What SMART Is Really Telling You

The handful of attributes that matter, and the many that do not.

01The dashboard that looks more useful than it is

Self-Monitoring, Analysis and Reporting Technology — SMART — was built into virtually every desktop hard drive from the late 1990s onward, and every consumer SSD carries a version of it too. Open any monitoring tool and you will see rows of attributes: reallocated sector count, spin-up time, seek error rate, temperature, power-on hours, command timeout, and a dozen more. The numbers feel authoritative. Most of them are not telling you very much at all.

SMART was designed for drive manufacturers and fleet operators, not for individuals diagnosing a sick disk in a hurry. The attribute list was never fully standardised — manufacturers assign different meanings to the same attribute ID, scale their raw values differently, and keep some of the most useful diagnostics proprietary. What survived the standardisation gap is a set of values you can read but sometimes cannot compare across brands or even across product lines from the same brand. The overall PASS/FAIL verdict the firmware posts is computed against thresholds the manufacturer chooses, optimised to avoid warranty false positives, not to give you early warning of impending failure. A drive can pass all its SMART thresholds on a Monday and be unreadable by Friday.

None of that means SMART is useless. It means you have to know which handful of attributes carry real signal, ignore most of the rest, and treat the overall verdict as a floor, not a ceiling.

The attributes worth reading, in plain language

Attribute	What it means	Does it matter	What to do
Reallocated Sectors Count	Sectors the drive found bad and swapped for spares.	Yes — any non-zero value is a drive telling you it has started dying.	Back up now. A rising count means replace, not monitor.
Current Pending Sector Count	Sectors that failed a read and are waiting to be reallocated.	Yes — arguably more than reallocated.	Image the drive before you do anything that writes to it.
Reported Uncorrectable Errors	Reads the drive could not fix with its own error correction.	Yes.	Treat as data loss in progress. Restore from backup rather than repair.
Spin Retry Count	Times the platters failed to reach speed on the first attempt.	Yes, on a mechanical drive — it points at the motor or bearings.	Power it down and image it on the next successful spin-up.
Power-On Hours	How long the drive has been running.	On its own, no.	Use it as context for the counters that do matter.
Wear Levelling Count / Percentage Used	How much of the SSD's rated write endurance has been consumed.	Yes, but predictably — this is a planned end, not a fault.	Plan the replacement. Expect read-only behaviour near the end.
Temperature	Drive temperature.	Only at extremes, sustained.	Fix airflow. Do not panic at a warm drive under load.

02The five attributes worth watching

Reallocated Sectors Count (ID 05). This is the single most important number on a spinning drive. When the disk's firmware finds a sector it can no longer reliably read or write, it remaps the data to a reserved pool of spare sectors and increments this counter. A small number of reallocations on a drive that has run for years is not automatically fatal — but any non-zero value means the surface has failed in at least one place, and the question is whether it has failed in one place or in the opening stages of a wider pattern. The raw value is what matters. If it was zero yesterday and is not zero today, stop writing to that drive and image it immediately.

Current Pending Sectors (ID C5). These are sectors the drive has flagged as unstable — it tried to read them and couldn't, but hasn't yet decided whether to reallocate or consider them recoverable. Pending sectors are often the leading indicator that appears before reallocated sectors climb. A non-zero count is urgent. Some pending sectors resolve themselves if a successful read or write operation recovers the data; many do not. Do not wait and hope. A drive with pending sectors needs to be imaged before anything else happens to it.

Uncorrectable Sector Count (ID C6). Once the drive has exhausted its error-correction and cannot read a sector at all — not even into the pending pool — it records the event here. Any non-zero raw value means you already have data loss at the sector level. Whether that translates to lost files depends on which sectors are affected and what the file system was storing there; you may or may not notice it at the application layer. But the counter tells you the drive has already crossed a line.

Reallocated Event Count (ID C4). Where ID 05 counts the total reallocated sectors, this attribute counts the number of reallocation events, including attempts that were triggered but did not succeed. On most drives the two values track closely together; a large divergence — many events, few successful reallocations — suggests the spare pool is running low or the firmware is struggling. Worth watching alongside 05, not instead of it.

Power-On Hours (ID 09) and Start/Stop Cycles (ID 0C). Neither of these predicts failure in isolation, but both supply context that makes everything else interpretable. A drive with reallocated sectors and ten thousand power-on hours reads very differently from one with the same reallocated count but fewer than a hundred hours. High start/stop cycles on a laptop drive that was docked and undocked daily are expected; identical numbers on a desktop machine that was supposed to spin continuously suggest the drive was losing power or entering an unintended sleep state, which is worth investigating on its own terms.

Raw values are vendor-specific and some manufacturers pack two numbers into one field. Read the normalised value against its threshold, and judge a drive on the direction a counter is moving.

03Temperature, seek errors, and the noise in the data

Temperature (ID BE, or on some drives ID C2) does matter, but not in the way most people expect. The headline danger is sustained operation above roughly 50–55 °C, which accelerates bearing wear on spinning drives and accelerates charge leakage in flash cells. What the raw value mostly captures, though, is operating environment: a drive in a poorly ventilated case will run hot before it shows any other sign of stress. Temperature trending upward over weeks — without any change to the workload — can indicate a failing fan or a heat sink that has lost contact with the controller, and that is worth acting on before other attributes deteriorate.

Seek Error Rate (ID 07) and Raw Read Error Rate (ID 01) are two attributes that worry people and usually shouldn't. Western Digital and Seagate in particular store these as composite values with a large raw number in the lower bytes and an error count in the upper bytes — the raw figure looks enormous and alarming, but the normalised value (shown in the 'current' column rather than the raw column) is what maps to actual failure risk, and on a healthy drive it remains high. Comparing the raw seek error rate across brands is essentially meaningless. If the normalised value drops toward the threshold the manufacturer has set, that matters; a large raw number on its own does not.

Spin Retry Count (ID 0A) is worth a glance on spinning drives. More than a handful of retries suggests the drive is struggling to reach operating speed — that can mean a weakening motor, a bearing problem, or a power supply that cannot deliver the current the drive needs at startup. Combined with clicking or grinding on spin-up, it points directly toward a mechanical problem that is not going to improve on its own.

But the counter tells you the drive has already crossed a line.

04The overall verdict is a lagging indicator

The single PASS or FAIL the firmware posts is useful as a hard floor: a FAIL almost always means imminent or active failure, and you should treat it as an emergency. But a PASS means only that no threshold has been crossed. Pending sectors at 3, reallocated sectors at 12, and an overall verdict of PASS is a perfectly plausible and genuinely dangerous combination. Read the individual attributes; do not let the headline verdict substitute for them.

SMART was never designed to give you certainty. It was designed to give you early signal — and for the five attributes above, it does that reasonably well. Check them regularly, set an alert if any of them leave zero, and treat the first non-zero reading not as a warning to file away but as the starting gun for a backup and imaging run.