Hercules Prophet 4500 in the test: The Kyro2 compared to the GeForce

Hercules Prophet 4500 in the test: The Kyro2 compared to the GeForce


After talking about the technology of the chip so often in the course of this article, I would like to turn to di I will speak here shortly afterwards. Like the Kyro2, the Kyro is a so-called tiler or display list renderer. As already briefly mentioned in the Features section, classic graphics hardware tries to process every polygon as quickly as possible, with the intermediate results being left in the graphics memory until they are finally completed. The order in which the polygons come is how they are rendered. It does not matter whether one polygon overlaps the other in the final image. The Kyro2 works completely differently at this point. Before the actual rendering of the image begins, the entire screen is divided into small tiles, which are then processed step by step. In this way, the chip can determine from the start which triangle covers another, e.g. because there is a wall in front of it. At this pixel by pixel level the chip carries out all important operations and stores the information forthis tile or tile in the memory.

That should mean that the Kyro2 first checks which triangles for are visible to the user. Then only the visible area of ​​the triangles is covered with textures, with any fading effects being taken into account at this point. In this way, the total number of memory accesses can be reduced immensely so that the memory bus is no longer the bottleneck all too often. These completely independent bricks are then combined into one picture. Due to the fact that these bricks are completely independent, it is theoretically also possible for these individual bricks to be created by different Kyro Chips. Multi-chip is the keyword here. Even if only single-chip solutions are available so far, the technology is able to work with several chips, in such a way that the screen is divided in two in the middle and the left half from the first chip and the right half from the second Chip is calculated.

On the next page: Benchmarks