Halo was one of the most anticipated and (to a lot of people) disappointing game releases in the last year. As much as Bungie and Gearbox tried, there wasn't too much that could save Halo from suffering from console port syndrome. All this, however, does not preclude me from using it to benchmark this 6800 GT. The game does take advantage of Pixel Shader 2.0, and it requires top-end hardware for maximum enjoyment at high resolutions.
For testing I used the game's built-in timedemo and the latest 1.04 patch. Sound was disabled and I forced the game to use Pixel Shader 2.0 via the command line. As Halo does not support FSAA, I ran the eye candy tests with 8x anisotropic filtering only.
Though I personally enjoy Halo much more on the Xbox (co-op mode rocks!) there's no denying that you get some pretty amazing visuals when you pair this game up with the 6800 GT. Performance also is very good. The pacing of Halo is much different than something like Unreal Tournament 2004. Here, a score in the 45-50 fps range is more than acceptable.
Here's a title that kind of fell under the gaming radar. Painkiller is much like Serious Sam in that it is a fast-paced shooter with some cool weapons and lots of enemies. Though there appears to be some thinly veiled plot, Painkiller is all about the killing. While the graphics won't blow away certain other A-list titles that are scheduled for release in the next couple months, the physics engine is what causes Painkiller to really stand out in the crowd. There's nothing like blowing away a baddie and have his body get tossed about the place.
I used Fraps to record my play for a section of the Opera House level. There is an in-game control for FSAA, but anisotropic filtering was set through the driver control panel. 1280x1024 resolution was unavailable for me (even though I could use it on my ATI cards) so I had to use 1280x960. All graphic options were set to their highest levels, including character shadows. Sound also was disabled, and I tested using version 1.2. A 1.3 patch was released, but I did not have time to test it.
CPU limitation rears its ugly head again here. And the 4x AA/8x AF numbers you read are correct over the average of five tests. Don't put too much stock in the difference, though. I attribute this oddity to some inconsistencies during the test since I didn't play the level exacly the same way every time. I'm very surprised the performance hit when using AA/AF wasn't greater at 1600x1200. Dreamcatcher, the game's developer, deserves some props here for some very tight coding. On an additional note, Painkiller is one of the games on NVIDIA's list that will support the GeForce 6800 series Shader Model 3.0. There hasn't been any word on exactly when we'll see it; I can only assume it will be once the 6800 series is more widely available.
UNREAL TOURNAMENT 2004
You can't call yourself an online FPS player if this game isn't in your arsenal. This LAN party favorite is challenging for players of all skill levels as well as computer systems of all types. There really isn't too much more to say about this game, it's Unreal Tournament! I remember playing the very first iteration on my PII 400/Voodoo3 PCI system. My PCs and this franchise have come a long, long way since then.
Unreal Tournament 2004
I wanted to find a timedemo that would really stress my system. The standard flyby and botmatch tests wouldn't do at all. Though I don't remember where I downloaded it, the demo I found was called UTBench. It shows an intense onslaught match. Vehicles, explosions, death - it's all here and ready to put the screws to your hardware. The game was patched to the latest release, and all in-game options were set to their maximums with sound disabled.
Unreal Tournament 2004 Results
It might have been a peculiarity of this particular timedemo, but these results aren't exactly what you'd call exciting or encouraging. UT 2004 loves processor power, and I apparently didn't have enough to get pleasing results.
Please bear in mind that most UT 2004 maps/games will probably not play out like this one. This is about as "worst case scenario" as you're possibly going to find. A lot of maps I personally ran through showed much better performance. If you happen to find yourself with a GeForce 6800 and want to play UT 2004, you'd do best to experiment with driver and in-game settings to tailor your experience to your liking.
nV News also has some additional information about the GeForce 6800 and Unreal Tournament 2004. You can read MikeC's analysis using this exact same demo here.
My results for this demo have inspired me to work on a CPU scaling analysis article. I have a 3.4 GHz Pentium 4 Extreme Edition CPU en route to me this weekend, and I'm going to spend some time seeing just how the extra 400 MHz and 2 mb of L3 cache increase performance with a video card like this.