Bitaxe Mining Log: What AxeOS Records, and What It Cannot

By Dominik Lederer, Founder, Hardware Engineer and Solo Mining Expert at Polarblocks. Network figures read from mempool.space on 14 September 2026 at 06:09 UTC. The exchange rate carries its own timestamp of 06:05 UTC.

The first question people ask me after the second week is never about hashrate. It is about the log. Line after line of accepted shares scrolls past, the numbers look enormous, and nothing happens. So they want to know whether the machine is working or quietly broken.

It is working. What confuses people is that a Bitaxe mining log is the most active looking record of total failure you will ever read. Every line in it is a genuine, verified piece of work, and none of it adds up to anything until one line does.

Mine sits on a shelf two metres from my desk. I have the log stream open in a browser tab most mornings while the coffee is brewing, and I have read tens of thousands of lines that meant nothing. That habit is the actual product here, and it is worth understanding before you buy.

What does a Bitaxe mining log actually contain?

A Bitaxe mining log holds four kinds of line. Work the device sent to the pool and the pool accepted, work the pool rejected, housekeeping messages from the network connection, and sensor readings such as chip temperature and fan speed. Alongside that stream AxeOS reports one standing figure, the best difficulty the device has ever reached.

Nothing in there is a balance. The log is closer to a flight recorder than to a bank statement, and that distinction is the single most useful thing to understand about it. A share entry says the chip did its job correctly. It says nothing whatsoever about being closer to a block than yesterday.

The AxeOS firmware behind it is open source, maintained by the OSMU community as ESP-Miner under the GPL version 3. The Gamma board that runs it is published separately under the CERN Open Hardware Licence, strongly reciprocal variant. Two different licences for two different layers, which is unusual enough in consumer hardware to be worth saying out loud.

Where does AxeOS store the log, and how do you read it?

AxeOS serves the log from the device itself, through a small web server the firmware hosts on port 80. The readme documents a websocket at /api/ws for the live text stream, a file download at /api/system/logs, and a separate stream at /api/ws/live for partial system updates. The dashboard at the device address in a browser shows the same material.

Three further endpoints matter for anyone who wants the numbers rather than the scroll. /api/system/scoreboard returns the twenty highest difficulty shares, and /api/system/statistics returns system statistics, with the readme noting that data logging has to be switched on for it. There is also /api/system/info for the current state.

The device is reachable by name as well as by address. ESP-Miner registers itself over mDNS with a default hostname of bitaxe, so http://bitaxe.local usually works, and it advertises itself as an AxeOS device for discovery tools. Current stable firmware is version 2.15.1, released on 29 August 2026.

Why does a log full of accepted shares still show no progress?

Because a share is proof of work done, not a step towards a block. Solo CKPool states a minimum difficulty of 10,000 and says plainly that shares there are cosmetic for feedback only. Our own calculation puts a 1.3 terahash device above that bar about every 33 seconds, so roughly 2,615 accepted shares a day.

Those 2,615 daily lines are real. They are also 12.7 billion times too easy to matter, because the only share that pays is one that beats the full network difficulty of 127,450,789,715,843. Nothing carries over between attempts. A device switched on this morning stands exactly where mine stands after more than a year.

This next part is worth knowing before you buy, and it comes from the service itself rather than from us. Solo CKPool describes the most satisfying part of the log as decoration. It also warns that hashrates below 100 gigahash are unrealistic and suggests treating such devices as a learning exercise. A Bitaxe Gamma 601 sits well above that line at 1.3 terahash per second, but the honesty of the framing is the point.

One figure deserves flagging, because two sources disagree. Polarblocks rates the Gamma 601 at 1.3 terahash per second and this article uses that number throughout. The upstream bitaxeGamma repository states about 1.2 terahash per second for the reference design. The gap moves the annual odds from about 1 in 13,352 to about 1 in 14,465, so here are both.

What does the best difficulty entry in the log mean?

It records the hardest single share the device has ever produced, and nothing more. Think of it as a personal record in a sport with one prize, awarded only for beating the world record. AxeOS also exposes the twenty highest such shares at /api/system/scoreboard, which makes the record easy to check without the dashboard.

