No independent stock-versus-DCENT_OS benchmark dataset is published yet. This protocol defines what a result must contain before D-Central will present it as performance evidence.
Required test record
Minimum evidence required for a reproducible DCENT_OS benchmark
| Category | Required fields | Why it matters |
|---|
| Hardware identity | Exact miner SKU, hash-board count and revisions, ASIC, control-board family/revision, PSU and cooling | Model names do not prove board-lane compatibility |
|---|
| Software identity | Baseline firmware/version, DCENT_OS commit, artifact filename, manifest digest and configuration | A result must resolve to code and a named build |
|---|
| Environment | Input voltage, wall-meter model, intake temperature/humidity, altitude and acoustic method | Power, thermals and noise are environment-sensitive |
|---|
| Mining outcome | Pool, pool-side hashrate, accepted/rejected/stale shares, achieved difficulty and timestamps | Connected or hashing is not accepted-share evidence |
|---|
| Telemetry | Wall watts, TH/s, J/TH, chip/board temperatures, fan RPM, throttling and error events | One headline number cannot describe stability or safety |
|---|
| Duration | Warm-up plus 1-hour, 24-hour and 72-hour checkpoints, with any reset or configuration change recorded | Short runs hide thermal and failover problems |
|---|
Raw-data format
Publish the unedited measurements as CSV alongside a human-readable report. The machine-readable benchmark field template is public; it is a schema, not a results dataset. No empty template is represented as performance evidence.
timestamp_utc,test_id,firmware,build_commit,artifact_sha256,miner_model,hashboard_revision,control_board,psu,input_v,ambient_c,wall_w,hashrate_ths,j_th,chip_temp_c,fan_rpm,pool_side_ths,accepted,rejected,stale,uptime_s,throttle_events,reset_events,notes
Comparison procedure
- Photograph and record the hardware identity before changing firmware.
- Measure the baseline firmware under a fixed power and environmental setup.
- Back up the unit and validate the recovery route before testing DCENT_OS.
- Use the exact same pool, worker, voltage source, meter and cooling arrangement.
- Publish failed tests and missing observations, not only favorable runs.
- Attach pool-side proof, raw CSV, configuration, artifact digest and test duration.
Independent reproduction
Independent miners and security researchers are invited to reproduce or falsify a result. Use the repository's platform bring-up report and link the raw dataset. D-Central will label first-party, community and independently reproduced measurements separately.
Protocol published 2026-08-08. No performance claim is implied by publishing a methodology.
