SSD and HDD in 2026: defragmentation, TRIM, and myths that need to die

TRIM and retrim: what Windows does to SSDs on its own

TRIM is the command by which the file system tells the drive: “these blocks are no longer needed”. Without TRIM the SSD does not know which internal pages it may pre-erase, and every write to a nominally busy block becomes a read-modify-write cycle — several times slower. Windows 11 sends TRIM automatically on the optimization schedule; doing it by hand is unnecessary. The defrag tool's /Optimize column shows how much garbage has accumulated.

Why does Windows then run “optimization” on an SSD monthly? That is a retrim — a batch TRIM for blocks marked but not cleaned for some reason (usually after power loss). No data defragmentation happens on SSDs: the kernel scheduler has a separate task type for solid-state drives. So “the optimizer touched my SSD, disaster” is a myth — the system does exactly what the drive needs.

The only defensible exception is filesystem micro-fragmentation on SSDs under specific workloads (SQL Server, giant video projects with hundreds of thousands of small files). Windows handles this with slab consolidation on its own; manual intervention remains exotic server territory.

HDD: when defragmentation actually pays

On a magnetic disk the head physically travels between file pieces; at 30%+ fragmentation, large-file read speed degrades severalfold. Microsoft's rule of thumb: under 10% fragmentation is normal and not worth touching; 10–30% gives a noticeable effect on large files; 30%+ means defragment now — the UI is already honestly slow on file operations.

The right HDD maintenance order: clean up first (the defragmenter needs 20%+ free space to maneuver), then analyze (for the fragmentation map), then defragment overnight with the disk idle. Windows does it with its own engine (defrag.exe) and that is enough: 2026-era third-party defraggers do nothing fundamentally better and happily move system files where they should not go.

Health check before any maintenance: chkdsk R: /scan (read-only, safe) surfaces bad sectors, and SMART attributes 05 (Reallocated Sectors), C5 (Pending) and C6 (Uncorrectable) are the main death predictors. A growing 05 means the drive is relocating data and running out of spare; any C5/C6 above zero means you need a backup not “someday” but today.

SSD myths and superstitions

“Never fill an SSD past half” — outdated truth. Modern controllers with over-provisioning work fine to 80–85%; past 90% write performance drops — not from “wear” but from a shortage of clean blocks for garbage collection. Keeping 10–15% free is sensible hygiene, not life support.

“Disable the page file to save the SSD” — an arithmetic error. Home-system paging writes a few GB a day; a modern TLC drive is rated for 600+ TB of writes, so paging alone would kill it in about a century. Disabling the pagefile makes sense only for a specific reason (no crash dumps needed, 32+ GB RAM), which is a separate topic from wear.

“Manual TRIM after deleting big files speeds the drive up” — not anymore: NTFS in Windows 10/11 sends TRIM synchronously with deletion. An extra optimization pass is harmless but useless. The only operations that genuinely restore speed to a stuffed SSD are freeing it to ≤10% full (cleanup) and a vendor firmware update, which brings garbage-collection improvements.

Questions and Answers

Read next