Planning
SSD Powered-Off Retention: Keep Temperature and Wear with the Rating
Programmed data in nonvolatile NAND can remain stored without external power. The useful buying claim is a supported retention interval at a stated temperature after stated prior use. Equal capacity and rated host writes do not establish equal powered-off retention. In the fictional comparison below, one year at 30°C covers the requested six months at 30°C on matching stipulated premises; three months at 40°C leaves that requirement unresolved.
Define the period in which the stored data must remain available
A server may finish a project with input data still on its SSD, then remain unused until the next phase. The question is whether that already stored data is supported through the intended period without drive power. Start with the required interval and temperature conditions, together with the use that precedes the hold.
Describe the retained data and the point at which it must be available again. Six calendar months after the last use is a different requirement from an unspecified later restart. Keep the anticipated temperature exposure attached to that interval; a duration copied from an offer without its temperature has lost part of its meaning.
The evidence must also apply after the relevant prior writing. A statement about a new, lightly used drive leaves a different question from a statement covering the offered device after its rated writing. For a used drive, the supplier must relate the applicable retention claim to its supported use history. A write total or wear display alone cannot fill in an omitted retention interval.
Understand why wear, time and temperature stay together
Intel's June 2012 application note, order number 325999-002US, explains retention using the floating-gate NAND of that period. Charge loss can change stored cell states, while program/erase cycling affects retention capability. The note identifies accumulated cycling, storage time and temperature, and the timing and temperature of prior cycling as relevant factors. Internal data moves require power.
Its JESD218A discussion places powered-off retention after writing to the endurance rating under the applicable workload and use basis. The important relationship is therefore prior use followed by a supported non-use interval. These historical explanations supply no numerical prediction for a modern drive.
References: [RET2]
Read the historical retention-use column with its active-use column
SNIA's Endurance of NVMe, SAS, and SATA SSDs contains Table 1, SSD Classes and Requirements JESD218B.01. The main text carries a 2021 copyright. It distinguishes active use with power on from retention use with power off.
Its client row pairs active use at 40°C for 8 hours/day with powered-off retention for one year at 30°C. Its enterprise row pairs active use at 55°C for 24 hours/day with powered-off retention for three months at 40°C. Those are different use bases, rather than two universal promises attached to the words client and enterprise.
The table and Intel's earlier JESD218A interpretation establish the historical context used here. They are not a reading of the current JEDEC standard or a product guarantee for an SSD now being offered. Ask which requirements the actual product documentation adopts.
SR-45: equal 4 TB offers, different evidence for a six-month hold
Consider two fictional offers, A and B. Each supplies a 4 TB SSD and states a rated host-write quantity of 1,000 TB. TB is decimal: 1 TB = 10^12 bytes. The 1,000 TB is cumulative host writing, not retained dataset size. These are invented offer terms for explanation, with no measured result or actual product selection.
Both offers explicitly adopt their respective historical application-class qualification bases after rated writing. A states one year of powered-off retention at 30°C on its client basis. B states three months at 40°C on its enterprise basis. For each offer, the applicable workload, prior-use and cycling premises are expressly stipulated to match its declared basis. The example does not infer those premises from the two equal host-write quantities.
The buyer requires already programmed input data to be retained through six calendar months without drive power at 30°C, after the stated prior use.
A's declared one year at 30°C encompasses six months at that same temperature. With the stipulated prior-use applicability, A establishes the requested retention interval. No conversion from DWPD, wear percentage or host-write quantity is needed to reach that conclusion.
B's declared three months at 40°C does not establish six months at 30°C. The required duration is longer and the temperature differs. A lower temperature can matter to retention, but these two offer fields provide no supported numerical extension. The buyer needs additional applicable evidence for B at the required interval and temperature.
B has an unresolved requirement; it has not been shown to lose data at three months or to fail the six-month requirement. Equally, A closes only this stated retention comparison. It does not establish enterprise compatibility, a complete application checkpoint, backup sufficiency or protection against every failure mechanism. The client label is not a recommendation to install A in an enterprise server.
Separate the transition into NAND from the time data remains there
Samsung's 2014 power-loss-protection application note describes capacitors temporarily powering an SSD so pending data can move from volatile DRAM into nonvolatile NAND after external power disappears. That explains a protection mechanism during the interruption.
The subsequent months without power are a separate interval. Completing a cache-to-NAND transfer does not specify how long the programmed data is retained under the later temperature and wear conditions. For SR-45, adding PLP to B's offer would leave the missing six-month retention evidence unresolved. The related PLP guide examines protected writes and mapping metadata in more detail.
References: [PLP1]
Obtain a response that answers the actual non-use requirement
Send the supplier a specific retention question tied to the offered model and capacity variant, relevant firmware applicability, documented prior use, intended powered-off interval and expected temperature conditions. Ask for the applicable product statement and an explanation of how those premises fit the requirement. The supplier should identify what its temperature value describes and which exposure conditions its statement covers.
For SR-45, a useful clarification would explicitly establish whether the offered drive supports retention of the programmed data for six calendar months at 30°C after the stipulated rated writing and prior-use conditions, with the supporting product evidence identified. Repeating 4 TB, 1,000 TB of host writes, enterprise or PLP does not answer that request.
If B returns applicable evidence covering the requested interval on matching premises, its retention comparison can be completed using that evidence. If the response expressly excludes the required conditions, the offer does not meet this criterion. If the response merely repeats three months at 40°C, the criterion remains unestablished. Resolve that gap with the supplier or compare another offer that addresses it; do not turn missing evidence into a predicted failure date.
For A, retain the one-year statement together with its temperature and prior-use basis when recording why the six-month requirement is covered. Revisit applicability if the actual non-use period, temperature exposure, offered variant or prior-use assumptions change. The result is a bounded equipment-selection decision that the appointed storage owner can use, rather than a generic declaration that an SSD is safe in storage.
Carry the selected equipment into a separate physical work scope
A powered-off retention assessment concerns stored data. Physical custody, equipment storage conditions and return to installation require their own agreed arrangements. Retention evaluation and data recovery remain separate from equipment handling.
If the selected server also needs physical installation, Leviathan's rack-and-stack service covers agreed equipment mounting, IT cabling, labeling and installation records. Use the service and project-inquiry links below with a nonconfidential equipment summary, released rack and connection plans, site constraints and required handoff. Identify the storage-selection owner separately so any equipment substitution returns to that owner for review.
References: [L1]
Primary sources
- [RET1] SNIA: Endurance of NVMe, SAS, and SATA SSDsChecked 2026-10-20. 2021 main text; Table 1, JESD218B.01, PDF page 4. Historical application-class context, requiring product-specific applicability.
- [RET2] Intel: Data Retention in Intel Solid-State Drives, 325999-002USChecked 2026-10-20. June 2012, printed pages 4–11; public Solidigm-hosted copy. Historical NAND explanation and JESD218A interpretation, without modern-device extrapolation.
- [PLP1] Samsung: Power loss protection (PLP), SSD Application Note for Data CentersChecked 2026-10-20. 2014 note, printed pages 01–02. Explains interruption-time cache-to-NAND protection; its product specifications are not transferred to another drive.
- [L1] Leviathan Systems: Rack & Stack / ColocationChecked 2026-10-20. Existing physical equipment installation and records. The agreed project defines the included work.