Benchmarks II
|
Now on to two more synthetic benchmarks that are usually for predicting possible performance. Firstly, the lovely VulpineGL, an OpenGL benchmark which renders two complex scenes. Higher resolutions rely on bandwidth.
Again, nothing has changed from the card's point of view. We see how the ABIT Siluro Ti 4200 just loves a fast host processor, especially at 1024x768x32. It is 15fps faster here than when used at 1600MHz. Cards that scale well usually have a greater longevity than those that can't. The Siluro Ti 4200 is fast enough that it simply waits for the CPU to feed it at lower resolutions. How about the impressive-looking Comanche 4 ?. This benchmark is characterised by the massive amount of geometric data required as simulations often cover vast expanses of land.
Talk about the Siluro waiting for the processor to feed it, here is a case in point. You'd think that we were benchmarking two different cards, such is the disparity between results. All we've done is to increase the core CPU speed and let the ABIT Ti 4200 do the rest. The very fact that at 1600MHz, the 1024x768x32 and 1280x1024x32 scores were similar, lead us to believe it was CPU-limited. This proves the case as we go to 2533MHz. This reiterates the belief that if you have a low-end CPU, and play at low-end resolutions (1024x768x32), it doesn't make an enormous amount of sense to use say a Ti 4600, it will simply wait for the CPU to feed it in most cases. The only caveat is if you use anti-aliasing or anisotropic filtering regularly. Now on to two real games, I'll conduct overclocking analysis on these to see how card 'clocking affects overall performance. |
