SMART Attributes That Predict Failure

Questions people actually ask

The drive says PASSED. Is it fine?

Not necessarily, and this is the most important thing to know about SMART. The self-assessment only trips when the manufacturer's own thresholds are crossed, and those thresholds are set so a drive stays inside warranty — not so your data stays intact. A drive with pending sectors and uncorrectable errors routinely reports PASSED. Read the attributes, not the verdict.

What is the difference between pending and reallocated sectors?

A reallocated sector is damage the drive has already worked around by moving the data to a spare. A pending sector is a block the drive has found unreadable and has not yet managed to relocate — meaning data currently at risk. Reallocated is history; pending is present tense. A small, stable reallocated count is survivable; any pending count is a reason to image the drive now.

I have UDMA CRC errors. Do I replace the drive?

Almost certainly not. CRC errors are interface errors — the data was corrupted between the drive and the controller, not on the platter. That points at the cable, the backplane connector, or the port far more often than the disk. Reseat and replace the SATA cable first, then try another port. Replacing the drive on CRC errors alone is a wasted part and a repeat visit.

Which attributes actually matter?

Current Pending Sector and Offline Uncorrectable first, because they represent data at risk right now. Then Reported Uncorrect and Reallocated Sector Count for accumulated damage, and Spin Retry Count for mechanical failure, which is late-stage. Power-on hours matter only as context — a clean drive at five years is a planned-replacement candidate. Schedule it; there is no need to move tonight.

What does SMART report on an NVMe drive?

A different set entirely — there is no attribute table. The ones to read are Available Spare, which is the pool the controller uses to retire worn cells and which triggers read-only mode when exhausted; Percentage Used, which is consumed write endurance and can legitimately exceed 100; Media and Data Integrity Errors, the equivalent of uncorrectable sectors; and Critical Warning, which is a bitfield rather than a number.

Does a rising count matter more than the absolute number?

Yes, and it is the measurement most people skip. Five reallocated sectors that have not moved in a year is a drive with a scar. Five that were two last week is a drive failing in front of you. Record the number, re-read in a week, and treat the trend as the signal.

Does high temperature actually shorten drive life?

Sustained operation above about 60 °C measurably does, and above 70 °C most drives begin thermal throttling. The important part is that this is usually an airflow problem rather than a drive problem — a disk in a dead-air pocket will fail early no matter what you replace it with. Check filters, fans and cable obstruction before blaming the hardware.