Common questions and fixes for CrystalDiskMark.
If you run CrystalDiskMark with Administrator rights, network drives do not appear in the drive list.
Fix: Run as a standard user (when UAC asks, choose No so the program runs without admin rights). Then the network drive will be available for benchmarking.
Results can differ because:
Use the same test size and similar settings when comparing. CrystalDiskMark results are not comparable between different major versions.
If the benchmark fails, it is often due to insufficient rights to create the test file on the selected drive.
Fix: Run CrystalDiskMark as Administrator (right-click → Run as administrator, or UAC: Yes). Do not use admin if you need to test a network drive (see above).
In CrystalDiskMark, MB/s means 1,000,000 bytes per second (decimal). 1 GB = 1000 MB in this context. For binary units, 1 GiB = 1024 MiB.
Benchmarking writes data to the drive and can shorten the life of SSDs and USB memory over time. Use benchmarking occasionally for comparison; avoid constant stress tests on drives you need to keep for a long time.
It depends on test size, number of runs, and drive speed. With default settings (e.g. 1 GiB, 5 runs), a fast NVMe might finish in one to two minutes; a slow HDD or USB stick can take much longer. Use a smaller test size (e.g. 64 MiB or 128 MiB) or fewer runs (1–3) for a quick check. The Demo profile is shorter than the full Default or Peak profiles.
For NVMe or SATA SSDs, 1 GiB (default) or 2–4 GiB is common and gives representative results. For HDDs, 1 GiB is usually enough. For USB flash drives or very slow storage, use 64 MiB or 128 MiB so the test does not take too long and to avoid filling the drive. Very large sizes (e.g. 32–64 GiB) can show sustained write behavior on SSDs with large caches but take a long time.
The ZIP (portable) version does not install anything: you unzip and run the executable. Use it if you want no trace on the system, want to run from a USB drive, or prefer to update by replacing the folder. The Installer adds shortcuts and registers the app for uninstall via Windows Settings. Both use the same program; benchmark results are identical. The installer may include ads (as noted on the download page); the portable build from the developer is ad-free.
CrystalDiskMark from the developer's official distribution or the Microsoft Store is the genuine software. The executables are digitally signed so Windows can verify the publisher. Always download from these official sources. If your antivirus flags the file, it can be a false positive (some security software is sensitive to benchmark tools that do raw disk access). You can check the file on VirusTotal or similar services. Do not download from unknown or third-party download sites.
SEQ (sequential) means the test read or wrote data in large, contiguous blocks (e.g. 1 MiB at a time in order). This reflects speed when copying large files. RND (random) means the test used small blocks (e.g. 4 KiB) at random positions. This is closer to real-world use when loading many small files or running applications. Sequential speed is usually higher than random. For SSDs and NVMe, both numbers matter: sequential for big transfers, random for snappiness.
Manufacturers often advertise sequential speeds under ideal conditions (empty drive, specific test). In practice, write speed can be lower because: (1) the drive is partially full and the controller has less free space to work with, (2) the drive has a write cache that fills up so sustained writes slow down, (3) test data (random vs 0-fill) and test size affect the result, (4) the PCIe or SATA link and driver matter. Running CrystalDiskMark with a larger test size (e.g. 4–8 GiB) can show how the drive behaves after the cache is saturated. This is normal and does not necessarily mean a defect.
Yes. You can run "All" or individual tests as many times as you want. Each run overwrites the previous result in the window. Use File → Copy or Save as text/image before starting the next run if you want to keep a snapshot. Running many write-heavy benchmarks in a row will wear SSDs and USB flash more; for occasional comparison, it is fine.