Storage capacity is the one specification that is never quoted the way it is measured. A drive labelled 1 TB shows 931 GB in the operating system, formatted capacity is lower again, and a "500 GB" drive that fails takes 93% of what you thought you had. None of this is fraud; it is a genuine mismatch of conventions that is worth understanding once.
Two conventions, one drive
SI units are powers of ten. IEC binary units are powers of two. They differ increasingly as they grow:
| Name | SI (decimal) | Bytes | IEC (binary) | Bytes | Ratio |
|---|---|---|---|---|---|
| kilobyte | 1 kB | 1,000 | 1 KiB | 1,024 | 1.024 |
| megabyte | 1 MB | 10⁶ | 1 MiB | 1,048,576 | 1.049 |
| gigabyte | 1 GB | 10⁹ | 1 GiB | 1,073,741,824 | 1.074 |
| terabyte | 1 TB | 10¹² | 1 TiB | 1,099,511,627,776 | 1.100 |
The gap grows by 2.4% at each step, and by the terabyte it is 10%. A "1 TB" drive is 10¹² bytes. Divide by 2^30 (one GiB) to get the number a system will display: 10¹² ÷ 1,073,741,824 = 931.3.
Manufacturers use decimal because 1,000,000,000,000 is a round number in the units they sell in and the capacity sounds better. Operating systems inherited binary from the 1980s when it was correct. The IEC introduced TiB to fix the ambiguity; most software still writes "GB" for both, so the argument continues. The data size converter accepts both conventions explicitly.
Formatting takes its cut too
Even expressed correctly, a formatted drive has less usable space than its raw capacity. Filesystems need space for their own structures:
- ext4 reserves about 5% by default for root. On a 1 TB drive that is 50 GB, and a non-root user cannot use it.
- NTFS reserves a small cluster for the master file table and keeps a shadow copy area.
- APFS reserves some space and reports container sizes as a maximum, so the free figure moves in discrete steps.
Put the three effects together and a 1 TB drive typically presents 900-931 GB to a user, sometimes less. On a 2 TB drive the fixed overheads are proportionally smaller, which is one reason capacity-per-pound improves with size.
What actually consumes space
The number that matters is not the file size on disk but the total per item, and the surprises come from a few categories:
- Photos. A modern phone takes 2-5 MB per photo. A 128 GB card with 10% reserved holds roughly 25,000 photos, but the same card as video holds under an hour of 4K.
- Video, overwhelmingly. 4K at a typical 50 Mbps bitrate is about 22 GB per hour; 1080p is 4-5 GB; a high-quality 4K master from a camera can exceed 100 GB per hour. Video, not photos, is what fills a family archive.
- Lossless audio. CD-quality WAV is about 10 MB/min, so an hour is 600 MB and a full album is 400-600 MB. FLAC is roughly 60% of that.
- Versioned data and backups. A "1 GB" dataset that is backed up locally, snapshotted nightly and mirrored to a second machine is three copies, not one. This is the most common reason people run out of space unexpectedly.
- Logs and caches. On a machine used for development, logs, package caches and build artefacts can exceed the source code by an order of magnitude and are usually deletable rather than needing storage.
Planning capacity with headroom
Three rules, in order of value:
- Budget in the unit you will see. If your operating system reports GiB, plan in GiB. Applying a 10% marketing gap twice — once on purchase, once on planning — is a common double-count.
- Keep 15-20% free. A filesystem that is 95% full is slow (ext4 reserves space for defragmentation), harder to extend, and more likely to fail on a large write. 20% free is not waste; it is the operating margin.
- Count copies, not files. Decide how many copies of each category you keep: working, backup, archive. Then multiply. For photos and video that are irreplaceable, three copies is the standard, and one of them should be off-site.
A worked example: a family wanting 10 years of photos and video. At 20,000 photos a year averaging 4 MB and 50 hours of 4K video a year at 22 GB, that is 80 GB of photos plus 1,100 GB of video — about 1.2 TB before any backup. With one local archive and one cloud copy, 2.5 TB of usable capacity is a sensible target, which means a 4 TB drive (usable around 3.6 TB) rather than two 2 TB drives.
Bit, byte and the prefixes that confuse everyone
Before the storage gap, one distinction that catches people out: a bit is one binary digit, a byte is 8 bits. Prefixes attach to both, and the usage is not consistent.
- bps (lowercase s) is bits per second — network speed.
- Bps (uppercase S) is bytes per second — file transfer rate.
- 100 Mbps = 12.5 MB/s — the factor of 8 is the whole story, and it is covered in the download time guide.
Storage is measured in bytes and their multiples, so a "700 MB" file is 700 million bytes, while a "700 Mb" file would be 87.5 MB. The lowercase b in a download progress bar is a real and confusing inconsistency.
