Half-Life 2 (Direct X 9.0)
For this test I first ran the timedemo again to get a general idea of where the cards stood at in terms of performance, and then for the actual game play testing I used dl_canals_12 which features lots of explosions, water, and the HL2 engine.
NX6600GT Single |
No AA/16xAF
|
2xAA/16xAF
|
4x AA/16xAF
|
1024x768 |
40.64
|
39.62
|
36.33
|
1280x1024 |
30.66
|
26.63
|
21.59
|
1600x1200 |
25.89
|
20.14
|
14.12
|
NX6600GT SLI |
No AA/16xAF
|
2xAA/16xAF
|
4x AA/16xAF
|
1024x768 |
42.24
|
40.94
|
33.09
|
1280x1024 |
35.66
|
19.83
|
21.54
|
1600x1200 |
27.37
|
20.45
|
16.66
|
Half-Life 2 Timedemo, NoAA, 16xAnsio. - There isn't much of a difference in terms of performance here. The SLI setup only improved the framerate by about 2 frames in each of the individual resolutions.
Half-Life 2 Timedemo, 2xAA, 16xAnsio. - The same pattern continues here. Something else to note is the faster gameplay at 1600x1200 than 1280x1024.
Half-Life 2 Timedemo, 4xAA, 16xAnsio. - Almost no framerate difference here between SLI and single configurations. I personally found this puzzling.
I didn't take the timedemos to too much heart however, as actual game play is quite different than a timedemo for many obvious reasons. I found that actual gameplay tells quite a different story in terms of performance.

NX6600GT Single |
Min
|
Max
|
Ave.
|
1024x768 (2xAA/16xAF) |
39
|
178
|
83
|
1280x1024 (NoAA/16xAF) |
29
|
159
|
69.3
|
1600x1200 (NoAA/16xAF) |
30
|
109
|
54.8
|
As this test shows very clearly, even in single form this card can perform very well at Half-Life 2 even on this very graphically intensive map. With a minimum in all 3 resolutions at 29 (it's basically 30&) or above.
With SLI things get even better.

NX6600GT SLI |
Min
|
Max
|
Ave.
|
1024x768 (4xAA/16xAF) |
25
|
168
|
74.2
|
1280x1024 (2xAA/16xAF) |
16
|
136
|
54
|
1600x1200 (NoAA/16xAF) |
21
|
112
|
57.7
|
This is a much larger improvement than what the timedemo showed. I was able to play the game with higher AA settings throughout. The difference in the 1600x1200 scores is mostly due to the run-through being different, but the average is still higher.
Unreal Tournament 2004 (Direct X 9.0)
This game of course is more or less a rehash of all the Unreal Tournaments that came before it. I used 3 separate timedemos to benchmark the cards on this particular game. All of the settings were set at their absolute maximum.
NX6600GT Single |
dm-rankin
|
as-convoy
|
br-colossus
|
1024x768 (4xAA/16xAF) |
123.36
|
81.44
|
107.43
|
1280x1024 (2xAA/16xAF) |
74.06
|
56.54
|
39.17
|
1600x1200 (NoAA/16xAF) |
107.43
|
76.6
|
48
|
NX6600GT SLI |
dm-rankin
|
as-convoy
|
br-colossus
|
1024x768 (4xAA/16xAF) |
209.03
|
134.42
|
98.53
|
1280x1024 (2xAA/16xAF) |
89.71
|
90.21
|
74.63
|
1600x1200 (NoAA/16xAF) |
170.51
|
141.17
|
91.27
|
Unreal Tournament 2004 Timedemo, 4xAA, 16x Aniso, No SLI - Not much to say here. Even when trying to press the card in single form to the limit, it's still able to hit averages above 30fps, or well above.
Unreal Tournament 2004 Timedemo, 4xAA, 16x Aniso, SLI - As if the game play needed to get faster, with two cards the lowest average score raises up to 74.63! I would have spend time to record real gameplay with FRAPS, but honestly this game just doesn't push these cards like Doom 3, FarCry, and Half-Life 2.
Thoughts about SLI
This is personally my first experience using SLI, and in relative terms it is a new technology, and as such there are many kinks still getting fixed up. First off I was lucky to be able to use the newest WHQL drivers that NVIDIA has recently released (71.84) which cure a lot of problems starting with what type of SLI (split fame, or alternate frame rendering) and issues with profiles etc. If you haven't read up on these things, I suggest you go read up about it on NVIDIA's webpage or on local forums or webpage's such as Viperlair's own.
I personally am a dual monitor user, and one of the greatest downfalls I had about SLI was the fact that when using both cards in SLI mode, it only allows one output. It seems sort of silly to me that with four outputs total across two cards that they couldn't have figured out a way to use some of that power to have it render a second monitor for even just 2D apps like web browsing. I'm not even asking for multi-monitor 3D applications, it simply leaves me most of the time turning off SLI and using two monitors instead of using SLI on only one (changing this mode in the drivers requires a restart which can be a hassle). Only the Quaddro line of cards supports this feature and I don't know if NVIDIA will support dual monitors for SLI home users anytime soon.
The other issue for most is cost benefit. Right now I think the most prudent way to buy honestly is the 6800GT, it probably offers the best performance to buck ratio if you prefer the NVIDIA flavor right now. The problems I see with the idea of buying two 6600GT's and using them in SLI is that with some applications a single 6800GT trumps it. Additionally the overall cost is the same. Some people have the idea of buying one card now, and then buying another later when they have more money, but the problem with that is by the time the money is acquired why bother with SLI at all? More than likely another card will have come down the pike offering better performance or similar performance than the pair of cards. It seems to me that the only real solution is buying a pair strait out, or sticking to a single card solution. Let's be honest SLI isn't for those on a tight budget whether or not it's 6600GT's or 6800Ultras. That said if you do decide to go the dual 6600GT route there is still a lot to be gained. Hopefully with more updates from NVIDIA more benefits from having SLI will form, and truly put the two GPU's up to their maximum potential. As long as PCI-E is the choice for graphic cards solutions exist, SLI will be here to stay. Logically 2 processors even at a slightly slower speed should do better than a single faster chip, and hopefully then there will be much more reason to go with a dual "budget" option then a single faster one, but that isn't the case now.
Conclusion
MSI has put out a very good product with their NX6600GT. It looks good with the red PCB and the MSI standard gold heatsink. The cards performed flawlessly through a barrage of tests with absolutely no hiccups, and I've been using them for tons of games for hours in my main machine for at least a good month. The performance of the pair on some tests is above that of the 6800GT on some tests and below the 6800GT on others according to other reviews I've seen around the web, so I leave it to you to decide if the games you play will benefit more from having a pair of cards or a single faster one. It's taken some time for NVIDIA to start maturing their drivers between actual WHQL certified drivers, and lets hope they step up to the plate like ATI has been doing with early and often driver releases in order to allow end users to get the full experience of having two cards, especially after shelling out so much hard earned money.

Pros
Nice red PCB, and Gold heatsink
Good performance
SLI is almost 2x as good.
Excellent midrange solution
Cons
SLI is still expensive (even with 6600GTs)
No SLI dual monitor support (not the fault of these cards however)
NVIDIA slow to release official WHQL drivers.
Bottom Line: The NX6600GT is a great upgrade from anyone still on a FX5900 or below or a 9800Pro or below. The difference in the speed of hardware took an enormous jump, and I would highly recommend it as a medium range card for anyone needing one. The 6600GT is the best midrange card on the market right now, and this MSI offering is among the best 6600GT offerings.
If you have any comments, be sure to hit us up in our forums.