1. Identify the factory bad-block class
KIOXIA's historical brief describes initial bad blocks detected during final die testing before customer shipment. It relates their presence to the NAND manufacturing process and distinguishes them from blocks that become bad later. The initial category therefore concerns the manufacturing and screening history of physical storage blocks. It is not, by itself, a record of previous customer operation. A description that treats every bad block as evidence of field wear would erase the distinction made by the manufacturer.
The category also explains why a generic unused description cannot settle the physical-block question. Absence of prior customer use and absence of manufacturing bad blocks are different assertions. The first concerns a usage history; the second concerns the storage structure identified through supplier testing. KIOXIA discusses how initial blocks are excluded from use, rather than redefining them as healthy because the product ships. This is a conceptual explanation from the September 2022 brief. It supplies no allowed count, screening instruction or conclusion about whether any particular surplus device contains an initial bad block.
2. Distinguish blocks that become bad in use
The brief separately describes bad blocks developing as NAND is used. Its overview links write/erase activity with stress on memory cells, and its functionality prose discusses operations that fail in the field. This later category has a different temporal basis from manufacturing screening. A block can enter the bad-block category during use rather than being identified as initially bad before shipment. Retaining that distinction gives the phrase bad block a more precise history context without reducing it to one assumed cause for every device.
Retirement describes what management does with a block identified as bad: it is marked and no longer used. That action is different from the failure event that led to the classification. The failure concerns the physical storage operation; retirement concerns subsequent exclusion from the usable mapping. A replacement decision does not reverse that classification or demonstrate restoration of the original block. These are mechanism meanings, without diagnosing an offered package, interpreting a status value or turning an architectural account of wear into a remaining-life estimate for the material.
3. Explain reserved replacement and logical identity
KIOXIA describes reserved blocks used as replacements and a logical-to-physical table that manages the relationship. The logical view identifies storage as presented through the management context, while the physical view concerns the actual blocks holding data. Assigning a reserved replacement changes which physical resource serves the logical relationship. It need not change the logical identity in the same way that it changes the physical block used beneath that identity. The logical identity and the physical resource can thus be discussed separately within one replacement event. This is the reason a mapping description contributes information that a physical bad-block label alone does not contain.
This distinction explains why replacement and repair are different descriptions. A reserved block provides another physical resource; the retired block remains excluded. Mapping connects the storage view to the resource currently assigned, rather than making a failed physical block healthy again. The reserve and the mapping also have separate roles: one supplies replacement resources, while the other records their relationship to the logical view. That account needs no address example, table construction or capacity arithmetic. It identifies the mechanism behind replacement while leaving the exact product's allocation policy, current reserve state and observed operating result outside the historical brief's use here.
4. Separate physical flags from managed visibility
The brief describes initial raw-NAND bad blocks as flagged for identification and managed-flash initial bad blocks as not used internally. This is a distinction in how a physical-block fact is represented or handled in the product context. A flag identifies a block at the physical level. A managed logical view can avoid presenting that block for ordinary use. The views concern different layers, so absence of a directly presented physical flag is not itself proof that no initial bad blocks existed.
Only this view distinction is needed here; the broader allocation of host and controller tasks belongs to the separate raw-versus-managed guide. It prevents a logical storage description from being read as a complete physical-block history. Likewise, an erased-data description concerns stored information, while a bad-block classification concerns the resource used to store it. One cannot establish the other by changing the label. The resulting explanation keeps initial screening, later retirement, reserved replacement and visibility separate. It provides no marker location, scan or erase procedure, health-register interpretation, acceptance criterion or present block count for offered devices.
Sources and further reading
KIOXIA Understanding Bad Block Management in NAND Flash Memory, September 2022 Rev 2.0: physical page 1 overview/functionality prose only. Figure 1, page 2 health material, numerical criteria, marker details and current-standards claims are excluded.