Some theories: (a) laziness (b) wanting to avoid discussions about whether the exact way something was written was fair to some GC or another (c) wanting to

Some theories: (a) laziness (b) wanting to avoid discussions about whether the exact way something was written was fair to some GC or another (c) wanting to avoid "can't reproduce these exact numbers on a different CPU or different JDK version" discussions (d) the code doesn't matter since these microbenchmarks are not the actual application you are interested in Seriously, if you are interested enough in this that you would be willing to run the benchmarks, you presumably have some JVM workload you care about. (Or you could just grab a standard benchmark suite.) If you are interested enough, you can run your workload with the different GC flags and see what they do for you, in your context. Being able to reproduce the author's microbenchmark numbers on your laptop would not tell you anything about how your workload of interest runs on the big servers (or whatever).