Kioxia rates the CM9-R 15.36TB E3.S drive for 11,000 MB/s sequential write. The wire article and the specification table do not say the same thing. Its line that the CM9-R "should offer better write speeds" is the weaker claim. ServeTheHome's specification table is the document I trust, and the number that decides it is 11,000 MB/s.

That is a ceiling, not a workload. A buyer sizing a high-availability array should not treat the dual x2 Gen5 option as proof that two hosts can pull the same peak without contention. Spec table proves the interface and the rated peak; it does not prove sustained write behavior, thermal throttling, or dual-port load. For the fab, the constraint is endurance: 1 DWPD over five years sets the replacement math, not the 14,800 MB/s read number.

Capacity sits in a 7.5mm E3.S body, and the BiCS FLASH generation 8 TLC stack makes the 1 DWPD rating plausible. Single x4 and dual x2 modes change the host, not the NAND. Dual-port arrays get redundancy at the connector level, but the table still does not show how the controller splits load between ports.

What remains unmeasured is the drive after the rated peak. Dossier leaves out sustained write behavior, thermal throttling, dual-port load, measured write endurance, and power draw. Those gaps matter more than the faceplate or the 7.5mm thickness.

The rated 11,000 MB/s write peak is a ceiling, not a workload guarantee. The binding constraint for a high-availability buyer is 1 DWPD over five years, not the dual x2 Gen5 option.

The document leaves out sustained write behavior, thermal throttling, dual-port load, measured write endurance, and power draw.

Treat the spec table as the source of truth for interface and endurance. Do not size dual-port load or sustained write from the review's expectation.