To give a good idea of ARM performance in practical situations (so you know what to expect), the resident ffmpeg ARM guru did benches with Cortex A8 and A9 for H.264 decoding without assembly optimizations.<p>With a Core 2 reference at 1.0, a Pentium 4 at 0.38 in terms of performance per clock. The Cortex A8 is 0.28 and the Cortex A9 is 0.4. The difference there is mostly the out-of-order execution added in the A9.