I think it’s pretty clear what the performance results are going to be here so there isn’t a lot of purpose on piling on the bad news. Our updated custom performance scripts make a couple of assumptions now for the all-important Observed FPS data:
- A dropped frame is noted and considered a 0ms result.
- An interleaved frame is added to the successive color for frame time. This means that we don’t count a dropped frame but instead consider it “runt” like in terms of how it affects the gaming experience.
Battlefield 3 sees a very similar drop in FRAPS to Observed frame rates for the CrossFire setup as we saw with our non Eyefinity testing; a pair of dual HD 7970s GHz Edition looks very similar to a single card. The Frame Times data indicates we are seeing nearly every other frame dropped from our testing resulting in half the perceived frame rate as reported by other performance metrics.
GRID 2 was a game that heavily witnessed the frame interleaving problem and because of it the frame times graph shows a combination of that anomaly with dropped frames as well. There is basically no observed scaling with CrossFire at 5760×1080.
Clearly for both AMD's Radeon HD 7970 GHz Edition and the NVIDIA GeForce GTX 770, single GPU results are right where you expect them to be – no frame pacing issues or other hiccups to be found. NVIDIA's SLI does have some issues with GRID 2 maintaining a consistent frame rate though AMD's CrossFire is obviously the outlier in both games.