Why Your 4TB Drive Shows 3.63TB: TB vs TiB Explained

You buy a 4TB drive, fit it, and the operating system reports 3.63TB. Nothing is faulty, nothing is missing, and no capacity has been lost. Two different definitions of a terabyte are being used, and this is the single most common storage question there is.

The two definitions

Drive manufacturers use decimal units, where a terabyte is 1,000,000,000,000 bytes — powers of ten, consistent with how prefixes work everywhere else in science and engineering.

Many operating systems report in binary units, where the equivalent figure is 1,099,511,627,776 bytes — powers of two, which is how memory addressing works.

The difference is roughly 9 percent at the terabyte scale, and it widens as capacity grows. A 4TB drive is genuinely 4 trillion bytes; expressed in binary units that is about 3.63 of them.

The correct binary terms are tebibyte (TiB), gibibyte and mebibyte. Some operating systems label their binary figures correctly as TiB; others display binary values while labelling them TB, which is where the confusion comes from.

Where the rest of the capacity goes

Unit conversion accounts for most of the gap. Three other things account for the remainder, and they are worth knowing when sizing an array.

Filesystem overhead. Formatting consumes space for metadata, allocation tables and journals. The amount varies by filesystem and by how it is configured.

Reserved space. Some filesystems reserve a percentage for the system, to prevent a full volume becoming unrecoverable. That space exists and is not available to you.

Over-provisioning on SSDs. Enterprise flash reserves capacity the controller uses for wear levelling and block management. A drive sold as 3.84TB contains more flash than that; the difference is deliberate and it is what keeps write amplification manageable. Our SSD endurance guide covers why.

RAID takes the largest share

In a server the unit conversion is the small part. RAID overhead is the big one, and it is where capacity planning actually goes wrong.

Twenty-four 1.2TB drives is 28.8TB raw in decimal terms. What you can actually use depends entirely on the array design.

RAID 10 mirrors everything, so roughly half.
RAID 6 gives up two drives to parity.
RAID 5 gives up one, and is inadvisable on large drives.
Hot spares come off the top before any of that.

Then apply the binary conversion, then filesystem overhead. The number a user finally sees can be a long way below the figure on the purchase order.

Our RAID levels guide covers the arithmetic in full.

Sizing an array properly

Work backwards rather than forwards, and the surprises disappear.

Start with usable capacity required, in the units your monitoring reports.

Add growth headroom. Storage is easier to plan than to expand, and a nearly full array performs worse — particularly on flash, where write amplification rises as free space falls.

Add filesystem overhead for whatever you are running.

Apply the RAID multiplier for the level you have chosen.

Add hot spares on top.

Convert to decimal to reach the drive capacity to order.

Doing this in advance avoids the common outcome of ordering exactly enough drives to hit a raw number and discovering the usable volume is well short of what the project needed.

Why the definitions differ at all

Briefly, because it explains why neither side is being dishonest.

Binary units arose because memory is addressed in powers of two — a memory chip’s capacity naturally lands on 1,024 rather than 1,000. Early computing borrowed the kilo, mega and giga prefixes as approximations, and the approximation was close enough not to matter at small scales.

Storage does not have that constraint. A drive can hold any number of bytes, so manufacturers use the standard decimal meaning of the prefixes.

As capacities grew, the gap between the two widened from negligible to significant. The tebibyte terminology exists to resolve it, and the remaining confusion is mostly software that still displays binary values under decimal labels.

Practical points

Replacement drives. When replacing an array member, match the stated capacity. A drive sold as 4TB will work alongside other 4TB drives regardless of how the OS displays it.

Slight variations between manufacturers. Drives of the same nominal capacity can differ marginally in exact byte count. In a RAID group the smallest member governs, so a marginally smaller replacement can occasionally cause problems — another argument for matching part numbers rather than capacity labels.

Documentation. When specifying capacity in a purchase order or a project document, state the units. "20TB usable" and "20TiB usable" differ by about 10 percent, which is a drive or two on a large array.

Common questions

Tell us the usable capacity you need and the RAID level, and we will work back to the drive count and capacity to order.