My roughly 10-month-old Kingston NV3 2TB (SNV3S2000G, firmware SDU00103) suddenly became unbootable after working normally the previous night. Windows automatic repair and Safe Mode failed, and resetting Windows also failed. CHKDSK repaired multiple corruptions and reported about 184 KB in bad sectors.
The BIOS NVMe tests were unusual: the short test passed, but the extended controller and namespace test failed at LBA 586848 (segment 7). The drive had about 2,040 power-on hours, 10.6 TB written, and was running at approximately 33°C.
I booted from a USB installer, erased the SSD, and installed a fresh copy of Windows 11. The same extended self-test then completed successfully. After reinstalling drivers and applications, I played the same game for around 12 hours without another problem.
SMART still shows 98% health, 100% available spare, 2% percentage used, no critical warning, and no current errors in the NVMe error log. However, the media/data integrity error counter is around 26,000 and the error information log counter is slightly higher. Those counters did not change in the comparison I made after the reinstall, although I did not record a smartctl reading before gaming. The self-test history still records the earlier failed test at LBA 586848.
Could the reinstall have allowed the SSD firmware to remap or retire a defective area, making the later test pass? Should I treat this as evidence of a failing drive and request a warranty replacement, or could the original problem have been caused by Windows or firmware corruption? I have saved screenshots and logs, but I am also concerned that securely erasing the drive before returning it might remove useful diagnostic information.
2 Answers
The counters are important, but they need context. Media/data integrity errors and error-log entries are lifetime counters, so reinstalling Windows will not reset them. A passing test afterward only describes the drive's condition during that later test; it does not erase the earlier failed result.
Before returning it, record the model, firmware, SMART output, self-test history, and the failing LBA. Do not rely on Secure Erase to preserve evidence. If you erase the drive for privacy, keep the diagnostic records separately and follow the seller's return instructions. In any case, back up anything important and avoid using this SSD as the only copy of data, even if it is used mainly for games.
The failed extended self-test and CHKDSK's bad-sector report are enough reason to distrust the drive, even though it currently passes and the health percentage looks good. SSD health percentages mainly describe wear and spare capacity; they do not guarantee that no individual NAND block, controller operation, or past uncorrectable read error has occurred.
A reinstall may have changed the data layout, caused the firmware to retire or remap a problematic area, or simply avoided the files that were previously unreadable. That can make the symptoms disappear without proving the underlying hardware is healthy. The retained self-test history and SMART counters are useful evidence, so save the logs and screenshots and use the warranty process rather than waiting for another failure.
So the safest interpretation is that the drive may have hidden or remapped the bad area, not that the original failure was definitely just Windows corruption. I was hoping the clean self-test meant it had recovered, but I will keep the logs and consider an RMA.