RAID 5 vs RAID 6: Capacity, Rebuild Risk and Which to Choose
The RAID 5 vs RAID 6 decision comes down to one question: how many drives can you afford to lose at once? RAID 5 survives one failure and gives you more usable space. RAID 6 survives two and costs you an extra drive’s worth of capacity and some write speed. With today’s large drives, that second layer of protection matters more than it used to, mostly because of what happens during a rebuild. Here’s how the two compare, and what each looks like from the data recovery side.
How RAID 5 and RAID 6 work
Both levels stripe data across all the drives in the array and add parity: calculated data that lets the array reconstruct the contents of a missing drive. Parity is spread across every member rather than kept on one dedicated drive. (If you’re new to the terms, start with what is RAID.)
- RAID 5 (single parity) stores one parity block per stripe. If one drive disappears, every missing block can be recalculated from the remaining data and parity.
- RAID 6 (dual parity) stores two independent parity blocks per stripe, calculated differently. With two separate equations, the array can solve for any two missing drives.
RAID 5 vs RAID 6 at a glance
| RAID 5 | RAID 6 | |
|---|---|---|
| Minimum drives | 3 | 4 |
| Survives | 1 drive failure | Any 2 drive failures |
| Capacity lost to parity | 1 drive’s worth | 2 drives’ worth |
| Read performance | Good | Good |
| Write penalty (small writes) | 4 disk operations per write | 6 disk operations per write |
| During a single-drive rebuild | No redundancy left | Still protected against one more problem |
| Best fit | Smaller arrays, smaller drives, data that’s well backed up | Larger arrays, large drives, data that must stay online |
Capacity
RAID 5 gives up one drive to parity; RAID 6 gives up two. The difference shrinks as the array grows:
- Four 8 TB drives: RAID 5 gives about 24 TB usable; RAID 6 gives about 16 TB.
- Eight 8 TB drives: RAID 5 gives about 56 TB; RAID 6 gives about 48 TB.
On a four-drive array, RAID 6 costs a third of your usable space compared with RAID 5. On a large array, the extra drive is a much smaller fraction of the total, which is one reason larger arrays lean toward RAID 6.
Performance and the write penalty
Reads are similar on both: data is striped across many drives, so large reads are fast.
Writes are where they differ. When you change a small piece of data, the array has to update the parity too. On RAID 5 that typically means reading the old data and old parity, then writing the new data and new parity: four disk operations for one write. RAID 6 has a second parity block to update, so it’s typically six operations. This is called the write penalty.
For large sequential writes (backups, media files), a good controller can write full stripes and the penalty matters less. For lots of small random writes (databases, virtual machines), RAID 6 will be noticeably slower than RAID 5, and both trail RAID 10. Controller and SSD caching can narrow the gap.
Rebuild times and the real risk
This is the part that should drive your decision.
When a drive fails, the array runs degraded. Replace the drive and the array starts a rebuild: it reads every block on every surviving drive, recalculates the missing data, and writes it to the new drive. On large modern drives, that can take many hours or several days, and it puts the surviving drives under sustained, heavy load the whole time.
During a RAID 5 rebuild, there is no redundancy left. Two things can go wrong:
- A second drive fails outright. Drives in an array are often the same model, same age, and same batch, with the same hours of wear. The stress of a full rebuild can be what tips a weak drive over.
- A surviving drive hits an unreadable sector. It doesn’t take a whole second drive failing. If the array needs a block from a surviving drive and that block can’t be read, there’s no second source to rebuild it from. Depending on the controller, that stripe is lost, or the rebuild stops and drops the drive, leaving the array offline.
Larger drives mean longer rebuilds and more data that must be read perfectly, so the window for trouble is wider than it was when drives were small.
RAID 6 changes this. If one drive has failed and the rebuild runs into an unreadable sector on another, the second parity can still reconstruct the block. It can even lose a second whole drive mid-rebuild and keep going. That’s the main reason many administrators treat RAID 6 as the default for arrays of large drives.
Neither one makes a rebuild risk-free. Our guide to degraded RAID 5 and failed rebuilds covers how to rebuild more safely, starting with copying critical data off before you begin.
When to choose RAID 5 vs RAID 6
RAID 5 can make sense when:
- The array is small (three to five drives) and the drives are modest in size
- Write performance matters and you can’t use RAID 10
- Everything on it is backed up elsewhere, so a failed rebuild is an inconvenience rather than a disaster
RAID 6 is usually the better choice when:
- You’re using large drives, where rebuilds take a long time
- The array has many drives, which increases the odds that one of them is weak
- The data needs to stay online through a rebuild
- Replacement drives might not arrive immediately, leaving the array degraded for a while
On NAS units, dual-redundancy options such as Synology’s SHR-2 give you RAID 6-style protection. Whichever you choose, keep a real backup. Parity protects against drive failure, not against deletion, ransomware, or a failed rebuild. The 3-2-1 backup rule covers the basics.
RAID 5 losing two drives vs RAID 6 losing three
Here’s what each level looks like when its protection runs out.
RAID 5 with two failed drives
With two members missing, a RAID 5 array can’t reconstruct its stripes, so the volume goes offline or the NAS reports a crashed pool. The important detail is how the two drives failed. Very often:
- The first drive failed some time ago, unnoticed, and has stale data.
- The second drive didn’t die completely. It dropped out of the array because of a few unreadable sectors or a timeout during heavy load, and most of it is still readable.
In that situation, the data is usually still there. A lab images every member, including the weak drive (after repairing it if it has a mechanical fault), identifies which drive is stale, and reconstructs the array virtually from the most current set of images. What hurts the odds is what often happens next on site: forcing the stale drive back online, starting a new rebuild, or reinitializing the array to “get it working again.”
RAID 6 with three failed drives
The same logic applies one level up. RAID 6 needs all but two members, so with three missing it goes offline. Recovery depends on getting enough of the failed drives readable to bring the set back within its tolerance: in practice, imaging at least one of the three “failed” drives well enough to work with. Dropped drives with bad sectors and drives with repairable mechanical faults are common, so this is often achievable.
In both cases, the less you’ve done to the array since it failed, the better. Our RAID failure guide has the full list of what to avoid.
The bottom line
RAID 5 gives you more space and faster writes but leaves you with zero redundancy during a rebuild, which is exactly when large drives are most likely to throw an error. RAID 6 costs an extra drive and some write speed, and in return it can absorb a second problem mid-rebuild. For arrays of large drives holding important data, RAID 6 is usually the safer choice. Either way, keep a backup.
If your array has already lost more drives than it can tolerate, our RAID 5 data recovery service also handles RAID 6 and other parity arrays by mail-in. You get a custom quote after a free evaluation, and you pay nothing unless we recover your data.