Variance in Array vs TypedArray performance test

I am comparing the operation of reverse in Typed vs Standard Arrays i.e:

const a = new Array(1e6).fill(0).map( () => Math.random())
a.reverse()
const b = new Float64Array(1e6).map( () => Math.random())
b.reverse()

Certainly we expect Typed to be faster and maybe also to be narrow in variation from the mean time.

However, the standard arrays are slower (expected) but the variation around the mean is very large in comparison. In my computer this shows 3% in TypedArrays, 30% in Arrays.

Conceptually, what is the cause of this ?

See Test