Not necessarily. The square root hardware could be heavily pipelined to get each stage short enough for one cycle. This increases latency, but lets you keep a o

Not necessarily. The square root hardware could be heavily pipelined to get each stage short enough for one cycle. This increases latency, but lets you keep a one-per-cycle throughput without slowing down the clock.