Understanding CrystalDiskMark Results
What the numbers mean and typical ranges for NVMe SSD, SATA SSD, and HDD. CrystalDiskMark reports sequential and random read/write speed so you can compare drives, verify upgrades, or document performance for reviews and support.
Results depend on the drive, test size, profile (Default, NVMe SSD, etc.), and system load. Use the same settings when comparing two drives. See our How to Use guide and FAQ for more.
Units Used
- MB/s – 1,000,000 bytes per second (decimal). 1000 MB/s = 1 GB/s.
- IOPS – I/O operations per second (useful for random 4K tests).
- μs – Average latency in microseconds.
In CrystalDiskMark: 1 GB = 1000 MB = 1000×1000 KB = 1000×1000×1000 B (decimal). 1 GiB = 1024 MiB (binary).
Typical Speed Ranges (Approximate)
Results depend on drive, controller, and test size. Use as a rough guide only.
| Drive type | Seq read (MB/s) | Seq write (MB/s) | 4K random (MB/s) |
|---|---|---|---|
| NVMe PCIe 4.0 SSD | 3000–7000+ | 2000–6000+ | 50–200+ |
| NVMe PCIe 3.0 SSD | 1500–3500 | 1000–3000 | 40–150 |
| SATA SSD | 450–560 | 400–530 | 20–80 |
| HDD 7200 rpm | 100–200 | 100–200 | 0.5–2 |
| USB 3.0 flash | 50–200 | 20–100 | 1–10 |
Numbers can be higher or lower depending on drive model, capacity, SLC cache, thermal throttling, and test size. Sequential tests use large blocks (e.g. 1 MiB); random 4K tests use 4 KiB blocks and reflect many small reads/writes.
Interpreting Your Numbers
When you run CrystalDiskMark, focus on:
- SEQ (sequential): High numbers mean fast large-file copy (movies, installers, backups). Important for video editors and anyone moving big files.
- RND 4K (random): Reflects everyday OS and app behavior (loading programs, boot, game levels). Low random speed often means the system feels slow even if sequential is high.
- Read vs write: Some drives have much higher read than write. Both matter: read for loading, write for saving and installing.
- If your result is far below the typical range, check troubleshooting (thermal throttling, background apps, driver, or wrong profile).
Example Benchmark Images
Optional: add screenshots to the images folder for reference.
How to Compare Drives Fairly
- Use the same CrystalDiskMark version and same test settings (test size, profile).
- Close other apps that use the disk. Run the test multiple times and note average.
- For NVMe use the NVMe SSD profile in Settings. For SATA use Default.
- Same test data (Random or 0 Fill) when comparing two drives.
What to Look For in Results
Depending on your use case, different numbers matter more.
- Boot and app load: Random 4K read (Q1T1) and random read IOPS. Higher is better for snappy feel.
- Large file copy (video, backups): Sequential read and write. This is the MB/s you see in file copy dialogs.
- Gaming (loading levels): Often a mix of sequential and random; sequential and 4K random both matter.
- Servers / VMs: Random read/write at higher queue depths (e.g. Q32T1 or Q32T16) and IOPS.
Common Use Cases
Why people run CrystalDiskMark and what they do with the results.
- Before/after upgrade: Run on the old drive, then on the new SSD or NVMe, to see the real-world improvement in numbers.
- Verifying a new or used drive: Check that sequential and random speeds are in the expected range for that model; unusually low values can indicate a problem or fake drive.
- Comparing two drives: Same test size and profile on both; compare average of several runs. Use for buying decisions or build logs.
- Troubleshooting slowness: If the PC feels slow, benchmark the system drive. Very low random 4K can indicate a failing drive, heavy fragmentation, or background activity.
- Reporting and reviews: Save as image or copy to clipboard to include in forum posts, reviews, or support tickets.