Inno3D Tornado GeForce3 Ti200 in the test: What does the little brother of the GeForce 3 do?
This benchmark was originally developed by PowerVR and served to illustrate the advantages of the Kyro2, as the overdraw in the VillageMark is particularly large. This means that most graphics cards also calculate the objects that are hidden by others and therefore actually do not have to be calculated. Thanks to its tile base rendering technology, the Kyro2 can recognize which polygons can be seen in the final image and which cannot, so that it does not calculate an unnecessary number of polygons.
You can be a bit surprised here. Why the GeForce3 Ti200 is significantly slower in 16-bit than in 32-bit is completely incomprehensible at first glance. But we got to the bottom of the matter and came to the following result: If you activate the 'LightspeedMemoryArchitecture' (LMA) feature in both 16 and 32 bits, you get the result shown above.However, if the LMA is displayed in both color depths, the results are confirmed in 16-bit, but the frame rates for 32-bit drop rapidly and are then worse than in 16-bit.
Now there are two options:
- Nvidia would say that this is an intended feature and that it should significantly improve performance in 32-bit color depth. Unfortunately, the new LMA only increases the speed in 32Bit.
- There is a bug in the drivers and the LMA simply does not work in 16Bit, but only in 32Bit so far, or LMA works in 16Bit generally not.
On the next page: 3DMark 2001