Guides · measured July 2026 · 7 min read
How to find the large files eating your disk.
The “low disk space” warning is older than most of the files that trigger it, and thirty years on, Windows still tells you the drive is full without telling you which files did it. To answer that, something has to measure every file on the disk, and the built-in tools do it the slow way: by walking the whole drive, a job that on a real machine takes about 67 seconds, long enough to refill your coffee. Here is how to rank every file on the drive by size the instant way, and why the OS makes you wait.
What the question really asks
It is a whole-drive sort. That is why it is slow.
Finding the large files eating your disk means ranking every file on the drive by size and pulling out the biggest, wherever they happen to sit. That is a question about the whole drive, not one folder, and the built-in tools answer it the only way they can without a stored index: by opening and measuring every file, one at a time. On a small folder that is instant. On 4.47 million files, it is a walk, and the walk is the wait.
The problem. Your disk fills up and the OS shows you the total, never the culprits, because finding the biggest files means measuring every file, and the built-in tools do that by walking the whole drive.
The slow way. File Explorer sorts one folder at a time; the command line and disk scanners open and measure every file on every run, so the cost grows with your drive, a 67-second-class job on 4.47 million files.
The fast way. An always-warm index already knows every file's size, so ranking the whole drive by size is a lookup, not a walk: the microsecond-to-millisecond range, about seven hundred whole-drive lookups inside a single blink.
It stays true. Delete a 9 GB file and it drops out of the results in about a millisecond. The index is live, so the size view is never a stale snapshot from the last scan.
Not just for you. The same instant, size-aware view is handed to any AI agent over MCP, so “what is eating my disk?” is one call, not a directory crawl.
The trade is the same one behind every fast search box. Measure everything once, in the background, and keep the numbers current, then every “which files are biggest” question is a jump to an answer that is already assembled. Walk the disk fresh each time, and you pay the full cost on every single ask.
The walk vs the index, drawn
One measures every file. One already knows.
Ask both approaches the same thing: every file larger than 1 GB, anywhere on the drive. A walk has no stored sizes, so it does the only thing it can: open the first file, measure it, open the next, measure it, all 4.47 million of them. An always-warm index already recorded every file's size when it read the drive, so the size filter simply lights up the handful of matches. Same question, same disk, two completely different amounts of work.
Top: the walk must open and measure all 4.47M files to know which are big, a job that grows with the drive, a 67-second-class operation on this machine. Bottom: the index already holds every size, so the filter resolves to the matching files in the microsecond-to-millisecond range, and it re-checks the instant a file is deleted. The cost of a walk is set by the size of the disk; the cost of a lookup is set by the number of matches.
Your options, honestly
Five ways to find the big files, and what each one costs.
Every Windows machine gives you the first three for free, and most people eventually download the fourth. They share one trait: none of them keeps a complete, current list of every file's size, so on a large disk they either work one folder at a time or walk the whole tree from scratch. Only the last keeps every size pre-recorded and live.
| The method | Where it is | What it does | Whole drive · 4.47M files |
|---|---|---|---|
| File Explorer | Sort a folder by Size | Sorts one folder at a time; totalling a folder means walking everything inside it | per-folder only |
| Storage settings | Settings › System › Storage | Buckets usage by category (apps, temp, documents) not by individual file; recomputes on open | slow, by category |
| The command line | PowerShell · Get-ChildItem -Recurse | Walks the whole tree and measures every file on every run, no stored sizes | grows with the drive |
| A disk-usage scanner | A separate download | Scans the whole drive each time you open it; the picture goes stale the moment you delete a file | a full re-scan each time |
| Interlinked | Taskbar app + agent tools | Whole drive pre-indexed with every file's size, kept live: rank by size is a lookup | 139 µs median |
Methodology: Ryzen 9 9950X3D · 64 GB DDR5 · NVMe · Windows 11 · 4,470,000 files. The Interlinked figure is the median whole-drive lookup across 20 queries on this corpus (85 µs on a single file, 2 µs best case). The walk-based rows scale with folder and drive size, so they are described by behaviour rather than one fixed number; on the same drive a full whole-drive walk is a 67-second-class job. For the name-search version of the same gap (412,000× faster than VS Code, ~480,000× faster than Windows Search), see how to search every file on Windows.
The anchors · Windows Search & Spotlight
The search boxes find words. Not the biggest files.
