CPU scaling from 2.67ghz to 4.4ghz. At 3.8ghz my GTX275 started to become the bottleneck.
(this was conducted on v1.02)
Settings: Very High. Memory was at or as close to 1600mhz CL8 as possible
R1 is the first run, and averaged of 2 1st runs.
R2 is the average of 3 runs after the 1st run.
NOTE: also keep in mind that i was raising my CPU frequency by raising the base clock (BCLK). On an i7 920 this means the whole motherboard was being overclocked. To get a better picture I would have to do these tests again with 1 BCLK and just change the multiplier, but I believe results would be the same within 1 or 2 FPS. 1-2 FPS can also be the variance in just running ArmA2Mark over and over again.)
Core i7 920
6GB (3x2GB) DDR3 1600 8-8-8-24
GTX275 (712core/1584shader/1296 memory)
WD Raptor 150GB
More memory bandwidth seems to help A2 over memory with lower latencies as well. This isn't necessarily a CPU optimization issue, but this tends to point in the direction of LARGE amounts of data being moved and processed by the CPU. If you look at a Memory Scaling review like one at TomsHardware and Anandtech you will see that the types of programs that perform better with faster ram are compression and rendering programs (WinRAR and Cinama4d, Lightwave) and games that load a a lot of data (HAWX, farcry2). In what I observed with A2, the game did not stutter as often, which feels more smooth.
compared to ArmA 1, which likes lower timings.
I had the intent of trying more settings (up to 2000mhz on the ram, and CL10 and 12 for 1600mhz) but so far nothing was very drastic and it is very time consuming to do these tests.