The issue is not Postgrex, but rather the fact it is comparing Ecto (which would be akin to a ORM):
With something that directly sends INSERT commands over the connection:
https://github.com/antonputra/tutorials/blob/main/lessons/230/go-app/device.go#L30-L33
And if you are going to set up 500 database connections, I would at least split over a few connection pools (by setting pool_count). Otherwise you are likely making the pool the bottleneck.
Overall, I’d expect the Go to have better throughput, but there is definitely a bit of apples to oranges going on. I honestly don’t understand why benchmark authors do not ask for some community vetting before publishing.
