Unreadable MFT clusters on NVMe but Windows hides SMART health data

I ran read-only diagnostics on a failing ADATA LEGEND 710 NVMe drive. chkdsk found four consecutive unreadable MFT file-record segments (0x71FD88–0x71FD8B), which is one bad 4 KB cluster inside the MFT metadata. Reading that cluster during the check generated 133 fresh bad-block events in the disk log.

Windows Cannot Expose Full NVMe SMART Data

I ran Get-StorageReliabilityCounter and MSFT_PhysicalDisk under admin. Windows does not expose NVMe SMART log page 0x02 on this device even when elevated. Available Spare and Media-and-Data-Integrity-Errors are unobtainable. The firmware only populates Temperature (49C), Wear (0%), and latency maxes.

Without the SMART trend data, I cannot prove whether the bad-sector set is fixed or still growing. The drive appears mostly healthy and unworn aside from the bad sectors.

Rescue Copy

I copied 105,310 files (12.37 GB) of irreplaceable data to Google Drive, then copied 1,700 files (228 MB) of loose project root files and config. Zero genuine media read failures during either copy.

Google Drive is a virtual mount with sync latency. I added a 60-second pause before verify after the first run reported 508 false missing-file results that all resolved on recheck.

I skipped three raw credential files and one 4.62 GB log file that exceeded the FAT32 4 GB file limit on Google Drive.

Traversal Performance

Get-ChildItem -Recurse then filtering afterward was catastrophically slow on node_modules and model trees. I switched to a manual pruning walker that excluded directories during traversal.