It's pretty well known graph in traffic engineering -- Max throughput isn't free flowing traffic. (Though I remember it at 35-40 mph on interstate/motorway). T

It's pretty well known graph in traffic engineering -- Max throughput isn't free flowing traffic. (Though I remember it at 35-40 mph on interstate/motorway). The basic reason is that free flowing traffic requires bigger gaps between vehicles. (safe anyway, but practically the way people drive generally follows the safe spacing function, but with a 'bit' of y offset). At the minimum latency (the free flow speed (ffs)) the spacing between cars is x seconds, so you get 3600/x cars per hour. At maximum flow rate, ~1/2 the ffs, the spacing is more like 1-2 seconds, so you get closer to 2k cars per hour. Even slower, the cars can be packed closer, but you don't clear the road nearly as quickly so the throughput is lower. See https://courses.washington.edu/cee320ag/Lecture/Freeway_LOSd1.pdf https://courses.washington.edu/cee320ag/Lecture/Freeway_LOSd... (page 13) and https://courses.washington.edu/cee320ag/Lecture/Freeway_LOSd2.pdf https://courses.washington.edu/cee320ag/Lecture/Freeway_LOSd... (page 9) It's a little bit like the tradeoff between latency and throughput in servers.