Investigations into Socket 939 Athlon 64 Overclocking

Battlefield 2 Performance

We're still working on coming up with some good benchmarks for Battlefield 2. For this test, the sample demo from Guru3D was used, which runs through the Operation Clean Sweep map. Unfortunately, there is only a single player in the map, so it's not a true indicator of performance. It does show the use of several vehicles, and it crosses a large distance, which is good for stressing memory usage. The results still seem too high, however.

Update (10/08/2005): I finally figured out the problem with the initial benchmark tests. Since I haven't done a lot with BF2 benchmarking, I made a serious error. It turns out that by default there's a 24 second pause after the map is loaded before the actual demo playback commences (unless you press ESC or click Join Game). The problem is that the benchmark already begins to record frame rates at the menu screen, so for 24 seconds the computer is rendering 200 to 1000 FPS. That skews the results so heavily that I have chosen to remove the graphs and information. I'll cover proper BF2 performance (with a proper demo file) in the next installment. The loading times below are still correct, so they can remain.

We also timed how long it took to load the level for benchmarking (after disabling the intro movies). We recorded the time for the initial load as well as the time for the second load of the same level after exiting and reloading.

It's interesting to see that the load times were also dramatically impacted by CPU performance, loading about 25% faster at 2.7 GHz than at 1.8 GHz. We've also done some testing (not shown here) of performance with 2x1024MB of RAM vs. 2x512MB of RAM. The 2GB RAM setups gain another 10 to 20% in loading speed, particularly on the second load. Unlike BF2 rendering performance, value RAM and performance RAM are nearly identical in load times - less than 2% at most CPU speeds.