Rendering Benchmark
9 January 2013
Tags
- Fairy Tale Twist
- Performance
Testing with a pure rendering example.
Adobe Air
- Tested on iPhone 4 (retina, 512MB), Release Build 126 sprites drawn @ 60 fps
- Tested on iPhone 4 (retina, 512MB), Export Release Build 292 sprites drawn @ 60 fps
- Tested on iPhone 4S (retina, 512MB), Export Release Build 56.4fps, average frame time 18ms, 2 draw calls per frame 644 sprites drawn @ 60 fps, 64MB RAM (15MB game objects, 18MB bitmaps, 1.2MB other), 35MB GPU RAM, 7MB download, 95% CPU utilization
Cocos2D-X
- Tested on iPhone 3GS (256MB) 891 sprites drawn @ 60 fps
- Tested on iPhone 4S (retina, 512MB) 5000 sprites with batching, 800 sprites drawn @ 60 fps, 50% of the time is spent building display lists while only 18% is spent in benchmark code
The app also spent 13% of the time waiting for the frame swap, i.e. we are GPU limited not CPU limited. Finally, 10% mach_msg_trap is overhead for iOS, this would decrease if we were CPU limited/had more code.