Guide
The web dashboard and hardware monitoring
Every sensor in the rig in a browser, charts, LAN access, the farm view, and the JSON API.
bzminer serves its own dashboard from inside the miner. There is nothing to install, no account, and nothing leaves the machine: the browser talks to the miner on port 4014, and the miner pushes it the same data the console shows. This guide covers what is on it, how to reach it from another machine, how to run it as a pure hardware monitor, and the JSON behind it.
Opening it
Start mining, then in any browser on the same machine:
http://127.0.0.1:4014/The Connect to rig button at the top of this site does the same thing for a rig on your network — type its address and port and it opens that rig’s own dashboard in a new tab. Nothing is sent to bzminer.com.
Three settings control the server, from the command line or config.txt:
--set http_port=8080 # a different port
--set http_address=0.0.0.0 # reachable from the rest of the LAN
--set http_enabled=false # offThere is no password on the dashboard. It binds to 127.0.0.1 by default, which means only the machine itself can see it. 0.0.0.0 opens it to anything that can reach the machine, including the overclock controls. Do that only on a network you trust, or behind something that asks for a login.
What is on it
Two tabs, plus a poll-interval selector (0.5s / 1s / 2s / 5s / off) and a live indicator.
Monitoring
- A card per GPU: usage, core, memory and hotspot temperatures, power, fan, clocks against their rated values, PCIe generation and width with TX/RX, and GDDR timings.
- Advanced sections for fan RPM, rail power, current and voltage, engine utilisation, PCIe errors and performance limits. Show all metrics reveals every vendor sensor as a tile — NVIDIA publishes well over a hundred on a modern card.
- An lstopo-style CPU map: packages, caches, cores and threads, with a Utilisation / Temperature toggle and a Boxes / Tree layout toggle.
- Per-DIMM RAM cards with DRAM timings and a temperature graph, and storage volumes.
- Click any metric to chart it over 10 minutes, 1 hour or 1 day. Histories are collected even while a tile is hidden, so opening one shows the past hour, not a blank.
Mining
- Per-device hashrate, accepted, rejected and pending shares, temperatures and power.
- The pool and rig summary, pushed live over a WebSocket rather than polled.
- On a farm proxy, every rig mining through it is a light-blue row named by its worker; click it to open that rig’s own devices.
Tune
Each GPU card has a Tune button when the miner can control that card. The panel shows only the knobs the card supports, with the ranges the driver reported, and every change goes through the same code the console and command line use — a change made in the browser appears in the terminal’s log, and one typed there shows up here on the next poll. The overclocking guide explains the knobs.
A hardware monitor that does not mine
Run bzminer with --dmon and it mines nothing at all: every sensor it can read, on
the console and in the dashboard, and nothing hashing. No wallet, no pool, no algorithm.
./bzminer --dmon
./bzminer --dmon --metrics temps,powers,clocks # choose the console columns
./bzminer --metrics ? # list every group and sensor on this machine
The same thing happens on its own if a run has no pool configured, or "pool": []
in config.txt — it drops into monitoring mode rather than exiting, and says so.
The shipped config.txt starts that way on purpose, so a fresh install does not hash
to a placeholder wallet.
Useful for a machine that is busy with something else, for checking a card’s sensors before committing it to a job, or simply as a permanent readout. One-shot versions, each of which prints and exits:
./bzminer --gpu-info # NVIDIA devices as CUDA sees them
./bzminer --cpu-info # CPU topology and live metrics
./bzminer --ram-info # memory modules, DRAM timings, SMBus temperatures
./bzminer --list-metrics # every sensor this machine exposesThe farm view
Every bzminer announces itself on the local network, so any one of them can show you the whole farm. Open the dashboard on any rig and the Farm tab lists every instance it can see, with combined hashrate, power and share counts; a click on a row opens that rig’s dashboard.
Nothing to configure. Each instance broadcasts a small summary on UDP port 4014 every five seconds; a rig that goes quiet for sixteen seconds drops off. The announce carries the rig name, web port, version, hashrate, power, shares and uptime — no wallet, no pool, no credentials.
- To follow the links, the other rigs need
http_address=0.0.0.0. A rig on the default setting still appears in the list and the totals, but is shown as local only. --disable_udpannounces nothing and listens for nothing; use it on a network you do not control. It is also aconfig.txtsetting ("disable_udp": true) for a mining OS that never assembles a command line.
JSON, for your own dashboard
Everything the dashboard shows is available as JSON on the same port. GET /help documents every endpoint, its units, and the placeholders.
| Endpoint | What it returns |
|---|---|
/api/snapshot | Everything: devices, pools, hashrate, shares |
/api/stream | The same, pushed as it changes (WebSocket) |
/api/gpus | Per-GPU detail and every vendor sensor |
/api/metrics | Metric histories for charting |
/api/topology, /api/ram, /api/storage | CPU topology, DIMMs, volumes |
/api/peers | The farm view, as the Farm tab renders it |
/api/oc | GET reads overclock state and per-knob support; POST applies or resets one |
/status | The full settings-and-stats document other mining dashboards expect, bzminer 1.x shape, hashrates in H/s |
/hive_status | Hive OS’s stats format, hashrates in kH/s |
Anything that changes the machine is a POST, never a GET, because a mutating GET gets fired by a browser prefetch or a plain refresh. A single overclock knob can be sent on its own; absent keys are left alone:
POST /api/oc {"pci_id":"0000:01:00","power_limit_w":320,"fan_pct":70}
POST /api/oc {"pci_id":"0000:01:00","reset":true}
The reply’s status is all, partial or none
and the message names what landed and why the rest did not — “needs
root/administrator” is the usual answer on Linux.
Problems
- Nothing at 127.0.0.1:4014. The port is taken or the server is off. The startup log says which port it bound, or that
http_enabledis false. - It is reachable on the rig but not from another PC.
http_addressis still127.0.0.1; set it to0.0.0.0. - The Tune button is missing on a card. The miner cannot control that device: on AMD only the fan is settable, on Intel nothing is, and on Linux every control needs root.
- Farm tab shows only this rig. UDP 4014 is blocked between the rigs, or
disable_udpis set.