People read that number as a distance travelled. It is not. If your best difficulty stands at, say, one hundred million, you have not covered a portion of the gap to 127,450,789,715,843, because there is no gap to cover. The next attempt starts from the same place as the first one did.

What the figure is genuinely good for is confidence that the chip and the pool connection are healthy. A best difficulty that grows over weeks tells you shares are being built, submitted and accepted. That is a diagnostic signal, and treating it as one makes the log far more useful than treating it as a scoreboard.

How does a year of log entries compare with a lottery ticket and an industrial miner?

A year of log entries is 52,560 block rounds, each one a fresh independent draw at a full block of 3.125 BTC plus transaction fees. That is a design target from the ten-minute block interval rather than a measurement. The comparison that matters is not the odds column but what each year costs to hold.

An Antminer S21 Pro reaches roughly 1 in 74 a year at 234 terahash per second, and draws 3,510 watts. At 230 volts that is about 15.3 amperes of continuous load, just under the 16 ampere rating of a German household circuit and not a load such a circuit is meant to carry around the clock. ASIC Miner Value lists the machine at 75 decibels. One lottery line played on both weekly draws costs 124.80 euros of stake a year plus a processing fee and leaves no log and no object behind, against 52.12 euros of electricity for the Bitaxe and 181.12 euros across its first year with the device counted.

Criterion	Bitaxe Gamma 601	Lottery line, two draws a week	Antminer S21 Pro
Winning chances per year	52,560 rounds a year, every one at a full block of 3.125 BTC plus transaction fees, minus the 2 percent service fee, annual odds about 1 in 13,352 at a difficulty of 127,450,789,715,843	104 draws a year on one line, annual jackpot odds about 1 in 1,344,598, our own division of the published 1 in 139,838,160 per draw	Annual odds of about 1 in 74 on a full block, reachable only under industrial power and cooling conditions
Running costs in a home setting	About 52.12 euros of electricity a year at 0.35 euros per kilowatt hour, roughly 4.34 euros a month from an ordinary wall socket, and separately a one-off 129 euros for the device, which makes 181.12 euros across the first year	124.80 euros of stake a year on one line plus a processing fee per ticket, due again in full every year with nothing carried over	About 10,762 euros of electricity a year at 3,510 watts and the same tariff, roughly 15.3 amperes of continuous load at 230 volts against the 16 ampere rating of a household circuit, not viable in a flat
Drawer risk (boredom factor)	Low, because at the Solo CKPool minimum difficulty of 10,000 the log produces a readable event roughly every 33 seconds and the best difficulty record visibly moves, though the service itself calls those shares cosmetic	High, because after the draw there is no record to read and no object to look at	Very high indoors, because 75 decibels of constant noise pushes the machine into a basement or a garage
Learning effect & blockchain fascination you can touch	High, because share difficulty, network difficulty, retargets and the block reward all become observable in your own log instead of staying abstract	None, because picking six numbers teaches nothing that transfers to anything else	Present, but noise, waste heat and power draw make the machine impractical to live with day-to-day

Which log entries point at a real fault?

Three patterns are worth investigating: rejected shares in bulk, a connection that reconnects in a loop, or a stream that stops entirely. A long run of accepted shares with no block is not a fault, and neither is a best difficulty figure that sits still for days, because both are the expected behaviour of a device this size.

One cause catches people out often enough that the ESP-Miner readme names it directly: some Wi-Fi routers block mining. The readme lists ASUS routers and some TP-Link models and says to disable AiProtection and IoT, adding that a router offering both options might need both switched off. If the log shows a connection that never settles, check the router before you suspect the board.

There is a second trap in the same document, and it is a hardware one. Only the ESP32-S3-WROOM-1 module in the N16R8 variant is supported, so modules without the right memory configuration will not run the normal firmware. And the settings page hides the frequency and voltage fields behind a URL suffix on purpose, with the readme warning that overclocking without additional cooling can overheat or damage the device.

One source conflict belongs in the open, since both pages are easy to check. The pool configuration example in the ESP-Miner readme uses solo.ckpool.org as the stratum address, while Solo CKPool's own configuration section names stratum.ckpool.org as the host, on the same port 3333. The operator's instruction is the one to follow.

