An NVMe Wear Percentage Is Not a Remaining-Lifetime Clock

Identify what the displayed percentage measures

A used-server offer may include a screenshot labeled SSD health, wear or life remaining. Preserve the actual field label and the reporting tool. A percentage detached from its definition leaves even its direction unresolved.

For example, Kingston's SSD Manager user guide V. 1.5.X.X, revision dated April 2, 2024, describes an SSD Wear Indicator that displays remaining write endurance and decreases from 100. That gauge is not the raw Percentage Used field, whose definition concerns consumption. Do not apply one display's interpretation to another tool's unnamed health score.

References: [K2]

Read endurance consumed without inventing time remaining

Microsoft's NVME_HEALTH_INFO_LOG reference defines PercentageUsed as a manufacturer-specific estimate based on usage and predicted NVM life. At 100, estimated endurance has been consumed, but subsystem failure need not follow. Values above 100 are permitted. The documented update interval is once per power-on hour outside sleep; values above 254 are encoded as 255.

A fictional offer showing Percentage Used 20 alongside a five-year rating has not established four years remaining. That projection assumes the endurance estimate maps directly to elapsed calendar time. Subtracting a percentage from 100 cannot supply that missing relationship.

Likewise, two dated readings show a change over the observed interval. Extrapolating that change into a failure date additionally assumes a future consumption rate and an endpoint that predicts failure. The field supplies neither. Keep workload-qualified DWPD and total-write comparisons in the linked SSD endurance guide.

References: [M1]

Read spare capacity against its own threshold

Kingston's NVMe SMART attribute reference defines Available Spare as normalized remaining spare capacity from 0% to 100%. Available Spare Threshold supplies the comparison point: falling below it is the spare-capacity warning condition.

Thus, a fictional spare reading of 9 with threshold 10 crosses that threshold. These invented values are not a recommended threshold for a product. The relevant threshold must accompany the actual reading.

This field does not measure free user storage or disclose how many physical NAND bytes the manufacturer originally reserved. The linked over-provisioning guide explains reserved-capacity denominators. An Available Spare reading alone cannot reconstruct that capacity accounting.

References: [K1]

Keep warning states beside the wear estimate

Critical Warning is a set of bits. Kingston's reference identifies separate conditions including low spare capacity, a temperature threshold condition, degraded reliability and read-only media. A volatile-memory-backup failure bit applies where that backup solution exists. Multiple conditions can be indicated together.

These are current states, not a persistent incident history. An all-zero warning field therefore does not establish that the device never experienced errors. Kingston separately identifies Media and Data Integrity Errors as a count of detected unrecovered integrity errors, and Error Information Log Entries as another count.

Ask what an indicated condition or recorded error means for the actual drive. A low wear estimate cannot cancel a read-only warning, and a wear percentage cannot replace the omitted warning fields.

References: [K1]

Compare A at Used 20 / Spare 90 with B at Used 110 / Spare 100

Consider two invented summaries, A and B, stipulated to use the same model, capacity variant, firmware and controller-log scope. A reports Percentage Used 20 and Available Spare 90. B reports Percentage Used 110 and Available Spare 100. Neither summary supplies thresholds, warnings or error history. These are interpretation examples, not observed hardware results.

A reports the lower consumption estimate under the shared stipulated implementation. Its spare value does not establish that 90% of user capacity is free. B reports consumption beyond the estimate's 100 point. Its spare reading does not restore unused endurance or contradict the Used reading: the fields describe different quantities.

Calling B failed solely because Used exceeds 100 goes beyond the field definition. Calling it suitable for continued service because Spare is 100 also goes beyond the evidence. The useful conclusion is that both offers need their missing context before the storage owner can assess them against the intended role.

A single combined health score would conceal the distinction that makes this comparison useful. Keep the two readings visible rather than averaging them or choosing whichever percentage looks more favorable.

References: [M1][K1]

Retain the controller, subsystem or Endurance Group scope

The October 4, 2022 NVMe Base Specification 2.0c describes a controller SMART / Health log whose Percentage Used concerns the NVM subsystem. This revision defines no namespace-specific health information. A namespace name on an export therefore does not establish isolated wear measurement for that namespace.

The separate Endurance Group Information log reports endurance and spare fields for the identified group. Figure 207 also defines an Endurance Group Critical Warning Summary that can flag conditions in one or more groups. Preserve the log identity and group identifier; a group reading must not silently become a subsystem reading.

This dated specification provides the definitions used here. Resolve the actual device's supported revision and implementation separately.

References: [NV1]

Resolve what the offered record can establish

For used or redeployed hardware, ask for the underlying report tied to the actual offered device, its capture date and the tool/version that produced it. Retain the exact model, capacity variant, firmware and reporting scope. Kingston's Manager guide illustrates why identity matters: the tool exposes model, serial and firmware information alongside health reporting. Share only the nonconfidential details needed for an inquiry.

Have the supplier or manufacturer explain the applicable estimate and any warnings or errors, then have the appointed storage owner assess the proposed role. Matching model and firmware makes a comparison better defined; it does not make two individual drives interchangeable.

An unattributed screenshot cannot establish that the offered device produced it. A historical report also leaves the intervening use unresolved. Request the missing evidence instead of assigning remaining years to the percentage.

References: [K2]

Bring selected equipment into the physical installation scope

Once the storage owner has resolved equipment suitability, Leviathan's rack-and-stack service can address agreed mounting, IT cabling, labeling and physical installation records. Drive diagnosis and the decision to reuse an SSD remain with the separately appointed storage owner and supplier.

Use the service and project-inquiry links below with a nonconfidential equipment summary, released rack and connection plans, site constraints and required physical handoff. Identify the selected configuration so a later equipment substitution returns to its decision owner.

References: [L1]