Turn it on and off in CCC and see the performance drop.
can do...give me a few minutes to run a benchmark
edit: a few benchmarks I ran using Crysis 2 and the Adrenaline benchmark app. I ran the central park benchmark as it has a lot of brick and rock structures that use tessellation in varying degrees and it's a repeatable test. Everything was run on Ultra in DX11 using the high resolution textures, 1920x1200, 0x AA, 16x AF. I changed the Tessellation setting inside CCC.
No Tessellation: Min 12.4fps average 42.4fps
AMD Optimized: Min 7.7fps average 28.5fps
Application controlled: Min 7.7fps average 28.5fps
8x Tesellation (used the slider in CCC): Min 9.7fps average 36.7
So this tells me either AMD optimized doesn't work or there is no Crysis 2 profile so it defaults to whatever the game uses. The slider works to control the tessellation level and to my eyes I could not notice a huge difference in quality, but the fps was better. The minimums were severely below the level you'd like to have by any stretch, but the averages were 8fps higher which is pretty good. No tessellation still looks good but there is a noticeable depth difference in brick surfaces and some rock structures. However, the FPS is still too low at my standard resolution to give a good experience if you judge the minimums as a basis for playability. I could not tell how long the fps dropped to this level, it is possible the timedemo started recording and read low fps prior to a fully loaded level, I do not know for certain. However, below 30fps average is substantially lower than I could play the game.
In Battlefield 3 There was no way to make a 100% repeatable test as I have not found any benchmark apps for the game. So I ran everything on ultra, MSAA off, blur on and slider set to the middle. 1920x1200 16x AF and ran a few times through Operation Guillotine's first few minutes with fraps benchmark. This mission in single player won't tell you how MP will perform but it seems to have some usage of tessellation with the terraign and rocks, walls, and structures etc. Also realize with this method there is a margin of error as it's impossible to do everything exactly 100% the same each time, but I did my best to repeat it.
No Tessellation: Min 45fps, max 82fps, Avg 62.58fps
Tessellation 8x slider in CCC: Min 42fps, Max 86fps, Avg 61.79fps
AMD Optimized: Min 39fps, Max 85fps, Avg 58.8fps
Application Controlled: Min 33fps, Max 82fps, Avg 55.91Fps
So I'm going to say that between no tessellation and using 8x, there is almost no difference whatsoever in the performance. I don't know what AMD Optimized is setting it to but there is clearly a profile in the drivers or CAP, but it's set lower than Application Controlled (64x?) and higher than 8x, or perhaps there was some dynamic tessellation rendering going on where things were adjusted on the fly but I don't know how it works so I could not say. There was almost zero visual difference as I believe there isn't much tessellation at all in the engine. There's definitely some because as you turn it down the game gets higher fps, but I don't think anyone should just turn it off here. 8x seems to be the sweet spot generally. It's only a few fps difference, but it's something.