Miner Burn-In Testing & Grading

Stage three: prove it. Hashrate, wall power, board temperatures and error rate are recorded over a sustained loaded run — then the unit is graded on those numbers. Part of our used miner services.

24 h

Full-load burn-in on repaired boards

5 metrics

Hashrate, power, temps, errors, stability

A / B / C

Grade tied to measured behaviour

Per serial

Test record goes with the machine

What the test actually records

MetricHow it is checkedWhy it matters
HashratePer board, against the model specification, over the runA board that reports chips but underperforms shows up here, not on a photo
Wall powerMeasured draw under load, compared with specFlags efficiency loss and PSU problems before the machine ships
TemperaturesChip and board sensors tracked through the runCatches weak thermal work, failing fans and hot domains
Error rateHardware error counters watched for creepA machine that "works" but errors is worth less than a clean one
StabilityRestarts, board drop-outs and fan faults during the runThe difference between a bench pass and a machine you can leave running

From measurements to a grade

The grade is a short label on top of a long record. What a buyer or a consignor actually receives is the record — the label just makes it comparable across a batch.

Grade A

At or near rated hashrate, temperatures stable through the run, no error-rate creep, all boards healthy. Cleaned and thermal refreshed.

Grade B

Inside a documented margin of spec after repair, stable through the run, cosmetic marks present. The fault history is included with the unit.

Grade C

Known derating or one unstable board, disclosed and priced accordingly. Parts, benches and short-horizon mining — not long-warranty stock.

If a unit cannot reach a sellable grade it is repaired again or sold as parts. It is never relabelled to make a batch look better.

Why test before selling

A used miner bought on a photo is a guess about three things: whether the chips are all alive, whether the thermal path still works, and whether it will hold up after the trip. A burn-in run answers all three before money changes hands.

For buyers

You see the numbers before you commit, instead of after the machine is racked and hashrate never arrives.

For consignors

Your batch is graded honestly, which protects the price you get — and tells you what is worth repairing first.

For farms

Replacement units arrive ready to rack, with a record you can compare against the machines already running.

For the bench

Test results feed back into repair: a repeated temperature spike on one domain is a fault we can still fix.

Testing questions

How long is the burn-in?
About 24 hours of full-load running for repaired boards, and a sustained loaded run for whole machines. A boot-and-show-a-number check is not a burn-in and we do not call it one.
What is measured?
Hashrate per board against spec, wall power under load, chip and board temperatures over time, hardware error rate, and stability (restarts, board drop-outs, fan faults).
Do I get the test data?
Yes — tied to the serial number, together with the fault history if the unit was repaired. Consignors get the same record for their own machines.
Can a batch be tested faster?
Testing can be split into screening and full burn-in when a deadline matters, but the grade always states which level of test was passed.

Want the numbers, not the adjectives?

Tell us the model and how many units you need tested — you get a test plan and a per-unit record.