CrystalDiskMark

Name: CrystalDiskMark
Availability: InStock
Author: Crystal Dew World

CrystalDiskMark is a free, open source disk benchmark tool for Windows by Crystal Dew World. It measures sequential and random read/write speeds for SSDs, NVMe drives, hard drives, and USB flash drives. Released under the MIT license.

What it does

Select a drive from the dropdown, click All, and CrystalDiskMark runs four tests. The sequential test (SEQ1M Q8T1) measures how fast the drive handles large contiguous blocks, reflecting file copy and video editing performance. The random 4K test (RND4K Q32T1) measures small block random access with multiple queue depths, the metric that matters most for everyday OS responsiveness.

Results appear in MB/s by default, but you can switch to IOPS or GB/s. The test file size is adjustable from 16 MiB to 64 GiB. Larger files give more accurate results on drives with SLC caching, where a small test might only hit the fast cache and miss slower NAND underneath. You can set the number of passes from 1 to 9, with the final score averaged.

Two measurement profiles are available: Peak mode pushes the drive to its theoretical maximum, while Real World mode simulates typical desktop workloads. Copy results to clipboard with Ctrl+C for sharing on forums or in spreadsheets.

Advantages

Simple one click operation, no configuration needed for a basic test
Tests both sequential throughput and random 4K performance in one run
Open source and completely free with no ads or bundled software
Widely used benchmark, making results easy to compare with online reviews

Drawbacks

Windows only, no macOS or Linux version
Does not test sustained write performance over long periods
No drive health monitoring (use CrystalDiskInfo for that)
Results can be misleading on drives with aggressive SLC caching if test file size is too small

Who it is for

I run CrystalDiskMark every time I install a new SSD or NVMe drive to verify it hits the advertised speeds. It is also the first tool I reach for when a machine feels slow and I suspect storage is the bottleneck. This is the standard benchmark that hardware reviewers and PC enthusiasts have used for years.

Frequently Asked Questions

Is CrystalDiskMark free?

Yes. CrystalDiskMark is free and open source under the MIT license. There are no paid versions or feature restrictions.

What do the CrystalDiskMark results mean?

SEQ1M measures large file transfer speed. RND4K measures small random access speed, which affects how fast your OS and applications feel in daily use. Results are shown in MB/s.

How do I get accurate results?

Use a test file size of at least 1 GiB to avoid SLC cache effects on SSDs. Close other programs during the test. Run multiple passes and let the drive cool between tests for consistent numbers.

Is CrystalDiskMark the same as CrystalDiskInfo?

No. CrystalDiskMark measures speed performance. CrystalDiskInfo monitors drive health through SMART data. They are separate programs by the same developer.

Features & How-To Guide

#	Feature	How to use
1	Random 4K read and write test	Click All › RND4K Q32T1 result shows 4KB random operation speed with 32 queues - important indicator for daily use.
2	Sequential read and write test	Choose drive from dropdown list › Click All › SEQ1M Q8T1 result shows sequential speed in MB/s.
3	Tested disk selection	Dropdown list at top › Choose partition (C:, D:, etc.) or external USB drive › Test applies to selected drive.
4	Copying results to clipboard	Edit › Copy (Ctrl+C) › Results in text format ready to paste on forum or spreadsheet.
5	Unit switching (MB/s vs IOPS)	Menu Theme › Choose MB/s or IOPS or GB/s › IOPS useful for comparing random disk performance.
6	Resize test file size	Dropdown list with size › Choose from 16 MiB to 64 GiB › Larger size gives more accurate results but test takes longer.
7	Change number of test passes	Dropdown list with number › Set from 1 to 9 passes › Final result is average of all passes.
8	Switch measurement mode peak or real	Settings › Queues & Threads › Profile: Peak (maximum performance) or Real World (typical load) › Results differ significantly.