RAID 5 with 4TB drives is reasonable. RAID 5 with 12TB or 16TB drives is a gamble. The problem is not RAID 5's design - it is drive capacity. As drives have grown larger, the time required to rebuild an array after a drive failure has grown with them. During a traditional block-level RAID 5 rebuild, the relevant member range is generally read from the surviving drives. Other implementations differ: for example, an OpenZFS resilver examines data known to be out of date rather than every physical block. On large drives, the probability of encountering an unrecoverable read error (URE) on a surviving drive during that read approaches uncomfortable levels. A URE during a degraded RAID 5 rebuild can make the affected stripe unrecoverable. Whether the rebuild aborts or broader array access is lost depends on the RAID implementation; affected data must be restored from backup if it cannot otherwise be recovered.
In short: RAID 5 is still acceptable with drives up to roughly 4-6TB, particularly if you use NAS-Pro or enterprise drives with higher URE specs. For drives 8TB and above, RAID 6 is the recommended configuration. For drives 12TB and above, RAID 6 should be considered the minimum. The amount read during a full rebuild generally grows with drive capacity, but actual URE risk also depends on the exact drive specification, array layout, utilisation and RAID implementation.
The URE Problem Explained
Every hard drive has a specification called the Unrecoverable Read Error (URE) rate - also written as UBER (Unrecoverable Bit Error Rate). This is the rate at which the drive is expected to encounter a bit that cannot be read, even after retries. URE specifications vary by manufacturer, model and capacity. For example, WD Red Plus is specified at less than 1 in 10^14, while current Seagate IronWolf 6TB, 8TB and 10TB models are specified at 1 in 10^15. Check the datasheet for the exact drive being considered.
A 10^14 specification corresponds arithmetically to one error per 12.5TB, but manufacturer datasheets state this as a maximum error rate; it should not be read as a prediction that a drive will encounter an error every 12.5TB. A pro-tier or enterprise drive extends this to approximately 125TB.
A conventional full block-level RAID 5 reconstruction reads the corresponding member range from every surviving drive. Other implementations differ; OpenZFS resilvering, for example, examines data known to be out of date. The total data read from the surviving drives depends on the drive size and array configuration. In a 3-drive RAID 5 array with 12TB drives, rebuilding one failed drive requires reading approximately 24TB from the two surviving drives. If 10^-14 is treated as an exact, independent per-bit probability, that model yields approximately 85% over 24TB. Manufacturer datasheets instead publish a maximum non-recoverable-error specification, so 85% is not an empirically established rebuild-failure rate. If a URE affects a stripe that is already missing a RAID 5 member, that stripe cannot be reconstructed from parity. The wider result depends on the implementation and may range from loss of affected blocks or files to an aborted or inaccessible array.
How Risk Scales With Drive Size
The URE risk during a RAID 5 rebuild scales directly with the total data read from surviving drives. Larger drives mean more data to read, which means a higher probability of hitting a URE before the rebuild completes.
Illustrative RAID 5 rebuild URE model by drive size (3-drive array). Percentages assume an exact, independent 10^-14 per-bit probability; manufacturer URE figures are maximum specifications, so these are not measured rebuild-failure probabilities.
| Drive size | Data read during rebuild | Approx. URE probability | Illustrative rebuild time; actual duration varies with NAS hardware, RAID implementation, drive performance, array utilisation and concurrent workload | |
|---|---|---|---|---|
| 4TB drives | 4TB | ~8TB | ~47% | ~18-24 hours |
| 6TB drives | 6TB | ~12TB | ~62% | ~28-36 hours |
| 8TB drives | 8TB | ~16TB | ~72% | ~36-48 hours |
| 12TB drives | 12TB | ~24TB | ~85% | ~55-72 hours |
| 16TB drives | 16TB | ~32TB | ~92% | ~72-96 hours |
| 20TB drives | 20TB | ~40TB | ~96% | ~90-120 hours |
These probabilities assume consumer NAS drives at the 10^14 UBER spec. The numbers look alarming but come with important context: a URE during rebuild does not automatically mean data loss in all cases. Modern RAID controllers and NAS operating systems can sometimes handle read errors gracefully, particularly for short error bursts. ZFS block checksums improve detection of silent corruption, but recovery from a device-reported URE still depends on available redundancy. RAIDZ1 does not gain an additional level of failure tolerance from checksums. And drives with better UBER specs (IronWolf Pro, WD Red Pro at 10^15) substantially reduce the URE probability.
But the directional finding holds: as drive capacity grows, the risk profile of RAID 5 degrades. The question is not whether RAID 5 is always unsafe - it is whether the risk level is acceptable for your data.
Rebuild Time: The Compounding Factor
The rebuild time problem compounds the URE risk. A RAID 5 array rebuilding a 4TB drive might take 18-24 hours. A 20TB drive rebuild at the speeds typical of a home or small business NAS might take 4-5 days. During that entire period, the array is running in a degraded state with no drive redundancy. A second drive failure during this window means total data loss.
