Eugen Standard
Standardized independent tests. No marketing.
English

External SSD — Full Drive methodology

Current version: full_drive_v1 · 25 September 2026

The primary methodology contains two tests: Fresh Sequential Write — Full Drive and Sequential Read — Full Drive. There is no fixed transfer size or duration. The test extent comes from the selected SSD, whether its nominal capacity is 500 GB, 1 TB, 2 TB, 4 TB, 8 TB or larger.

Preparation and repeatability

Before every write attempt, including a retry after interruption or failure, SSD Studio verifies the selected device, quick-formats the volume as NTFS, and performs mandatory TRIM (ReTrim). The write begins only after ReTrim completes successfully. If ReTrim is unsupported or fails, the attempt is blocked.

Four pauses are configured independently: before the first write in the block, between writes, before the first read in the block, and between reads. A zero first-write pause starts that write after mandatory preparation without the configured interval between writes. Write pauses follow NTFS formatting and ReTrim; they never replace ReTrim. On a resumed run, the first-pause setting applies to the first scheduled attempt of each block. A timed pause does not prove that the SSD has reached a particular temperature. During each pause the SSD remains powered and connected, without other workloads. All four settings and actual pauses, model, measured usable capacity, firmware, operating system, fio version, test date, NTFS file system and ReTrim result are recorded. Results with different pause settings retain distinct comparison contexts. Test logs and results are stored outside the volume being filled.

1. Fresh Sequential Write — Full Drive

fio writes sequentially into the test file until the file system has no further writable user space. The runner determines available space after preparation; no capacity is entered into the profile. The workload uses one job, 1 MiB blocks, queue depth 1, direct I/O and the Windows asynchronous I/O engine.

“Full Drive” refers to the available user space on the selected, freshly prepared volume. File-system metadata and allocation alignment are not test data. Nominal capacity on the box is not substituted for the measured extent.

Completion is verified against fio’s actual completed byte count, the resulting file and remaining usable space. A cancelled run, quota limit, disconnected device or I/O failure is not a successful full write. Progress and remaining time are estimates until completion is validated; a speed decrease after the SLC cache changes the time estimate.

Published values include average speed, sustained speed, SLC speed/size/duration when observed, actual written GiB and total elapsed seconds. No unobserved cache phase is invented.

2. Sequential Read — Full Drive

After all write attempts finish, the final planned successful write supplies the exact file and extent for every read attempt. Each read uses the same unchanged file and reads all its data from beginning to end in one sequential pass, using the same block size and queue depth. The source must already exist: a read does not create or extend a file. Reading fewer bytes than were written is not a successful full read. If the final write fails, the read phase cannot fall back to an earlier file that was overwritten during preparation.

There is no formatting or TRIM after the final write or between reads. Published values are average read speed, actual read GiB and test duration; idle pauses are not included in the measured fio duration.

Attempts and results

The normal result set contains three valid writes followed by three valid reads:

  1. NTFS + ReTrim → first-write pause → Write 1.
  2. NTFS + ReTrim → between-write pause → Write 2.
  3. NTFS + ReTrim → between-write pause → Write 3.
  4. First-read pause → Read 1 → between-read pause → Read 2 → between-read pause → Read 3, all using the file created by Write 3.

The primary table reports each test’s median of three selected attempts and preserves individual measurements, raw logs, failures and provenance. The read sequence is recorded as writes_then_reads_v1; earlier immediate reads are not combined with it. A previously completed valid write can be retained when the remaining write attempts are resumed, because its preparation and write procedure are unchanged.

Calculator

The current calculator accepts only full_drive_v1 measurements. Full Drive shows each SSD’s measured full-run duration and actual data volume; different capacities therefore represent different workloads. Smaller transfers are estimates within the measured range. Write estimates use the observed cache shape calibrated to the measured full-write size and duration; read estimates use the measured read size and duration. Extrapolation beyond measured data is disabled.

Before real Full Drive retests are published, the calculator shows an empty state.