Windows users should also be careful with a popular but misleading premise: a built-in health indication does not mean Windows is continuously forecasting the condition of every hard drive and SSD. The native Drive Health coverage discussed here applies to NVMe SSDs, not SATA SSDs or mechanical hard drives. That distinction affects what Windows can tell you, what a third-party SMART report means, and how much confidence to place in any single value.
A Windows storage alert is a risk signal, not a diagnosis
For an NVMe SSD, the underlying health information comes from the drive's NVMe SMART/Health log. The device returns this telemetry through its Get Log Page command, and it can include a critical-warning state, available spare capacity, percentage used, power-on hours, media errors, and an error-log count.
Three NVMe critical-warning categories are particularly important:
- Low available spare capacity. The SSD's reserve space is running low.
- Reliability degradation. The device has reported significant media-related errors or an internal error that degrades the reliability of the NVM subsystem.
- Read-only media. The drive has reached a state in which writing is no longer available.
None of those conditions is a reason to wait for a more convenient time to protect data. Back up first. A successful copy does not prove the drive is healthy, and a warning does not establish the exact component or mechanism at fault, but the combination makes data preservation the sensible first action.
On systems that expose the native NVMe Drive Health area, the relevant Settings route is:
Settings > System > Storage > Advanced storage settings > Disks & volumes > Properties
Treat that screen as one source of evidence rather than a full forensic report. It can help identify an NVMe condition worthy of attention, but it cannot turn health telemetry into a precise remaining-lifetime estimate.
Windows does not make SATA and HDD health universal
A major practical limit is interface coverage. Windows' built-in Drive Health monitoring should not be read as a universal health service for every storage device connected to a PC. Its documented focus is NVMe SSDs. A SATA SSD or conventional hard disk may therefore have no equivalent native warning even if its own SMART data contains relevant indicators.
That has two consequences.
First, the absence of a Windows health notification is not a clean bill of health for a SATA SSD or an HDD. Important failures can occur without the particular Windows feature being applicable to the device at all.
Second, a screenshot from a disk-monitoring utility must be interpreted according to the interface and manufacturer. An NVMe value and a SATA SMART attribute may sound similar—both may use words such as “wear,” “errors,” or “health”—but they do not necessarily share definitions, scales, thresholds, or implications.
Before interpreting a warning, establish the basics: whether the affected device is NVMe, SATA, or behind USB; which physical disk holds important data; and whether the drive is internal or housed in an external enclosure. Those details determine which telemetry should be available and how much weight to put on it.
NVMe Percentage Used is not a countdown clock
One of the easiest NVMe fields to misunderstand is Percentage Used. It is a vendor-specific estimate of life consumed, based on the drive's actual use and the manufacturer's predicted NVM endurance. In other words, it is an endurance model supplied by the drive maker, not a universal measure of immediate failure probability.
A reading of 100 has a specific but limited meaning: the manufacturer-estimated endurance has been consumed. It does not necessarily mean the SSD has already failed. The field can even exceed 100.
That does not make the value unimportant. A high or rising percentage-used reading is useful context when planning replacement, especially when combined with critical warnings, media errors, or reliability problems. But it should not be converted into an unsupported statement such as “the drive has exactly 26% life left,” nor should a reading below 100 be used to dismiss all other warning signs.
The right question is not “Has this number reached a magical failure threshold?” It is “What does this manufacturer-specific estimate say alongside the device's critical state, error history, and the value of the data stored on it?”
Why SATA SMART labels demand more caution
SATA SMART information has a different interpretive problem. Individual SMART records contain an ID, status information, a normalized value, and vendor-specific bytes. Those vendor-specific bytes are not defined uniformly by the specifications. As a result, raw SATA SMART attributes and labels cannot be assigned universal meanings or replacement thresholds without the drive maker's documentation for that model family.
This is why broadly confident interpretations of labels such as Media Wearout Indicator, Grown Bad Blocks, or other model-specific counters are risky. A value that appears to be a percentage might be a normalized score on one SSD and something else on another. A raw count may be encoded, scaled, or thresholded differently across vendors and generations.
