I've handled ~1 TB DBs in Postgres with about 2000 read queries/sec. Technically these were stored function invocations, so wrapped a considerably larger number of queries inside.

This didn't seem to be a problem. It was the simultaneous write operations that created real limits, banging on the disks/disk controller like that.