Allocation unit size: a reproducible 4 KiB vs 64 KiB space example

Reproduce how four small files occupy 20,480 or 262,144 bytes in a simple 4 KiB vs 64 KiB cluster model. Separate space rounding from real filesystem behavior and performance.

Published 2026-10-24 · AI-assisted Ithy editorial. Primary references checked 2026-10-24. Original examples are labeled; this is not a named human expert review.

An allocation unit, often called a cluster, is a filesystem allocation boundary. It is different from a drive's physical sector, an SSD flash page and the number of bytes in a file. Changing it is not a universal way to make an SSD faster or last longer.

An original four-file example

Imagine four ordinary files containing 1, 1,000, 4,096 and 4,097 bytes. Their combined logical size is 9,194 bytes. In a deliberately simplified model, each nonempty file gets enough whole clusters to contain its content.

Illustrative allocation, in bytes; not a disk benchmark
File bytesAllocated with 4,096-byte clustersAllocated with 65,536-byte clusters
14,09665,536
1,0004,09665,536
4,0964,09665,536
4,0978,19265,536
Total20,480262,144

The unused tail space in this model is 11,286 bytes with 4 KiB clusters and 252,950 bytes with 64 KiB clusters. Those are arithmetic results for these four invented file sizes. They are not estimates of a typical user's disk, measured performance or write endurance.

Reproduce the numbers without touching a drive

For a file of s bytes and cluster size c, compute ((s + c - 1) // c) * c. The integer division rounds up to a complete allocation unit. The attached Python program prints an eight-row CSV and does not inspect, format or write any drive.

Download the original Python model and its computed CSV. The program was executed for this article and its totals checked independently. One KiB means 1,024 bytes; the unit is explicit to avoid confusing decimal KB with binary KiB.

Where the simple model stops

Real filesystems have metadata and special cases. Small NTFS files can have resident data; compression, sparse allocation, deduplication and other features can change physical storage. Therefore the rounding formula should not be advertised as an exact reconstruction of every “size on disk” value.

Microsoft's NTFS overview describes features and cluster-size-dependent limits. Its format reference documents filesystem-specific allocation choices. A filesystem cluster is not a promise that a drive performs one hardware operation of that exact size.

A practical decision sequence

  1. Choose a filesystem supported by the devices that must read the drive.
  2. Start with its appropriate default unless the application vendor documents a specific requirement.
  3. Separate a workload dominated by tiny files from one dominated by large files.
  4. Compare candidates only with a representative, repeatable workload and sufficient backup/recovery preparation.

A larger cluster can increase rounding waste for small files in this model. It does not follow that it always improves throughput, reduces fragmentation or extends SSD life. Those are different claims requiring actual measurements and a defined environment.

Inspecting a setting is different from formatting

For NTFS, Microsoft's fsutil fsinfo documentation explains information queries such as fsutil fsinfo ntfsinfo C:; permissions may be required. This article did not execute that command on a Windows device. The format command is a data-changing operation, not a harmless way to inspect a setting.

Cluster rounding is also different from the unit conversion that makes a 1 TB drive approximately 931 GiB. Diagnose the specific boundary before changing a drive to address a display difference.

For a correction, send the specific claim and a supporting source through Ithy Help. Research methodology · More research guides.