Sequential read
The best-case number. Large blocks read in order, which is what copying a big file or loading a level looks like.
DiskSpeedTest runs the same four tests every time: sequential read, sequential write, random 4K and access latency. Local only, no account, no cloud upload. Eighty seconds and you know exactly what the hardware does.
A single big number hides the truth. Drives behave differently for large files, small files and idle access, so DiskSpeedTest reports each separately.
The best-case number. Large blocks read in order, which is what copying a big file or loading a level looks like.
The same test in the other direction. This is where cheap drives and full caches fall over.
Small scattered reads, measured in IOPS. It predicts how an operating system feels more than any headline figure.
The gap between asking for a block and getting it. Low latency is why a good NVMe drive feels instant.
An optional pass that copies a generated file set and times it end to end, including the filesystem.
Every run is stored locally with a timestamp. Compare a drive before and after a firmware update, or export CSV.
Indicative figures from healthy drives on a modern desktop. Treat them as landmarks, not targets.
| Drive class | Seq read | Seq write | 4K read | Latency |
|---|---|---|---|---|
| NVMe Gen4 | 5,000 to 7,400 | 4,000 to 6,800 | 180K to 400K | 0.03 ms |
| NVMe Gen3 | 2,800 to 3,500 | 1,800 to 3,000 | 120K to 300K | 0.05 ms |
| SATA SSD | 520 to 560 | 480 to 530 | 60K to 95K | 0.08 ms |
| 7200 RPM HDD | 150 to 220 | 140 to 200 | 0.8K to 1.6K | 8 ms |
Operating systems and games mostly move small blocks in unpredictable order. A drive that wins on sequential throughput can still lose on everyday responsiveness, and that is the gap this test exposes.
Higher queue depth lets the controller overlap requests. Real desktop work rarely saturates it.
Four things: sequential read, sequential write, random 4K throughput in IOPS, and average access latency. Together they describe how a drive behaves for large files, small files and everything between.
Yes, briefly. The write tests create a temporary file, measure against it and delete it. You choose the size before the run, from 256 MB to 4 GB, and nothing is left behind afterwards.
No. A standard run writes a few gigabytes at most, which is a rounding error against a modern drive's endurance rating. The endurance profile exists for people who want a longer sustained test.
Any drive Windows mounts, including NVMe, SATA SSD, USB flash, memory cards and spinning disks. External drives work over USB as long as the connection is stable during the run.
Download DiskSpeedTest, pick a profile and get four numbers that mean something.