Here are the results for the other demos :
| Demo file | FPS at 1024x768 | FPS at 1280x1024 | FPS at 1600x1200 |
|---|---|---|---|
| Demo001 | 153.7 | 121.9 | 90.1 |
| DEMO002 | 157.8 | 131 | 98.6 |
| CRUSHER | 85.1 | 68.6 | 50.7 |
| TORTURE1 | 80.5 | 57 | 40.5 |
| AVERAGE | 147.9 | 124.1 | 93.1 |
Please note that the average scores noted here take the first list of results into account, in order to make a clearer picture of performance.
Now, what is interesting is that I redid all tests at 1024, but with AA mode Quincunx activated. Why 1024 ? Well, I don't know about you, but I read an interesting article on CRT monitors and performance (on Tom's Hardware of course). Basically, it says that it's not because a monitor can do 1600x1200 that makes a good reason for the monitor to do it. Besides, if you run anything at 1600, you don't need AA anyway (in my opinion).
So, after running the demos with Quincunx, I charted the results from best to lowest score. This is the result (attention : the bottom level is 40fps, not 0fps) :
As you can see, the results show that Quincunx gives a serious performance hit compared to the highest scores, but it also maintains the same average level of performance longer. The sustained level of 80-90fps only fails at the very hardest demos (Crusher and Torture1), then falling below the 60fps mark. For me, that seems to be a good sign.