Where your RAID capacity actually goes.
The number on the box is never the number you get. Enter your drives to see what’s really left — and what took the difference.
—
after RAID overhead—
the same bytes, counted in 1024s—
on a fresh NTFS volumeThe first two are the same bytes — only the counting changes. The third is day-one free space; it drifts down as metadata grows.
Results
Enter your disks above and the full breakdown appears here.
Enter your disks above to see how this level lays them out.
RAID 6 (dual parity)
Striped with two disks worth of parity. Survives any two failures.
Two independent parity calculations are distributed across the array, so a second failure during a rebuild is survivable.
Rewards
- Survives a failure during rebuild — the case that kills RAID 5
- Still efficient once you pass six disks
- Sensible default for large arrays of large disks
Risks
- Two disks of capacity gone
- Heavier write penalty than RAID 5
- Rebuilds are slow and stress the remaining disks
Rebuild time
—
| Disk | Idle | Light use | Busy |
|---|
After you format it
Filesystem metadata takes another cut. Figures are approximate — real overhead moves with volume size, cluster size and how many files you store.
| Filesystem | Overhead | Free space | Notes |
|---|---|---|---|
| FAT32 | ~0.1% | — | Maximum file size 4 GiB. Maximum volume 2 TiB, and the Windows format tool stops at 32 GB. |
| exFAT | ~0.05% | — | No journaling and no permissions. Fine for removable media, weak for an array. |
| NTFS | ~0.4% | — | Master file table plus a 64 MB log. The MFT zone is reserved but reclaimable. |
| ReFS | ~1% | — | Integrity metadata. Windows Server and Pro for Workstations only. |
| ext4 | ~6.6% | — | About 1.6% metadata plus 5% reserved for root. tune2fs -m 0 gives the reserve back. |
| XFS | ~0.5% | — | Grows online but cannot shrink. Plan the size before you commit. |
| Btrfs | ~1% | — | Metadata is duplicated on rotational disks by default, so real overhead runs higher. |
| ZFS | ~4.2% | — | Roughly 1% metadata plus a 3.2% slop reserve the pool keeps for itself. |
| APFS | ~0.3% | — | Container metadata. Snapshots and clones change the picture over time. |
| HFS+ | ~0.5% | — | Catalog file plus journal. Superseded by APFS on modern macOS. |
Every RAID level compared
Minimum disks, what each level survives, and the formula behind its usable capacity. n is the number of disks, g the number of parity groups.
| RAID level | Min disks | Survives | Usable capacity | Mixed disk sizes |
|---|---|---|---|---|
| Single disk (no RAID) | 1 | None | One disk | Used in full |
| JBOD / linear span | 2 | None | Sum of all disks | Used in full |
| RAID 0 (stripe) | 2 | None | n × smallest | Truncated to smallest |
| RAID 1 (mirror) | 2 | n − 1 disks | Smallest disk | Truncated to smallest |
| RAID 5 (single parity) | 3 | 1 disk | (n − 1) × smallest | Truncated to smallest |
| RAID 6 (dual parity) | 4 | 2 disks | (n − 2) × smallest | Truncated to smallest |
| RAID 10 (striped mirrors) | 4 | 1 per mirror pair | ⌊n ÷ 2⌋ × smallest | Truncated to smallest |
| RAID 50 | 6 | 1 per group | (n − g) × smallest | Truncated to smallest |
| RAID 60 | 8 | 2 per group | (n − 2g) × smallest | Truncated to smallest |
| RAIDZ1 (ZFS) | 3 | 1 disk | (n − 1) × smallest | Truncated to smallest |
| RAIDZ2 (ZFS) | 4 | 2 disks | (n − 2) × smallest | Truncated to smallest |
| RAIDZ3 (ZFS) | 5 | 3 disks | (n − 3) × smallest | Truncated to smallest |
| RAID-TP (QNAP) | 5 | 3 disks | (n − 3) × smallest | Truncated to smallest |
| RAID-TM (QNAP) | 3 | 2 per mirror set | ⌊n ÷ 3⌋ × smallest | Truncated to smallest |
| SHR (Synology) | 2 | 1 disk | Sum − largest disk | Used in full |
| SHR-2 (Synology) | 4 | 2 disks | Sum − two largest | Used in full |
| unRAID · 1 parity | 2 | 1 disk | Sum − largest disk | Used in full |
| unRAID · 2 parity | 3 | 2 disks | Sum − two largest | Used in full |