Implicit defaults for most modes are good, but we should additionally safe-guard users.
Pushing them to define the scale and mode in which the benchmark is run is arguably good
for making a reliable benchmark suite.