DriveSpeedTest

An honest external-SSD benchmark that says whether the drive, cable or port is the bottleneck.

Role
Design and engineering
Areas
Tools, Apps
DriveSpeedTest diagnosis tab: a ranked verdict with confidence levels above a sustained-write chart that drops sharply once the SSD's cache fills.

Quick drive-speed tools often write compressible zeros, never flush to the device and read the data back from RAM, so a USB drive can appear faster than its own link allows. DriveSpeedTest measures the actual device instead, with unbuffered, write-through, sector-aligned I/O and incompressible data.

Six-step flow of a DriveSpeedTest run: refuse the system drive and size the test to free space, run sequential writes and reads with 1 MiB blocks, run random 4K writes and reads at queue depth one, write 4 or 16 GiB in flushed 64 MiB chunks to chart sustained speed, grade the claims and rank bottlenecks, and delete the test files.
Each sustained chunk is flushed before it's timed, so a late buffer drain can't hide the cache cliff.

What it checks

  • Throughput: sequential and random 4K reads and writes with the cache bypassed.
  • Sustained writes: several gigabytes charted over time to expose the point where an SSD's fast SLC cache fills and speed drops.
  • The connection: the negotiated USB link speed, found with the same device tree walk that USBView uses, plus host controllers, TRIM and filesystem. Because Windows sometimes reports a SuperSpeed device as USB 2.0, the tool infers the real tier from the device descriptor and confirms it against measured throughput.
  • SMART health, through smartctl when available and Windows reliability counters otherwise.
  • Marketing claims: paste the vendor's blurb and each claim is graded against what was measured, with a ranked list of likely bottlenecks and fixes.

A React dashboard streams each phase live over a WebSocket and exports JSON or text reports. Playwright drives a demo mode to regenerate the documentation screenshots. The app refuses the system drive by default and cleans up its test files.

Three-column comparison of DriveSpeedTest depths. Quick: 1 GiB sequential, 3 seconds per random test, no sustained write. Full: 2 GiB sequential, 4 seconds per random test, 4 GiB sustained write. Sustained: 1 GiB sequential, 3 seconds per random test, 16 GiB sustained write.
Only the sustained depth writes enough to get past most portable SSDs' fast cache.

It pairs with b2cFastCopy, which reacts to the same cache cliffs during a copy.

Bar chart of the usable throughput ceilings DriveSpeedTest assumes for each USB link: 450 MB/s at 5 Gbps, 1050 MB/s at 10 Gbps and 2100 MB/s at 20 Gbps.
If the link's ceiling is below the advertised speed, no drive can reach the claim over that connection, so the shortfall is pinned on the port or cable.