if you want to use partitions, you want to list out all the partitions. most modern NVMe devices perform best with ~4 partitions, but ideally you want to ensure that the number of ‘devices’ does not exceed the number of CPU cores.
how did you overprovision? are you sure your model of disk benefits from OP? a lot of disks dont. ideally you should test performance with ACT before/after overprovisioning to ensure you’re actually getting some gain.
We are using aws i3 instances with nvme devices. The number of partitions is equal to number of cores on the machine, so we end up with 8 partitions of 200GB each per device.
Did not benchmark overprovisioning, but aws does not OP their disks.