able to include the GeForce4 Ti 4200, Kyro II and Radeon 7500 although we will be doing future comparisons with those cards included.
Memory Controller & Occlusion Culling Performance - VillageMark
VillageMark is a benchmark that was developed by PowerVR to show off the benefits of their tile rendering architecture which you may be most familiar with from the Kyro II card. Based on a tile or deferred rendering architecture, the Kyro II's core had the unique capability of not rendering any pixels that wouldn't be seen by the user. The VillageMark test has an incredible amount of overdraw (where the engine instructs the graphics unit to draw a large number of pixels that will never be seen) in order to stress the importance of features such as z-occlusion culling and other memory bandwidth optimization techniques.
While none of the architectures we're comparing today can be classified as deferred renderers, they do have some very advanced techniques borrowed from the deferred rendering world and thus VillageMark becomes a great benchmark of them. This way we can find out exactly how much of an improvement NVIDIA's LMA II is over the original Lightning Memory Architecture.
The first thing that should be mentioned is that the latest drivers from ATI resulted in a bit of a performance drop here. Normally the Radeon 8500 comes out with a 113 fps score, but with the latest drivers the score is pegged at 92 fps. Using the latest drivers fixed some significant bugs in one of the games we used to benchmark and thus we stuck with the latest drivers for this review.
Comparing clock for clock the GeForce4 to the GeForce3 Ti 500, both running at 240/500, we can see that there is definitely a significant improvement in the GeForce4's Z-occlusion culling algorithms. While the improvement isn't the 50% that NVIDIA is boasting, the 25% improvement that can be seen here is impressive none the less. The increased clock speed of the shipping GeForce4 boards make the improvement all that more noticeable.
The GeForce4's LMA II is clearly more than just a new marketing term.