Benchmark Results: Battlefield 3 And StarCraft II
Battlefield 3
We’re incorporating a new quartet of games into our 2012 System Builder Marathon test suite, replacing former favorites that often taxed our budget-friendly graphics hardware. Of these new games, Battlefield 3 is the most graphically-demanding.
In order to get acquainted with the game’s hardware demands, (and unlock every mission), I played through the entire campaign on the overclocked December $600 PC (yes, sometimes play accompanies work). The Ultra quality preset was far too taxing at 1920x1080, and frame rates also dropped too low at 1680x1050. But things smoothed out acceptably at 1920x1080 using the High quality preset. Unfortunately, it turns out we won't be using High in our SBM series.
This is obviously a graphics-heavy test, evidenced by noticeable resolution scaling already at Medium detail levels. Armed with a Radeon HD 6950, this quarter's PC leads our previous configuration, despite its respectably overclocked Radeon HD 6870. Even still, both machines are easily playable at test settings really meant for a $500 machine sporting less powerful graphics hardware.
Battlefields 3’s Ultra quality preset was exactly where I hoped the Radeon HD 6950 would flex its muscles and allow the current PC to deliver a better gaming experience than what last quarter's build could muster. There are more graphically demanding areas encountered within the game, so I’d argue that an average of 45 FPS is a safer minimum target than our typical 40 FPS mark.
The current machine leads, delivering good performance through 1680x1050. But the platform still falls shy or where we'd want to be at 1920x1080. Likewise, the former PC is capable through 1280x1024. But, as mentioned, it's too slow at anything higher than 1680x1050. This quarter's build takes a victory in Battlefield 3.
StarCraft II
This 60-second Fraps-based benchmark is demanding right at the start, but eases up as enemies are eliminated from the map. We’ll stay consistent by running the same four resolutions, but note that in StarCraft II, lower settings don't do the game justice.
Including 1280x720 in our testing often helps establish where bottlenecks are happening. If you remove that resolution, we're going to be inclined to overemphasize the importance of the GPU. But because we have it, we see a drop in performance as the aspect ratio gets wider, indicative of a CPU limitation. The fact that the Core i5-based machine easily wins at every resolution backs up our hypothesis, while also providing evidence that this game exploits more than two processing cores.
Similar CPU-limited results emerge at the highest graphics and texture settings, although the two builds are now grouped more tightly together. It’s interesting to note that the stock Radeon HD 6870 is definitely working hard at 1920x1080, as average frame rates drop off more precipitously, while the Radeon HD 6950 in the current build is still held back by its host processor, losing only 2 FPS.
Both systems breeze through this multi-player map. While graphical demands bring the average frame rates of our stock machines at 1920x1080 to within 2 FPS of each other, minimum frame rates are more CPU-limited. The current build never drops below 72 FPS, while the December PC maintained 10 more frames per second than that. Despite very playable performance from both machines, we must award victory to last quarter's PC. Its more powerful processor delivers smoother performance in the game's biggest and most intense battles.