← All guides

How to read a mining dashboard: hashrate, shares, and rejected work explained

Monitoring · 7 min read

General information only, not financial advice. Healthy ranges below are the ranges published by the dashboard and troubleshooting guides cited in the text, checked 25 October 2026; firmware versions differ in how fields are labelled. Nothing here predicts mining earnings.

Start with one question: is work actually reaching the pool, right now? Not the hashrate, that number can look healthy on a miner that has stopped producing anything useful. The first field to read, on any dashboard, is the time since the last accepted share. Everything else on the page is commentary until that one checks out. The reading order below works the same way on an Antminer status page, on the AxeOS dashboard that runs a Bitaxe, and on the pool’s own worker page: liveness, then agreement, then error rates.

First: liveness, or the LSTime test

On the Antminer status page the field is labelled LSTime, the time since the last share. Counted in seconds, it is normal background. Climbing from seconds into minutes means something specific and bleak: the miner has effectively stopped producing valid work, even if the screen still shows a hashrate. The hashrate is what the chips report doing. LSTime is what the pool has actually received. When they contradict each other, believe LSTime.

The companion field is accepted work. DiffA on an Antminer page. It should climb steadily. A miner whose accepted count is flat while its fans spin is consuming power and returning nothing, and no other metric on the page changes that verdict. Check these two fields first, every time, before reading anything else.

Second: local hashrate versus pool hashrate

Two hashrate numbers exist because two machines measure different things. The miner reports its own rate directly. The pool estimates a rate from the shares it receives, and that estimate lags the device by 10 to 15 minutes, so a pool graph checked two minutes after switch-on is a picture of the past, not a fault.

Real-time hashrate also fluctuates normally, by plus or minus 10 to 15 percent on a Bitaxe-class dashboard. The session average is the number that matters for comparisons; the live figure is weather, not climate. The sourced agreement test is this: over an hour, device and pool hashrate should agree within roughly 10 percent. Inside that band, the difference is estimation noise. A pool reading much lower than the device’s average over that hour means shares are being lost in transit, a network problem, not a chip problem.

Third: four verdicts on submitted work

Shares come back with four different labels, and they are not four spellings of failure.

FieldWhat it meansHealthy readingWhen it rises
Accepted (DiffA)Work the pool countedClimbing steadilyNormal operation, this is the point of the exercise
Rejected (DiffR)Work submitted and refused between miner and pool0–1% in one guide’s range; below 1–2% in another’sUnstable overclock, high latency, or a flaky pool endpoint; on a Bitaxe, also overclocking or WiFi instability
Stale (DiffS)Work submitted for a block the network had already moved pastOccasional; a latency symptom when frequentSlow connection path between miner and pool
DiscardedWork dropped because a new block arrivedNormal churn, not an errorRises with block arrivals; no action implied

Two distinctions carry most of the diagnostic weight. Discarded is not rejected: work discarded on a new block is the system clearing its desk, and counting it as failure manufactures a problem that does not exist. And stale is a timing diagnosis, not a quality one, the work was valid, it simply arrived for a block that no longer existed, which points at latency rather than at the machine.

The rejected-share ranges deserve a note on sources, because two guides publish two ranges: 0–1 percent and below 1–2 percent. Both are printed in the table rather than blended. The practical reading rule that follows from the pair: under 1 percent sits inside every published range, 1–2 percent sits inside one and outside the other, worth watching, not yet worth a teardown, and above 2 percent is outside both, which is where investigation starts.

The error rate that never leaves the chip

One percentage on an AxeOS-style dashboard is routinely confused with rejected shares, and the two measure different journeys. Hashrate error percentage happens inside the ASIC chip itself, before any share exists. Rejected shares happen afterwards, between miner and pool. A high error percentage with a low rejection rate localises the problem to the device. A low error percentage with a high rejection rate localises it to the path. Same symptom family, opposite ends of the cable.

The published healthy range for hashrate error is 0–2 percent, alongside the 0–1 percent rejected-share range from the same guide. Nearby sits one more field beginners over-read: Best Difficulty, the highest-difficulty share found in the session. It is a trophy statistic. It does not affect performance, and chasing it changes nothing.

When the two dashboards disagree

PatternMost likely readingFirst move
Device hashrate normal, LSTime climbing into minutesMiner effectively stopped producing valid work despite the displayTreat as stopped: check the pool connection before touching settings
Device average normal, pool rate much lower over an hourShares lost in transit, network issueConnection path and pool endpoint, not the miner’s tuning
Pool rate low in the first 15 minutes onlyNormal pool-side lag, the estimate is 10–15 minutes behindWait, then re-check against the one-hour agreement test
Rejected ratio risingUnstable overclock, high latency, or flaky pool endpointBack off the overclock first; it is the cause under the operator’s control
Stale shares frequent, rejected shares normalLatency: valid work arriving after its block passedShorten or stabilise the connection path
Discarded count rising, everything else steadyNormal churn from new blocks arrivingNone, record it and move on

A five-minute check, repeated weekly

  1. Read LSTime (or the pool page’s last-share time) first. Seconds: proceed. Minutes: stop and treat the miner as down.
  2. Confirm accepted work. DiffA or its equivalent, is climbing.
  3. Compare the device’s session-average hashrate with the pool’s figure over the last hour. Note the percentage gap.
  4. Read the rejected rate and the stale count, and place each against the ranges in the table above.
  5. On dashboards that show it, read the hashrate error percentage and note whether it or the rejection rate is the elevated one, that locates the problem.
  6. Write the five numbers down with the date. A single reading diagnoses; a dated series proves.

What success looks like: LSTime counted in seconds, accepted work climbing, device and pool within roughly 10 percent over the hour, rejected shares under 1 percent, hashrate error inside 0–2 percent, and a dated log line that lets next week’s check be a comparison instead of a fresh guess.

The dashboards reward sequence over scrutiny. Read in the order above, liveness, agreement, errors, and most visits end at step two with everything inside its range. Read in the order the screen happens to present, hashrate first and largest, and the one failure that matters, a miner quietly producing nothing, hides behind the most reassuring number on the page.

, Dex Carter, Tech Writer