New SSD Lot Slower but Within Spec: When to Hold Release
- Slower benchmark results within published datasheet minimums do not inherently prove an SSD lot is defective.
- Baseline drift must be verified by standardizing host hardware, firmware, capacity, fill level, preconditioning, thermals, and workload.
- A written acceptance failure or credible application timing risk can justify a documented hold pending investigation.
A new SSD lot can benchmark below a reference lot yet still meet its published specification. That result calls for investigation, not an automatic defect label. The release decision depends on which limit the project actually agreed to and whether the device's application workload remains safe.
Keep three comparisons separate
| Comparison | Question |
|---|---|
| Datasheet statement | Is the measured result comparable with the vendor's stated test conditions? A headline or typical number is not necessarily a guaranteed minimum. |
| Purchase or qualification criterion | Did the buyer and supplier agree to a particular test and acceptance range for this part and revision? |
| Application service limit | Does the lot meet the required throughput and latency during the real peak workload, including recovery? |
Micron's SSD measurement brief shows why fill state, write history, preconditioning and system configuration can change observed performance. Compare the reference and new lots sequentially under the same controlled setup, or use equivalent calibrated setups. Match capacity, firmware record, host, interface, driver, temperature, workload, queue depth, run duration and metric. Capture distribution or percentile latency where the application depends on tail behavior. Repeat enough units and runs for the project's sampling plan; a single pass is weak evidence.
Decide whether to hold
- Re-run a retained reference sample. If it too has shifted, inspect the fixture and software before blaming the lot.
- Request the supplier's change record, relevant PCN and firmware information. A change notice may explain a difference but does not itself establish acceptability.
- Compare results with the written acceptance test. If that test fails, quarantine the lot under the agreed quality process and notify the supplier.
- If the written test passes but application timing is at risk, pause release while engineering checks the production workload and decides whether additional qualification is needed. Record the decision owner, affected units and disposition.
- If both criteria pass, document the observed shift and monitoring plan. Avoid claiming that every slower lot is defective—or that every in-spec lot is safe for this application.
For a project-specific comparison, include both lots' measurements, test conditions and acceptance limits with a quote request.
References
We publish measured usable capacity and define trial-batch verification against the quoted product and agreed test scope.