Current x86 chips are spending huge amounts of their transistor and power budget on tiny increases in instructions per clock.

When you have twice as many cores to play with then you can have better total performance with a lot less resource cost.