Actually comparing Arm versus x86 is really like comparing apples to oranges. One is RISC, other is CISC. One is good with running multiple threads on single core with highly sophisticated branch prediction, other concentrates having many low powered cores each hoarding threads. And this is very, very crude simplification.
Benchmarks tend to be always favor other camp, depending which attributes you are testing.
Actually Intel chips are since long RISC based, with a CISC translation layer. The real difference is in the more advanced pipeline, prediction and cache and and of course the 32 vs 64 bit + other architectural features. Hence the intel chips use a lot more power than an ARM based processor.
This is not exactly true. In most cases intel's chips are more power hungry but the low power atom processors bring better performance than an ARM cortex A9 while still having lower power consumption. (mostly by racing to sleep faster) And this is a three year old architecture i'm talking about.
Build your own benchmark. Measure whatever you want to do. Or if you insist on standardized testing, use something like SPECInt 2006. Results here: http://www.spec.org/cpu2006/results/res2013q1/ (sadly no ARM results). Of course that is only for integer calculations, if you need fp, use SPECfp.
