by Digby » Feb 25, 2003 @ 8:46am
That particular benchmark is near worthless for trying to judge the performance of games.
It would at least need to test the speed to which the CPU could copy DWORDs from system memory to the buffer returned by GAPI's GXBeginDraw().
When you look at the BitBlt test result and it is several times faster than the numbers reported by other similar devices, that's a clue that it has some sort of hardware assist in copying pixels. However, if the BitBlt is much faster AND if the StretchBlt performance is slower than most other devices, that means that the CPU is having problems writing to display memory (slower RAM, narrow bus, etc.). This is a hint that games will probably run slower on this device.