CrystalDiskMark: what it does and who it is for
CrystalDiskMark is a disk benchmark that quantifies how fast a storage device transfers data. Each run measures sequential reads and writes in large blocks, random reads and writes in small blocks, and the same random patterns at a deeper queue depth, reporting throughput in megabytes per second and IOPS alongside microsecond latency. Because the workload set is fixed, results are comparable between machines, between drives, and before and after a change such as enabling a different storage driver or moving a drive to another port. The output grid is designed to be copied straight into a support thread or a buying decision.
How the workflow fits together
Select the drive, choose the number of test passes, the test size and the queue and thread depth, then start the run. The default profile covers the five measurements people quote most often, and the progress display updates live as each block completes. When the run ends, the result grid can be copied to the clipboard or saved as an image or text file. Checking the drive letter first matters, because benchmarking a volume that holds the running system will mix in background activity from the operating system and from anything else currently writing to that disk.
Configuration, limits and trade-offs
Test size is the main trade-off. A small size such as 1 GiB finishes quickly but can be served largely from the drive's cache, while 8 GiB or more reflects sustained behaviour on a large solid-state drive. The queue and thread depth changes what is being measured: depth 1 approximates light desktop use, while deeper queues approximate server or heavy multitasking load. Write tests consume endurance, so they are worth running sparingly on a nearly full drive. Results also depend on the interface and link width, so a fast drive in a slow enclosure reports the enclosure's ceiling rather than the drive's, and caching layers in the operating system can flatter small sequential writes.
When to choose something else
Choose another benchmark if you need to test across many files or directories, examine mixed workloads, or produce a vendor-style endurance report; tools aimed at professional storage validation cover those cases. For a quick sanity check that a newly installed drive works, the manufacturer's own utility or copying a file of known size is often enough. It is also worth remembering that benchmark numbers measure peak conditions, not the everyday speed of applications.
- Best for
- Buyers and builders who want comparable before-and-after numbers for a new drive, a new port or a changed storage setting.
- Good to know
- Small test sizes can be served from cache, write tests consume endurance, and results describe peak conditions rather than everyday speed.
CrystalDiskMark runs a fixed workload so results can be compared between drives and machines.
How to get started
- Install CrystalDiskMark and select the drive to test.
- Choose a test size such as 1 GiB for a quick run.
- Leave the default queue and thread depth unless you need a specific profile.
- Press All to run the full set of measurements.
- Copy the results to the clipboard or save them as an image.
- Repeat with a larger test size when you want sustained numbers.
Questions & answers
Why do my numbers differ from the drive specification?
Specifications quote peak sequential figures under ideal conditions. Real results depend on the interface, the enclosure and background load.
Is benchmarking bad for an SSD?
Write tests consume a small amount of endurance, so running them occasionally is fine and running them constantly is not.
Should I test the system drive?
You can, but background activity from Windows will affect the result. A secondary or empty volume gives cleaner numbers.
