Do not treat one number as the whole diagnosis

SSD health percentages are derived from controller telemetry and vendor interpretation. Read the raw NVMe SMART fields, firmware and time series instead of assuming that 99% means immediate failure.

The concern is the relationship between wear, actual writes and error indicators. A steady decline after heavy writing differs from a large drop on a nearly idle system.

• Save dated SMART snapshots with percentage used, host writes, critical warnings, media errors, firmware and temperature.
• Judge the rate of change against actual writing and errors, not an isolated rounded health percentage.

Use the 1,200TBW limit correctly

Samsung’s warranty table lists the 2TB 990 PRO for five years or 1,200TBW, whichever occurs first. TBW is a warranty condition, not a promise that every drive fails at that exact point.

Estimate remaining endurance from measured workload cautiously. Workload, temperature, firmware and write amplification can change the rate.

• Save dated SMART snapshots with percentage used, host writes, critical warnings, media errors, firmware and temperature.
• Judge the rate of change against actual writing and errors, not an isolated rounded health percentage.

Record total host writes over time

Capture Total Host Writes on the same day and in the same unit. Compare daily or weekly change with expected workloads such as video caches, virtual machines, game updates, databases and AI model processing.

A background process can write far more than the user realizes. Find the source before blaming normal endurance consumption on a defective SSD.

• Save dated SMART snapshots with percentage used, host writes, critical warnings, media errors, firmware and temperature.
• Judge the rate of change against actual writing and errors, not an isolated rounded health percentage.

Interpret Percentage Used

NVMe Percentage Used is intended as a normalized endurance estimate. It may not move in perfectly linear one-percent steps and software can display rounded values.

A persistent rapid increase without corresponding host writes deserves firmware verification and Samsung support evidence, especially when the drive is within warranty.

• Save dated SMART snapshots with percentage used, host writes, critical warnings, media errors, firmware and temperature.
• Judge the rate of change against actual writing and errors, not an isolated rounded health percentage.

Critical warnings change the priority

A non-zero critical warning, increasing media/data integrity errors or repeated unsafe shutdowns requires more attention than a rounded health percentage alone.

Back up important data and save the SMART report before running benchmarks or firmware changes.

• Save dated SMART snapshots with percentage used, host writes, critical warnings, media errors, firmware and temperature.
• Judge the rate of change against actual writing and errors, not an isolated rounded health percentage.

Historical 990 PRO health concerns

The 990 PRO had widely discussed historical reports of health declining unexpectedly, followed by firmware action. Current diagnosis must use the actual unit’s firmware and telemetry rather than assuming every 990 PRO is affected.

Do not erase the history, but do not turn it into a universal failure claim. Samsung’s current official firmware page is the controlling source for present release information.

• Save dated SMART snapshots with percentage used, host writes, critical warnings, media errors, firmware and temperature.
• Judge the rate of change against actual writing and errors, not an isolated rounded health percentage.

Temperature and health readings can interact

Samsung’s firmware notes include a release addressing high-temperature reports logged in Magician. Confirm firmware before interpreting an isolated software temperature or health display.

Actual sustained overheating remains a separate problem. Check airflow, heatsink contact and throttling evidence.

• Save dated SMART snapshots with percentage used, host writes, critical warnings, media errors, firmware and temperature.
• Judge the rate of change against actual writing and errors, not an isolated rounded health percentage.

Unsafe shutdowns and power events

Unexpected power loss can increment unsafe-shutdown counters without proving NAND wear. Repeated events can still increase filesystem and stability risk.

Investigate power supply, sleep-state behavior, laptop battery events and forced resets when the counter grows unexpectedly.

• Save dated SMART snapshots with percentage used, host writes, critical warnings, media errors, firmware and temperature.
• Judge the rate of change against actual writing and errors, not an isolated rounded health percentage.

Run diagnostics after backup

Use Samsung Magician’s health and diagnostic features when the drive is stably detected. Short non-destructive checks are appropriate before heavier testing.

Do not repeatedly stress a drive with media errors, intermittent disappearance or important inaccessible data.

• Save dated SMART snapshots with percentage used, host writes, critical warnings, media errors, firmware and temperature.
• Judge the rate of change against actual writing and errors, not an isolated rounded health percentage.

Calculate a practical write rate

Divide the increase in host writes by the number of observed days. Compare the annualized rate with 1,200TBW, but present it as a planning estimate rather than a failure countdown.

A 100TB-per-year workload is materially different from 5TB per year. The tool should make that difference visible.

• Save dated SMART snapshots with percentage used, host writes, critical warnings, media errors, firmware and temperature.
• Judge the rate of change against actual writing and errors, not an isolated rounded health percentage.

Prepare a warranty case

Save proof of purchase, the exact model and serial, firmware, SMART reports, screenshots and diagnostic results. Samsung coverage may depend on seller authorization and is not transferable under the published terms.

Do not sanitize or return the drive before understanding the data-security implications and support instructions.

• Save dated SMART snapshots with percentage used, host writes, critical warnings, media errors, firmware and temperature.
• Judge the rate of change against actual writing and errors, not an isolated rounded health percentage.

Replacement strategy

When errors or instability accompany health loss, move workloads to a verified replacement before the old SSD becomes inaccessible. A secondary backup remains necessary even with a new drive.

Compare the 990 PRO with alternatives by current price, cooling, platform generation and warranty source.

• Save dated SMART snapshots with percentage used, host writes, critical warnings, media errors, firmware and temperature.
• Judge the rate of change against actual writing and errors, not an isolated rounded health percentage.