[monocasa]

But there's more work being done per average x86_64 instruction due to RMW ops. Hence why they just look at an entire binary.