CrystalDiskInfo's verdict is a summary. The evidence is in five raw numbers underneath it. Read those, watch whether they move, and you will know about a failing drive weeks before the status dares to change from Good.
Every SMART tool shows two layers. The top layer is a verdict — Good, Caution, or a percentage — computed by comparing values against thresholds set by the drive manufacturer. The bottom layer is the raw counts themselves: how many sectors the drive has already replaced, how many it cannot currently read, how many transfer errors it has logged. The verdict only changes when a threshold is crossed, and manufacturers set those thresholds to avoid warranty returns, not to give you early warning. The raw layer has no such diplomacy. This page teaches you to read the five rows that matter and — more important — to read them as a trend, because a hidden fault announces itself by moving numbers long before it changes the verdict.
05 Reallocated Sectors Count is the number of damaged spots the drive has already given up on and replaced with spares from its reserve area. C4 Reallocation Event Count records how many times it has had to do this. A healthy drive sits at zero for both, for its whole life. The raw column is what counts — ignore the big "current value" and "worst" scores next to it, those are the manufacturer's grading curve. Three reallocated sectors will not change your health status. Three reallocated sectors that were zero last month means the surface is actively deteriorating.
C5 Current Pending Sector Count counts sectors that failed to read and are waiting for a decision — the drive will retest them when something tries to write there. C6 Offline Uncorrectable counts sectors that failed even the drive's own background self-tests. These two are the earliest honest signals there are: pending sectors appear while the drive still works almost normally, because Windows retries and often succeeds on the second or third attempt. That retrying is exactly the freeze-then-recover behaviour users describe. Any C6 value above zero means data has already been found unreadable by the drive itself.
C7 UDMA CRC Error Count is different: it counts errors in data travelling between the drive and the motherboard — the cable's territory, not the platter's. A rising C7 with flat 05 and C5 values usually means a cable or connector problem, which is cheap and fixable. That is why this row matters so much in hidden-fault diagnosis: it splits the suspects in half. Platter numbers (05, C5, C6) rising means the drive. Interface numbers (C7) rising means the connection. Treating a cable fault as a dying drive wastes money; treating a dying drive as a cable fault wastes data.
Here is the practice that separates guessing from diagnosis:
Numbers that are identical are old history — a scare from months ago that the drive survived. Numbers that grew are a process still running, and a running process of surface damage does not stop on its own. Memory is unreliable here; last month everyone "thinks" the number was about the same. Written values settle it.
Be honest about the blind spots. There is no SMART attribute for a weakening read head, worn bearings or a tiring spindle motor — the parts that produce clicking, shaking and catching. A drive can be mechanically dying with a flawless SMART table. That is precisely why the video pairs the software check with the hold-and-listen test: your hand reads the attributes SMART does not have. Use both. SMART growth with no symptoms means plan a replacement calmly. Symptoms with clean SMART means believe the symptoms and keep investigating — the table is not the whole drive.
Watch the video for the full hidden-fault check — the physical test plus the software readings, on one real drive.
▶️ Watch Full VideoFive raw values cover almost everything. 05 Reallocated Sectors and C4 Reallocation Events show damage already replaced. C5 Current Pending and C6 Offline Uncorrectable show damage in progress. C7 UDMA CRC Errors points at the cable or connector instead of the drive. On a healthy drive all five read zero. Always read the Raw Values column, not the normalised scores beside it.
A small count that never changes is usually survivable history — the drive found a few weak spots and replaced them. A small count that grows between checks is active deterioration, and that is the dangerous case regardless of size. Write the value down, recheck in a week, and let the movement decide. Keep backups current in both cases.
No tool can. SMART shows accumulating damage, and damage trends tell you risk is rising — drives with growing pending and reallocated counts fail at far higher rates — but the same drive may run for a week or a year. Treat growing values as a deadline to replace on your schedule, not as a countdown timer.