http://www.techspot.com/review/448-battlefield-3-beta-performance/page5.html
Doesn't use MSAA.
I ain't saying that H's review is flawless but when I look at the other reviews I find concerns. Oh well, lets see what surfaces on release.
Just wanted to quickly point out that Techspot's review states, or at least implies via its wording, that they were using the Ultra and High
presets. As they are presets, the game... and I quote from the
testing methodology page...
"When set to Ultra every setting is maxed out with the exception of anti-aliasing post which is set to medium. The anti-aliasing deferred settings is set to 4xMSAA while anisotropic filtering is set to 16x. Other quality settings such as texture, shadow, effects, mesh, terrain and terrain decoration are all set to Ultra.
The Ultra preset was extremely demanding, so we also tested using the High quality preset. This turned anti-aliasing deferred off and left the anti-aliasing post settings on medium. All other visual quality settings as detailed above are turned to high. We'll be looking for an average of 60fps for stutter-free gameplay."
In other words, the High preset is what is turning off MSAA, not the tester. Furthermore, anything the tester did change from the preset would thus change the
"Graphics Quality" setting in the screenshots to 'Custom' from 'Ultra' or 'High', indicating tickering.
I'm not saying the techspot review is perfect either, but it certainly does seem to be the data with the most integrity in that group of links. Even more so because it compares different CPU performance (a lot of people, including myself, still have those legendary i7 920s and didn't have a reason to upgrade to every testers poster child Sandy Bridge) really well, and is something not even my preferred GPU testers (Anandtech/Tom's) do often enough.
Finally, for the record, all of these performance reviews both linked and elsewhere while interesting for the curious, are all rubbish to the objective. With the beta comprised of one map (I would've loved to see data on Caspian Borders, though I can imagine the difficulty a tester would've had pulling that off), incomplete Ultra engine implementation, rushed Beta drivers from both nVid and AMD and a bug filled play environment with map clipping, unoptimized foliage and animation glitches abound, none of this means anything until its November, nVid and AMD put out WHQL drivers (or at least further optimized preview ones) and a comparison is done between the FPS results of a tight Squad Deathmatch versus a 64 player Conquest map taking place in the desert or urban environments EA were so happy to market in Fault Line and E3.
I'll be particularly interested in an analysis of what exactly 'Anti-Aliasing Post' does exactly and why it remains at 'Medium' when Anti-Aliasing Deferred' itself is turned 'Off.'