General Performance Testing Comments
Ten parts (three lathe/MTM, five fixed milling, and two rotary milling) were used as a comprehensive FlashCPR performance test suite. All tests were performed with the latest available drivers from the video chipset manufacturer and at factory clock rates. All tests were performed twice to verify low testing variance.
Lower scores (shorter bars) are better. The row titled GDI Generic is a baseline score obtained by disabling hardware acceleration. Row titles are given by the renderer name reported by the driver; in all but one case, this is the exact model of the card. The Radeon X1800 Series tested is actually a Radeon X1800 XT. Please note that AGP and PCI-E results are not directly comparable to each other; the baseline for the AGP testbed system is somewhat slower. See the graphs for details.
AGP Performance Testing Results
Test bed used for AGP performance testing: AMD Athlon 64 3000+ on a Gigabyte GA-K8NS-939 (nForce3 chipset), with 1GB of dual-channel DDR-400.
PCI Express Performance Testing Results
If you are using a relatively modern video card (GeForce2/Radeon 8500 or newer) with a reasonably large amount of memory (128Mb+), turning Display Lists on can improve performance significantly. Note, however, that if you are using a card without hardware geometry acceleration, or one with less than 128Mb of memory, turning Display Lists on can actually result in a significant slowdown.
Partial Updates can increase performance on certain configurations, and has no effect in others; several of the quality problems mentioned in the first part of this document are caused by the Partial Updates setting, however.
Vertex Arrays, like Display Lists, can be used improve performance on modern cards with large amounts of memory. In some limited-memory situations (particularly those in which the video memory is shared with the system memory), Vertex Arrays can cause a slowdown; in general, however, it is recommended to leave Vertex Arrays turned on.