How much does printing to stdout in parallel tests skew the benchmark? Isn't stdout be definition synchronized and you are better off benchmarking without stdou

How much does printing to stdout in parallel tests skew the benchmark? Isn't stdout be definition synchronized and you are better off benchmarking without stdout involved?