Supporting benchmarks that might be interesting alongside the OP:

https://github.com/johnj/llds#wall-timings-in-seconds

Same concept applies, reducing translations for page lookups reduces latency.