“RAID 5 is dead” is a headline about a real failure mode: an unrecoverable read error during a rebuild. It is not dead everywhere — but on large drives the maths is unkind. See the risk for your own array in the RAID 5 calculator.
What a URE is
An unrecoverable read error (URE) is a sector the drive cannot read back, even after retries and ECC. Manufacturers quote a rate: roughly one URE per 10¹⁴ bits read for consumer drives, 10¹⁵ for enterprise/nearline, and 10¹⁶–10¹⁷ for SSDs. In a healthy array a URE is repaired from parity; the problem is when there is no parity left.
During a RAID 5 rebuild there is no redundancy — the array is already down one drive and is reading all the survivors to reconstruct the replacement. A URE on any surviving drive during that read cannot be repaired, and the rebuild fails for that stripe.
Why drive size makes it worse
The probability of hitting a URE rises with the number of bits you must read, which rises with drive capacity. Rebuilding a RAID 5 of five 4 TB consumer drives reads ~16 TB; at a 10¹⁴ rate that is a high chance of at least one URE. The calculator computes this from the standard 1 − (1 − UER)^bits model.
Enterprise drives (10¹⁵) and SSDs (10¹⁶+) shift the odds dramatically — which is why URE rate is a selectable input in our calculator. Understating it would understate the risk, so we default conservatively.
The fix: dual parity
RAID 6, RAID-DP and RAIDZ2 keep a second parity, so a URE during a single-drive rebuild is reconstructed rather than fatal — the rebuild continues. This is the real reason dual parity replaced single parity for large-capacity arrays, as covered in RAID 5 vs RAID 6.
So RAID 5 is not universally dead: on small arrays of enterprise drives or SSDs, with a good backup, it can still be reasonable. But for large nearline HDD pools, RAID 6 (or RAID 60) is the safe default.