Real NAS rebuild speeds are slower than theoretical maximum sequential throughput. Concurrent user I/O can reduce rebuild throughput, but actual speed depends on the NAS hardware, drives, RAID implementation, rebuild settings and workload. Rebuilding 20TB of data at 60MB/s takes approximately 93 hours. Nearly four days with zero drive redundancy and, under the article's simplifying independent 10^-14 error model, a calculated probability approaching 96%. Manufacturer URE figures are maximum specifications, so this is not a statistical certainty.
Practical implication: the argument for moving to RAID 6 on large drives is not just about URE probability during any single rebuild event. It is about the combination of URE probability and the extended rebuild window that large drives create.
RAID 6: What Changes
RAID 6 uses two parity drives instead of one, allowing the array to tolerate two simultaneous drive failures without data loss. More importantly for the large-drive URE scenario: RAID 6 can survive one URE during a rebuild. If the reconstruction process encounters a single unrecoverable read error on a surviving drive, the second parity set provides the redundancy needed to complete the rebuild despite the error.
RAID 6 is not free. The additional parity drive costs one drive's worth of capacity compared to RAID 5. In a 4-drive array: RAID 5 gives you 3 drives of usable space, RAID 6 gives you 2. Write performance is also lower due to double parity calculations. RAID 6 generally has additional write overhead relative to RAID 5, but the performance difference depends on the controller, caching, drive count, workload and write pattern.
Synology's SHR-2 is the equivalent of RAID 6 for mixed-size drives: it uses two parity drives and tolerates two simultaneous failures. For NAS devices with Synology DSM, SHR-2 is the appropriate selection when the drive failure risk profile argues for moving beyond single-parity protection.
How ZFS Changes the Equation
ZFS-based NAS systems (TrueNAS, QNAP QuTS Hero) handle the URE problem differently from traditional RAID implementations. ZFS stores a checksum for every block of data written to the pool. During a rebuild (called resilvering in ZFS terminology), each block is verified against its stored checksum. A block that fails its checksum can be detected and, if the parity allows for it, corrected or flagged.
This block-level integrity checking does not eliminate the URE risk but changes the failure mode. A URE is a device-reported read failure, not silent corruption. ZFS checksums additionally detect returned data that does not match its stored checksum; repair still requires sufficient redundancy. In RAIDZ1 (the ZFS equivalent of RAID 5), a detected bad block during resilvering can still cause failure if no correctable copy exists. In RAIDZ2 (equivalent of RAID 6), one bad block can be reconstructed from the second parity set.
ZFS also has regular data scrubbing built in. Scheduling a weekly or monthly scrub reads every block on the pool and verifies checksums, detecting and, where sufficient redundancy exists, correcting silent corruption. Btrfs also checksums data and supports scrub and repair; on Synology DSM, silent-corruption detection requires the shared folder's data-checksum option to be enabled. This scrubbing process is one of the strongest arguments for using a ZFS-based NAS OS (TrueNAS, QuTS Hero) when data integrity is a priority.
The Practical Decision: When to Use RAID 5 and When to Use RAID 6
The following guidance is based on drive size, drive tier, and the file system in use:
| Up to 4TB drives, consumer NAS (ext4/Btrfs) | RAID 5 is acceptable. URE risk during rebuild is real but the rebuild window is short enough to be manageable. Keep backups current. |
|---|---|
| Up to 4TB drives, ZFS (TrueNAS/QuTS Hero) | RAIDZ1 (RAID 5 equivalent) is acceptable. ZFS error handling reduces practical risk. |
| 4-8TB drives, consumer NAS (ext4/Btrfs) | RAID 5 is workable but RAID 6 is recommended if the NAS holds important data. Rebuild times are 24-48 hours with meaningful URE exposure. |
| 4-8TB drives, pro-tier (IronWolf Pro, WD Red Pro at 10^15 UBER) | RAID 5 is acceptable. Pro drives reduce URE probability by 10x compared to consumer drives. |
| 8TB+ drives, consumer NAS | RAID 6 / SHR-2 recommended. URE probability during rebuild is too high on standard NAS drives to be comfortable. |
| 8TB+ drives, pro-tier or enterprise, with ZFS | RAIDZ1 is defensible with pro drives and ZFS scrubbing. RAIDZ2 is better practice and removes the question entirely. |
| 12TB+ drives, any configuration | RAID 6 / RAIDZ2 minimum. The rebuild window and URE exposure at this capacity make single-parity impractical regardless of drive tier. |
The one rule that applies regardless of RAID level: RAID is not a backup. A RAID 6 array with current backups is the correct setup. A RAID 5 array with no backup is always inadequate, regardless of drive size. See the RAID is not a backup guide for the full explanation.
Australian Buyers: What You Need to Know
Drive pricing in 2026 and the RAID 6 capacity cost. The capacity penalty of RAID 6 over RAID 5 is real: in a 4-drive array, you lose one full drive to the second parity set. With NAS-grade drives priced above $200 AUD for 4TB and considerably more for larger capacities, the effective cost of the additional RAID 6 parity is now a meaningful dollar amount. Factor this into the drive purchase decision rather than treating RAID level as a configuration choice independent of hardware cost.