The two search tools everyone reaches for were built to answer a different question: where is the file called X, or which files contain the word Y. Neither is built to rank an entire volume by size. Windows Search covers only a slice of the drive by default and falls back to a walk on a miss, a 67-second median on our test drive. Spotlight is faster because macOS indexes the whole volume, but it still sorts one folder at a time in Finder, not the whole disk by size. And the OS built-ins each answer only part of it: File Explorer sorts a single folder, Storage settings buckets usage by category, the command line walks the tree fresh every run.
Windows
Finds by name and word; covers a slice of the drive; no whole-drive size ranking
67 s median on a miss
macOS
Finds across the whole volume, but sorts a folder by size, not the whole disk
instant to search, not to total
Windows
Every file plus its size, always warm: rank the whole drive by size directly
139 µs median
The lesson of both anchors is the same one from Spotlight vs Windows Search: coverage and freshness are what make an answer instant. Cover every file, keep every size current, and “which files are biggest” stops being a walk and becomes a lookup, on the whole drive, not one folder at a time.
The instant answer
Every file's size, already known. Look up, never walk.
Interlinked keeps a live index of every file on the machine (names, contents, and sizes) and stays current as files change. Because every size is already recorded, ranking the whole drive by size is a lookup, not a walk: it lands in the same microsecond-to-millisecond band as a filename search, and it never falls back to measuring the disk. Every repo, plus the ninety percent of the drive that was never in a repo: the caches, the virtual-machine disks, the old backups where the giants actually hide.
And it is not only for you at the keyboard. The same warm index answers any AI client that speaks MCP. One install auto-configures 19 of them (Claude Code, Cursor, Copilot, Codex, Cline and more), so when you ask an agent to clean up build artifacts or find what is eating the disk, it gets a size-aware whole-drive view in one call instead of a directory crawl. That file-hunting tax on agents is the larger story in the backbone of the LLM economy: the same job that took one agent 6 minutes 57 seconds and 71 tool calls comes back in 16 milliseconds and one.
Questions people ask
Finding big files, answered.
What is the fastest way to find large files on Windows?
An engine that already stores every file's size and keeps it current, so ranking the whole drive by size is a single lookup rather than a fresh walk of the disk. The built-in, walk-based tools get slower as your drive fills; a warm index does not. The mechanics of Windows file search (and why it crawls) are laid out in how to search every file on Windows.
Why is my C: drive full when I can't find the big files?
Because the OS reports totals, not the individual offenders, and finding them means measuring every file, a whole-drive walk it avoids until you force it. The giants usually hide outside your user folders: a virtual-machine disk, a video export, a forgotten node_modules, an old system backup. An index that covers the whole drive sees all of them.
Can Windows Search or Spotlight sort my whole drive by size?
Not really. Both are built to find files by name or by a word inside them, not to rank an entire volume by size. Windows Search also covers only a slice of the drive by default and falls back to a walk on a miss, which is why it took a 67-second median on our test drive. The Spotlight comparison covers the macOS side.
How is this different from a disk-usage scanner?
A scanner walks the whole drive each time you open it, and the picture goes stale the instant you delete something, so you rescan. An always-warm index keeps every file's size current as files change, so the answer is both instant and never out of date. You never wait for a scan bar again.
Does keeping every file's size indexed slow down my machine?
No. At rest the whole engine sits around 44 MB (less than a single browser tab) and the index on disk stays under 1% of the drive. It updates as files change rather than rescanning, so a saved or deleted file is reflected in about a millisecond.
Do this today
Stop scanning for the giants. Look them up.
Install once. One signed installer sets up the whole-drive index (every file's name and size) and auto-configures 19 AI clients: Claude Code, Cursor, Copilot, Codex, Windsurf, Zed, Cline and the rest. No settings to tune, no JSON to edit.
Let it read the drive once. Filename and size search work immediately; contents fill in behind them. From then on, ranking the whole drive by size is a lookup instead of a walk: the thing that made the built-ins slow simply can't happen.
Sort the whole machine by size. Not one folder: the entire drive at once, biggest first, in the microsecond-to-millisecond range. The virtual-machine disks, the forgotten caches, the old backups all surface together.
Delete, and watch it stay true. Remove a giant and it drops out in about a millisecond, because the index is live. No rescan bar, no stale snapshot, and your agent sees the same current view over MCP.
Everything local is free forever.
Everything (the app plus the same index exposed to every AI client over MCP) is free forever, on all your devices, no card. Hosting your projects starts at $5.99/mo. Download for Windows.
Keep reading: how to find a file you forgot the name of, Interlinked vs Windows Search, and the backbone of the LLM economy.