Power-on hours have a similar limitation. They provide useful context about accumulated operation, but hours alone do not establish an SSD's remaining endurance or prove an imminent failure. The workload, drive design, firmware behavior, temperatures, write volume, and the manufacturer's own attribute definitions all matter.
For SATA drives, the sound approach is to preserve the complete report, note whether any values are changing over time, and consult the exact drive maker's documentation before treating an attribute as a definitive health percentage. A single static reading without the model, firmware, raw data, and vendor definition is usually insufficient for a precise conclusion.
CRC errors point to the link, not automatically the flash
A SATA SMART CRC Error Count is often treated as an instant verdict on the SSD itself. That is too simple. Samsung's description of SMART ID 199 says the CRC engine tallies errors when there is a problem between the host and the SSD. This supports an important distinction: the count can reflect a problem in the path between PC and drive rather than NAND-media failure alone.
The word “path” matters. A nonzero CRC count does not identify the faulty item. It cannot, by itself, prove that a SATA cable is loose, that a cable is defective, that a motherboard port is at fault, or that the SSD is innocent. It identifies a host-to-drive communication issue worth investigating.
For an internal SATA device, reseating connections and checking the broader host-to-drive link are reasonable diagnostic steps after backups are safe. What matters next is evidence: does the count remain stable, or does it increase after the connection has been checked? A stable historical count has a different meaning from a counter that continues to climb. Neither observation alone substitutes for protecting important files.
External USB drives can hide health data
An external drive presents another complication: the USB-to-storage bridge between Windows and the actual SSD or hard disk. SMART access depends on whether that bridge and the software being used can pass through the necessary commands. Some bridge configurations do not support the SMART-status command.
Accordingly, a blank, partial, or unavailable SMART report from an external enclosure is ambiguous. It may indicate a bridge pass-through limitation, unsupported telemetry, operating-system behavior, or tool compatibility. It does not, on its own, establish that the disk has no health data or that the disk is malfunctioning.
This is especially relevant when comparing an internal drive's report with an external version of the same drive. Different enclosure hardware can change the visibility of diagnostic information without changing the condition of the storage media itself.
SMART is valuable, but it cannot promise advance notice
SMART and NVMe health data are useful because certain error patterns correlate with elevated failure risk across large drive populations. They can identify conditions that deserve prompt action. Yet correlation is not an individual forecast.
A large field study of disk drives found that more than 56% of failed drives had none of four strong SMART signals examined. More than 36% had no counts across the other SMART variables examined as well. The study's central practical conclusion remains important: SMART data alone is unlikely to predict failure effectively for each individual drive.
That finding cuts both ways. A clean-looking report cannot guarantee safety. A concerning attribute can justify caution without revealing an exact date of failure. Disk health tools are best understood as risk indicators and diagnostic inputs, not warranties or crystal balls.
The most effective protection is operational rather than predictive: maintain backups that are separate from the drive being monitored, verify that important files can be recovered, and avoid relying on a drive-health screen as the only safeguard against loss.
A measured response plan
When Windows reports a critical storage condition, or a system displays a SMART error, use this order of operations:
- Back up important data immediately. Prioritize irreplaceable documents, photos, project files, and any data not already recoverable elsewhere.
- Identify the actual device and interface. Confirm whether the affected disk is NVMe, SATA, or external USB storage. Do not assume Windows' NVMe-oriented Drive Health information covers every disk.
- Record the available evidence. Keep the warning text and complete health readings. For SATA, the exact model and firmware are essential to interpreting vendor-specific attributes.
- Look for critical states and changes, not just one number. NVMe critical warnings, media errors, and a growing error-log count are more meaningful in context than an isolated percentage or label.
- Investigate connection-related counters appropriately. A SATA CRC count calls for attention to the host-to-drive link, but it does not identify a specific failed cable or component.
- Plan replacement around risk and data value. A critical warning, read-only state, or reliability degradation shifts the balance strongly toward replacing the drive after data is secured. A noncritical vendor-specific reading may justify monitoring and documentation, but should not be overstated.
The key lesson is simple: storage telemetry can give useful warning, particularly for supported NVMe devices, but it has boundaries. Windows alerts and SMART values are reasons to protect data and investigate carefully—not proof that every drive failure is foreseeable, and not a substitute for backups that already exist before the warning appears.