Why the 931 figure is not a shortchange
The arithmetic is unambiguous once the two conventions are named. A manufacturer sells 1,000,000,000,000 bytes and calls it 1 TB because 10^12 is a round number. The operating system counts in binary multiples, so it divides by 2^30 to display GiB:
10^12 ÷ 1,073,741,824 = 931.32
The 6.9% difference is the accumulated drift of 2.4% per step over three steps from MB to TB. At exabytes it would be 9%. This is why the standards bodies introduced TiB, and why the drive label and the operating system are both technically correct.
Format overhead, in numbers
Beyond the unit gap, the filesystem claims space for itself. On a 1 TB ext4 drive with the default 5% reserved-blocks setting, root can use the whole drive but any other user sees about 5% less, and that reserve exists so the system keeps working when the disk is nearly full. Windows NTFS reserves a portion for the master file table. APFS reports container size as a fixed maximum, so free space moves in jumps.
Practical usable capacity on a nominal 1 TB drive: 900-931 GB, and closer to 900 after formatting. On a 2 TB drive the proportional loss is smaller, which is part of why larger drives offer better value per byte.
What consumes capacity, measured
Per-item figures that make the planning concrete:
- Photo, modern phone: 2-5 MB each, so 20,000 photos is 40-100 GB
- Video 4K, 50 Mbps bitrate: 22 GB per hour
- Video 4K, camera master 100 Mbps: 45 GB per hour
- Video 1080p, 8 Mbps: 3.6 GB per hour
- WAV audio: 10 MB per minute, so 600 MB per album
- FLAC: about 55% of WAV
- RAW photo from a DSLR: 25-45 MB, roughly ten times a JPEG
The pattern is that video dominates everything else. A decade of family photos is tens of gigabytes; the same decade of family video is multiple terabytes. Anyone planning storage by "how many photos" is planning for the wrong thing.
Copies, not files
The question that determines capacity is not "how much data" but "how many copies of it do I keep". A modest photo library of 200 GB becomes 600 GB with one local archive and one cloud backup, and 900 GB with a second local copy. Backups are not free in space, and the 3-2-1 rule (three copies, two media types, one off-site) is a statement about total capacity as much as about safety.
A worked plan for ten years of family photos and video, at 20,000 photos a year averaging 4 MB and 40 hours of 4K video a year:
- Photos: 20,000 × 4 MB = 80 GB per year
- Video: 40 h × 22 GB = 880 GB per year
- Ten years: 9.6 TB of originals
- Two copies plus 20% headroom: 9.6 × 2 × 1.2 = 23 TB
That is two 16 TB drives, or one 32 TB drive plus a cloud archive — and it is a very different answer from the one most people give when asked how much storage they need.
Frequently asked questions
Why does a 1 TB drive show as 931 GB?
Manufacturers count in decimal (1 TB = 10^12 bytes) and operating systems display in binary multiples (1 GiB = 2^30 bytes). 10^12 divided by 2^30 is 931.3, so the 6.9% gap is the accumulated 2.4% drift of three prefix steps. The IEC's TiB name exists to remove exactly this ambiguity.
Is a 2 TB drive twice as useful as a 1 TB drive?
For capacity, yes. For reliability, two 1 TB drives give you two independent failure domains while one 2 TB drive does not. For value per dollar, larger drives are usually cheaper per byte, but they concentrate more data behind a single failure point.
How much space do 10 years of family photos take?
About 800 GB of JPEGs at 20,000 photos a year and 4 MB each. If you also keep 4K video at 40 hours a year, video adds roughly 880 GB per year and dominates completely — about 9.6 TB for the decade before any backups.
Why is my new drive not showing the advertised capacity?
Three stacked reasons: the decimal/binary gap (6.9% at 1 TB), the filesystem reserving space (typically 5% on ext4, less on NTFS), and the partition layout. Together they usually account for the difference; a significantly smaller figure suggests a faulty drive rather than a convention mismatch.
Frequently asked questions
Why does a 1 TB drive show 931 GB?
Manufacturers use decimal units where 1 TB = 10^12 bytes, while operating systems use binary units where 1 TiB = 2^40 = 1,099,511,627,776 bytes. 10^12 divided by 2^30 is 931.3, so the same drive reads as 931 GiB. Nothing is missing; the labels count differently.
How much usable space does a drive really have?
Less than the label, for two reasons: the decimal/binary gap above, and the filesystem's own overhead. A 1 TB drive typically offers about 900-931 GB, and after formatting shows around 900 GB with a small amount reserved for the filesystem structure.
Is 500 GB better value than 1 TB?
Usually not on price per unit, but sometimes on total cost. Two 500 GB drives can cost more than one 1 TB drive of the same total capacity, but buying two smaller drives gives you two independent failure domains — if one fails, the other still holds your data, which a single larger drive does not.
How much storage do I need for photos and video?
A modern phone takes 2-5 MB per photo and 100-400 GB of 4K video per hour. A 128 GB phone holds roughly 30,000 photos or about 5 hours of 4K. For a family archive, budget 1 TB per 2,000 photos plus video, and add 20% headroom.