Gaming Tests
The heart and soul of figuring out game play. These are the true tests of and video card setup.
Doom 3 (OpenGL) - This test was run with the version 1.1 patch.
I ran the time demos first to see where I could play the game at in terms of performance vs. visual quality. These were used to get a ball park area of where the cards perform. Check out the in game graphs to see the bottom line.
| NX6600GT Single |
No AA/16xAF
|
2xAA/16xAF
|
4x AA/16xAF
|
| 1024x768 |
77.9
|
60.5
|
63.1
|
| 1280x1024 |
56.3
|
40.8
|
N/A
|
| 1600x1200 |
41.4
|
27.1
|
N/A
|
| NX6600GT SLI |
No AA/16xAF
|
2xAA/16xAF
|
4x AA/16xAF
|
| 1024x768 |
95.7
|
87.8
|
67.2
|
| 1280x1024 |
84.4
|
69.0
|
43.8
|
| 1600x1200 |
68.4
|
48.7
|
28.7
|
Doom 3 Timedemo, no AA, 16x Aniso. - These speeds are scorching. In both single and dual configurations all of these resolutions are easily playable.
Doom 3 Timedemo, 2x AA, 16x Aniso. - It scales pretty well right here as well. The 1600x1200 score on the single setup is starting to dip into unplayable however.
Doom 3 Timedemo, 4x AA, 16x Aniso. - Even in SLI form the 6600GT's really start to choke out when running 4xfsaa.
After running through with the Timedemo, I played the game myself and used FRAPS to monitor the action. Using the information I gathered earlier it opened up some secrets to how the card(s) will perform while actually playing. I decided to choose the best looking settings possible at each individual level to show what the cards were comfortable playing at. I played the same level for each run-through and attempted to play it as close to the same as possible each time. Something to note is that I did not use a custom .cfg file. All play settings are at the maximum values as they are in the options, and no additional tweaking was performed.
| NX6600GT Single |
Min
|
Max
|
Ave.
|
| 1024x768 (4xAA/16xAF) |
24
|
63
|
47.3
|
| 1280x1024 (2xAA/16xAF) |
22
|
63
|
48.4
|
| 1600x1200 (NoAA/16xAF) |
21
|
62
|
47.4
|
With a single card Doom 3 performed well at the settings show in the graph. It was very rare for the card to dip below the 30fps mark and all times it did that was momentary during a scripted event. It's impressive to see that even at 1600x1200 a single 6600GT can perform well. Also it is noted that this game doesn't improve very much graphically from having AA on as in other titles but improves much more from running a higher resolution.
| NX6600GT SLI |
Min
|
Max
|
Ave.
|
| 1024x768 (4xAA/16xAF) |
34
|
63
|
58.2
|
| 1280x1024 (2xAA/16xAF) |
37
|
64
|
58.3
|
| 1600x1200 (NoAA/16xAF) |
32
|
62
|
55.3
|
You'll notice with SLI while running at the same resolutions and AA settings (for 1024 and the 1280 tests respectively, the 1600x1200 test was run at 2xAA) the framerate never even gets to 30fps, and runs near the 60fps frame cap for most of the duration of the test, even the budget 6600GT when put into pairs can be an impressive powerhouse. Look at the minimum and average FPS in the table and the improvement is very obvious.
FarCry (DirectX 9.0) - This was run with the version 1.1 patch.
| NX6600GT Single |
No AA/16xAF
|
2xAA/16xAF
|
| 1024x768 |
51.18
|
44.8
|
| 1280x1024 |
36.56
|
31.27
|
| 1600x1200 |
26.75
|
22.89
|
| NX6600GT SLI |
No AA/16xAF
|
2xAA/16xAF
|
| 1024x768 |
59.86
|
58.06
|
| 1280x1024 |
52.47
|
43.74
|
| 1600x1200 |
39.53
|
31.95
|
FarCry Timedemo, No AA, 16x Aniso. - In FarCry this setup peters out really quick especially when moving to higher resolutions. It just goes to show that still NVIDIA doesn't have a good handle on Direct X based games compared to ATI.
FarCry Timedemo, 2x AA, 16x Aniso. - Something interesting to note is that the performance hit from running fsaa isn't as bad as it could be in other games. However the performance of the single card at these settings is virtually unplayable.
Then I moved on to actual play. I played the level "Fort" and recorded game play for 120 seconds trying to make sure each run-though was as close to one another as possible. For this I used God mode with no other special settings.
| NX6600GT Single |
Min
|
Max
|
Ave.
|
| 1024x768 (NoAA/16xAF) |
7
|
55
|
34.4
|
You'll notice that I only got through playing FarCry at 1024x768 with no fsaa on. This is because I found anything above this resolution or setting unplayable. I experienced a lot of mouse lag and even with God mode on could waste many magazines of ammo trying to kill single enemies. 1024x768 ran well enough though. You'll notice that the Minimum FPS shown here is "7". I would ignore that as that was only at the very beginning of the test while the level was fading in. Something to note however is the average which is above 30, and for the most part the graph which shows that through most of the test it was above 25. For some these play settings might also be considered "unplayable" settings but all things considered the mouse moved smoothly and aiming and firing even while moving was still easy to do accurately, even with the sniper rifle.
| NX6600GT SLI |
Min
|
Max
|
Ave.
|
| 1024x768 (NoAA/16xAF) |
6
|
72
|
37.8
|
| 1280x1024 (NoAA/16xAF) |
0
|
57
|
25.9
|
Once again don't pay attention to the minimum fps because that was taken from the beginning of the test while the game was fading in. You'll notice that the FPS on the 1024x768 is still low even though improved and the 1280x1024 test is absolutely terrible. I struggled through that test at almost an unplayable framerate. I think it's sad to think that even in a SLI configuration that this pair of cards performs so poorly on this test. I've noticed that around the web that many people compare 2 6600 GT's as being at around the same speed as a single 6800GT in this case, that might hold true.
NEXT
