Since it's in the context of a homelab you usually don't change your hardware for one application, using the same resources in both test seems logical (could argue that the test should be pg vs redis + pg).
And their point is that it's good enough as is.
It's a homelab. If it works, it works. And we already knew that it would work without reading TFA. No new insights whatsoever. So what's the point of sharing or discussing?
In a home lab you can go the other way around and compare the number of requests before saturation.
e.g. 4k/sec saturates PG CPU to 95%, you get only 20% on redis at this point. Now you can compare latencies and throughput per $.
In the article PG latencies are misleading.