Pro-tier drives worth considering in AU. The Seagate IronWolf Pro and WD Red Pro are available through Scorptec, Mwave, and PLE at a premium over consumer NAS drives. For large-capacity RAID 5 deployments where the operator is resistant to moving to RAID 6, using pro-tier drives with 10^15 UBER rates substantially improves the URE probability profile. This is not a substitute for RAID 6 at 12TB and above, but it meaningfully changes the risk at 6-8TB.
Synology DSM drive compatibility note. Under DSM 7.3, 2025-and-later DS Plus models permit third-party HDDs that are not on the incompatibility list for new storage-pool creation. Synology still recommends drives on the compatibility list, so verify the exact NAS and drive model before purchase. M.2 NVMe drives for cache and storage still require drives on Synology's official compatibility list. Enterprise and rackmount Synology models maintain stricter compatibility requirements. Verify your specific model's compatibility list at the Synology website before purchasing drives for a new pool.
QNAP QuTS Hero for large drives. QuTS hero is available only on supported QNAP NAS models and requires at least 8GB RAM. Some models also require a specific firmware version or hardware configuration, and dual-controller models are not supported. For buyers running 8TB and above drives and wanting the strongest practical protection against URE events during rebuild, QuTS Hero with RAIDZ2 configuration provides better data integrity guarantees than QTS (ext4) with RAID 6. The RAM requirement and setup complexity are higher, but the data protection improvement is meaningful at these drive capacities.
Related reading: our NAS buyer's guide.
Use our free RAID Calculator to work out usable capacity across RAID levels.
See also: our complete Synology NAS Australia guide.
Rebuild risk is largely a function of the drives themselves. Our NAS hard drive guide covers which drives are rated for it and why capacity changes the calculation.
Is RAID 5 dead?
No. RAID 5 remains appropriate for smaller drive sizes and pro-tier drives with better UBER specs. The popular claim that RAID 5 is dead came from the transition to multi-terabyte drives in enterprise environments during the early 2010s. For a home NAS with 4TB drives and regular backups, RAID 5 is still a functional and common configuration. The risk scaling at 8TB and above is real and worth taking seriously, but it does not make RAID 5 categorically unsuitable at all drive sizes.
What is RAID 6 in Synology and QNAP terminology?
Synology calls their dual-parity configuration SHR-2 (for mixed drive sizes) or RAID 6 (for matching drives). QNAP calls it RAID 6. TrueNAS uses RAIDZ2 terminology. Unraid can use RAIDZ2 in a ZFS pool, while its main array uses dedicated single or dual parity. All refer to the same underlying concept: two parity blocks per stripe, allowing the array to survive two simultaneous drive failures. The specific implementation differs but the data protection outcome is the same.
How can I reduce RAID 5 rebuild risk if I cannot switch to RAID 6?
Several steps reduce risk. First, use pro-tier drives (IronWolf Pro, WD Red Pro) with 10^15 UBER rates rather than consumer drives at 10^14. Second, reduce rebuild time by replacing a failed drive as quickly as possible and minimising other I/O activity on the NAS during the rebuild. Third, ensure current backups exist before any rebuild begins. Fourth, run scheduled SMART tests and RAID scrubs to detect marginal drives before they fail outright rather than discovering them during a rebuild.
Does RAID 6 require more drives than RAID 5?
RAID 6 requires a minimum of four drives (versus three for RAID 5). In a 4-drive array, RAID 5 gives you three drives of usable space while RAID 6 gives you two. The capacity cost is one full drive. In larger arrays (6-bay, 8-bay), the capacity penalty of RAID 6 becomes proportionally smaller. An 8-drive RAID 6 array gives you six drives of usable space (75% efficiency), which is close to RAID 5's seven drives (87.5% efficiency) while providing substantially better rebuild resilience.
Does Synology SHR automatically choose RAID 5 or RAID 6?
SHR (Synology Hybrid RAID) defaults to single parity (equivalent to RAID 5) and is labelled SHR. SHR-2 uses dual parity (equivalent to RAID 6) and must be selected explicitly when creating the storage pool. On a 4-bay Synology NAS with four drives, selecting SHR-2 provides dual-parity protection and tolerates two simultaneous drive failures. SHR-2 requires a minimum of four drives. For 8TB and above drives on a 4-bay Synology, SHR-2 is the recommended configuration.
What happens to data if RAID 5 rebuild fails due to a URE?
It depends on the file system and NAS platform. On QNAP QTS systems, which use ext4, and on Synology systems configured with ext4, A URE affecting a stripe that is already missing a RAID 5 member leaves that stripe unreconstructable. Depending on the RAID implementation, this may produce affected-block errors, abort the rebuild, fail another member or make the array inaccessible. If no usable backup exists, specialist RAID recovery may be required. Australian pricing varies widely with disk count, failure mode and damage, so obtain a case-specific quote; successful recovery is not guaranteed. On ZFS-based systems, the error is detected and handled more explicitly, but recovery still depends on whether the second parity set has sufficient information to reconstruct the affected blocks.
Understanding RAID risk is the starting point. The RAID guide covers all RAID levels, capacity calculations, and how to choose the right configuration for your NAS and drive combination.