calin_banc

It's not only about the quality (or how much calculations is doing every unit), but also about quantity, numbers. You don't actually have tens or hundreds of units at the same time, on the same map in Crysis. The difference between SP and MP, might be explained through render distance/object distance (you have them lower in SP). If you increase them, you'll need a rather much faster CPU than a faster GPU. Perhaps it's the draw calls or something like that 'cause the GPU usage drops like a stone when those two (draw/object render distance) go above a certain value.

Does ArmA 3 even starts on single core CPU? The low specs are: Processor:Intel Dual-Core 2.4 GHz or AMD Dual-Core Athlon 2.5 GHz

The problem with the ArmA's engine, it's that it doesn't utilize very well the CPU. To see that you're limited by the processor due to poor scaling on multiple cores it's quite annoying. At least Cryengine 3 and Forstbyte can use up a CPU a whole lot better. We'll see how things will set in with the final version, but I don't believe in miracles at this point.

http://www.overclock.net/t/1362591/gamegpu-crysis-3-final-gpu-cpu-scaling AMD CPUs can hold their on ground if the game is programed accordingly with that architecture and it can come equal with the latest i7 CPUs. It's up to Bohemia to do that optimization code along side AMD - not the very best ultra high end i7, but still a good choice.

100% will performance on most computers.

BF 3 is CPU heavy due to the physics in it. For multi GPU setups you'll need something with more than 4 threads (2600/2700 and so on) in order for the GPU usage to stay up and avoid some low FPS. Also, Frostbyte uses more of the CPU compared to ArmA's 2 engine. Although the maps are smaller, the game looks and runs much better and more importantly, the control of the character and the handling/feeling of guns, how they shoot and the way they look, are far ahead of ArmA 2. There is really no point in comparing those two. Of course, a modded version of ArmA can add some nice stuff and take back some territory, but in the end is not a very smooth experience, but it has it's own strong points that still keeps us playing it. Regarding this Alpha, personally I wouldn't expect much of an improvement of performance in the final version. Some tweaks here and there, but the overall numbers should be the same. All the games I've played so far (not much, but still) that had a pre-launch testing time, ran +/- the same in the final version. If BIS is the exception from this rule, than bravo!