Every diagnosis in this series ends at the same desk: a decision. Replace the drive now while it still works, or keep it under watch? Here are the signs that settle the question — and the backup habit that should change either way.
The video closes on the question every viewer actually has: when to replace the hard drive. By now you may hold several kinds of evidence — a health tool still saying 100 percent, some raw SMART values, a filtered event timeline, a benchmark graph with craters, a clone with two skipped areas, a drive that shudders slightly in the hand. None of these alone is a verdict, and waiting for the tool's headline to turn red means waiting until the drive is nearly dead. This page converts the evidence into a decision with two honest outcomes — retire now or keep under monitoring — and then addresses the deeper issue a suspect drive always exposes: it was holding the only copy of something.
Replace the drive without waiting for more proof if you have any of the following:
Notice what is not on the list: the health percentage. It is the last thing that will change, and a replacement decided by evidence today happens on your schedule, with your data intact. The same replacement forced next month happens on the drive's.
Monitoring is a legitimate verdict, but it must be earned with flat evidence, not granted by optimism. The profile that qualifies: the only movement was C7 (the cable counter) and it froze after a cable swap; all platter values — 05, C5, C6 — have stayed at zero or frozen for a month of normal use; no new disk entries in the event log; no symptoms the user can feel. A drive in this state is demoted, not trusted: use it for copies, downloads, game libraries — things that exist elsewhere or can be fetched again. Recheck the five SMART rows monthly, on a calendar, because the entire arrangement depends on noticing the week the numbers start moving again.
A suspect drive is never only a drive problem — it is proof your data had a single point of failure. Fix that while the lesson is fresh, in plain terms: three copies of anything important, on two different kinds of storage, with one copy outside the PC (an external drive you unplug, or cloud storage). The suspect drive, or its replacement, may hold copy one. It must never hold the only copy. Test the arrangement once: open real files from the backup, because an untested backup is a rumour. Most data loss stories do not start with a failed drive — they start with a drive that warned for a month while its owner meant to get around to backing up.
If the verdict is retire, the replacement choice deserves the same honesty as the diagnosis. A new mechanical drive is the like-for-like option: cheap per terabyte, and subject to the same mechanical failure family you just met — keep the monitoring habit. An SSD removes that entire family — no heads, no platters, no clicking — but fails its own way, by wear and electronics, and is checked by a different health reading (a wear percentage, not sector drama). Either is a good answer; the bad answer is a second-hand drive of unknown history in the role of sole data keeper. Whatever you install, record its SMART baseline on day one — the written starting values future-you will need for exactly this comparison.
A retired drive still holds your data, including the parts you deleted — deletion is not erasure. Before the drive leaves your hands — sold, recycled, handed down, binned — deal with that: a full overwrite pass for a working drive, or physical destruction of the platters for one that no longer spins. A drive that failed with your documents on it and ends up in a general e-waste stream is a privacy incident on a delay. Keep the dead drive's details (model, serial, its final SMART readings) noted with your records if it is under warranty — manufacturers' warranty checks often ask, and a drive showing documented growing damage within its warranty period is a straightforward claim.
Watch the video — when to replace, the recovery warnings, and the complete hidden-fault check in one place.
▶️ Watch Full VideoIf the symptoms are mechanical — clicking, grinding, a catching shudder in the hand — replace it. There is no SMART counter for head and bearing wear, so a clean table cannot overrule what the drive is physically doing. Back up or clone immediately, because a mechanically distressed drive owes you no warning before it stops. If the symptoms are only slowness with clean SMART and a clean physical test, keep investigating other causes while you monitor weekly.
Monthly is the practical rhythm — often enough to catch growth early, light enough that you will actually keep it up. Read the same five raw values (05, C4, C5, C6, C7), compare against your written baseline, and check the filtered event log for new disk entries. The arrangement only works if the numbers are written down; a monitoring habit that relies on remembering last month's values is not monitoring.
For a system drive, usually yes: it removes the mechanical failure family entirely and transforms boot and loading times. But it is not immortal — SSDs wear with writes and can fail electronically, and their health is read as a wear percentage rather than sector counts. For bulk storage where cost per terabyte decides, a new mechanical drive with a monitoring habit and proper backups remains a perfectly honest choice.