LOSTCIRCUITS

Performance

For what we want to show it is necessary to look at the bottlenecks first to see where CPU, GPU or memory limitations are putting a cap on the frame rates, if such limitations are becoming dominant, we can change the AGP interface as much as we like and we still won't see any differences unless we just choke the entire system with AGP 2X or 1X.

There is no difference between AGP 4X and AGP 8X when using the Granite Bay chipset in ViewPerf. Some of the internal benchmarks show increase in frame rates when we overclocked the system to 145 MHz FSB, however, Ugs-01, ProE-01 and 3DS Max results did not change which is consistent with the GPU being the limiting factor here. With tape-forcing AGP2.0, the results did not change either with the one exception of DRV-08 dropping to 90.22 fps. Selecting AGP 1X instead of 4X did not cause any further performance drop either which leads us to believe that the BIOS switch simply does not work.

ASUS P4G8X

As workstation chipset-based mainboard with huge memory bandwidth, the ASUS P4G8X could theoretically be the one to show off the differences between the AGP 4X (2.0) and 8X (3.0) protocols, however we failed to see anything significant. On second view, there is at least a possible explanation.

Next Page:    => A7N8X =>