What does a year of log entries cost?

About 52.12 euros of electricity. A Bitaxe Gamma 601 draws roughly 17 watts, which is 148.92 kilowatt hours across 8,760 hours, at a house assumption of 0.35 euros per kilowatt hour. At the BDEW household average of 37.0 cents the same year costs 55.10 euros, or 4.59 euros a month.

Those 17 watts are a rounded input and the cents that follow are arithmetic rather than measurement. Product data give a range of 15 to 19 watts, so the honest band is 45.99 to 58.25 euros a year. The same caution applies to the hashrate: 1.2 terahash per second gives about 1 in 14,465, and 1.4, which the Eco Line reaches only overclocked, with better cooling and therefore outside that watt band, gives about 1 in 12,398.

Against that sits the expected value. One block on 14 September 2026 carried 3.125 BTC plus transaction fees, and at 67,058 euros per bitcoin the bare reward alone was worth 209,556.25 euros, exchange rate dependent and excluding fee income. Spread across 13,352 years of mean search time that is about 15.69 euros a year, or about 15.48 euros once average fee income is added and the 2 percent Solo CKPool fee is taken off.

The figure moves both ways, and not symmetrically. Fees push it up: across 807 data points in the week to 14 September 2026 the average block carried 0.01974 BTC in fees, with a median of only 0.01576, which adds roughly ten cents a year. The retarget due around 19 September 2026 pushes it down, and the projection itself drifts: mempool.space put the change at plus 4.41 percent at 06:09 UTC and plus 4.53 percent at 06:33 UTC, with 750 blocks still to run. Plus 4.41 percent costs 4.22 percent of the expected value, about 66 cents a year, and plus 4.53 percent costs 4.33 percent, about 68 cents. The downward effect is therefore roughly seven times the upward one, 6.7 at the lower projection and 6.9 at the higher.

As a ratio, using the pre-fee expected value of 15.69 euros, the return is 30.1 percent measured against electricity at 0.35 euros per kilowatt hour, 28.5 percent at 0.37 euros, and 8.7 percent measured against a full first year of 181.12 euros including the 129 euro device. Using the post-fee 15.48 euros those three figures become 29.7, 28.1 and 8.5 percent. Every one of them names its own denominator, which is the only way such a percentage means anything.

Electricity alone would break even at about 10.54 cents per kilowatt hour on the bare reward, or about 10.39 cents on the post-fee figure. That is a statement about the power bill and nothing else. Once the purchase price is counted there is no positive break-even in year one at any German household tariff. The full arithmetic sits in our profitability breakdown, and the mechanics behind the log are covered in how a Bitaxe works.

A Bitaxe mining log is the most honest document in home mining, because it never pretends to be an account balance. At the pool minimum difficulty of 10,000 it logs about 2,615 accepted shares a day, all of them real work and none of them progress, and anyone who sells that scroll as a growing position is selling a feeling rather than a machine.

Dominik Lederer, Founder Polarblocks

Frequently asked questions

How do I download the mining log from a Bitaxe?
The AxeOS firmware serves a log download at /api/system/logs on the device itself, and a live text stream at the websocket address /api/ws. Both are documented in the ESP-Miner readme, alongside /api/system/scoreboard for the twenty highest difficulty shares.

Why does my Bitaxe mining log show accepted shares but no blocks?
Because a share only proves the chip did valid work at the pool's difficulty. Solo CKPool sets a minimum difficulty of 10,000 and states that shares there are cosmetic for feedback only. The only share that pays is one beating the network difficulty of 127,450,789,715,843.

What does best difficulty mean in a Bitaxe log?
Best difficulty is the hardest single share the device has ever produced. It is a health signal rather than stored progress, because every hashing attempt is independent. AxeOS also exposes the twenty highest difficulty shares at /api/system/scoreboard.

How many shares should a Bitaxe Gamma 601 log per day?
At the minimum difficulty of 10,000 that Solo CKPool states, a device running 1.3 terahash per second clears that bar roughly 2,615 times a day, so about one share every 33 seconds. That is a Polarblocks calculation, and pools raise the difficulty over time, which thins the stream.