Guide

The console: screens, hotkeys and the command line

The four screens, what every column means, the hotkeys, and the commands behind the c key.

Four screens, all compiled in, all showing the same run. Press o to cycle through them, or start on one with -o <name>. This guide walks the screens, then the mining table column by column, then every key, then the commands behind c.

The four screens

ScreenWhat it shows
tuiThe boxed dashboard: a hardware box, an algorithm box per pool, a scrollable log pane. The default whenever there is a terminal.
logA plain scrolling log, with the device table reprinted every 30 seconds. The default when output is redirected — a full-screen dashboard sent to a file is just cursor codes, so a mining OS gets this without asking.
monitorOne dense sensor table, nvidia-smi dmon style, redrawn in place. The default in --dmon.
deviceA full-screen hardware inspector, one page per device: identity, firmware, board, PCIe link, memory type and lockable clocks, DRAM timings, every sensor. d from any screen, Esc back.
./bzminer -o monitor ...        # start on it
./bzminer -o log ...            # what a service manager or a log file wants
./bzminer -o device --device-select 1 --device-metrics all > rig.txt   # a full hardware report for one card

-o is short for --output. In bzminer 1.x it meant the pool; a URL given to it is still taken as a pool, with a warning, so an old start script keeps working.

How fast they refresh

Three knobs that are easy to confuse:

--interval 1000              # how often the DATA is sampled and published: every screen and the web UI read this
--tui-interval 400           # how often the tui REDRAWS; faster than the data changes just costs CPU
--log-table-interval 30000   # how often the log screen reprints the device table (0 = every update)
--hashrate-window 30         # seconds the reported hashrate averages over

So a screen that feels too fast wants --interval or the i / I keys, not --tui-interval. The hashrate window is a different kind of thing: a GPU’s hash counter advances one batch at a time, so an un-averaged rate would read 0, 0, 0, BURST, 0 — the window is what makes it a rate. Raise it for a calmer number, lower it to see a tuning change sooner. The avg figure is unaffected and is always since start.

Reading the mining table

The tui has two tables with two column settings, because they answer different questions:

--device-columns free,total,core,mem,fan,power,temp        # the HARDWARE box; same vocabulary as --metrics
--mining-columns id,cfg,tbs,shares,eff,poolhr,hr,status,poolinfo   # the ALGORITHM box
./bzminer --list-columns                                   # every name all three tables accept on THIS machine

Four columns are worth knowing by name:

  • cfg says how each device is configured, as opposed to what it is doing: i64 is a GPU at intensity 64, i0 a GPU on auto, t16 a CPU mining on 16 threads. The quickest way to confirm a setting actually took.
  • tbs is the measured average time between shares found — what a device delivers, against the est. tbs its difficulty predicts.
  • poolhr is the pool’s credited rate as a percentage of the miner’s own. 100% means the pool is crediting exactly what your rig reports; below means it is finding fewer shares than its hashrate implies; above means luck has run your way. It is the number to watch when a miner looks fast but pays badly. It swings over minutes and settles over hours.
  • poolinfo is the pool’s own column: height, difficulty, estimated time between shares and latency. Warthog leaves it out by default because its status column carries the rig summary and wants the width.

The log pane scrolls with the wheel, the arrows, PageUp/PageDown and Home/End. While it is scrolled up it holds still and the bar says held N up [End] live, so a frozen pane is never mistaken for a stalled miner. Selecting and copying text keeps working at the same time on every Linux and macOS terminal and on every modern Windows one; the old conhost window is the exception, so run bzminer under Windows Terminal there.

Every hotkey

The same on every screen. h prints this list in the miner, with the current setting beside each key that has one.

KeyWhat it does
hthis list (again to close)
copen the command line
onext screen
ddevice detail page
upool details: url, wallet, worker, password (again to close)
ppause / resume mining
qquit
+ / -more / less log detail, live
i / Irefresh faster / slower
t / TCPU mining threads, up / down
naim the tuning keys at the next card
g / Gcore clock offset, ±15 MHz
l / Lcore clock lock, ±15 MHz
m / Mmemory clock offset, ±50 MHz
k / Kmemory clock lock, through the card’s own steps
w / Wboard power cap, ±5% of the card’s range

u is the one to know when a pool is crediting you nothing. It prints every configured pool in failover order with its url, algorithm, wallet, worker name and password, read from the config rather than the snapshot — because the snapshot only carries the joined wallet.worker login, in which a wallet typo and a worker typo look identical. The password never leaves the console; the web host does not answer that field.

The tuning keys move one card, not the rig. n picks which, and its row is marked > in the tables and ▸ on the device page. Rig-wide is oc all core=+150 from the command line, where you have to type the word.

The command line

Press c, type, press Enter. Esc cancels.

CommandWhat it does
helplist the commands
pause / resumestop and restart mining without exiting
oc ...overclock — see the overclocking guide
intensity <n>mining intensity on every device; 0 is auto
level <name>log level: network, debug, info, warn, error
output <name>switch screen by name
quitexit

The device page

A selector table on top, a scrollable detail page below, for every device in the machine — GPUs that are mining, GPUs that are not, and the CPU. ← / → move between devices and aim the tuning keys at the one you are looking at; a toggles between the compact view and every available sensor; / filters rows by substring. It refreshes on its own timer and does not wait for mining snapshots, so a monitoring-only run keeps it ticking.

Also useful here: a memory-clock lock is only accepted at one of a card’s discrete states, so the memory group lists them — 405, 810, 5001, 10251, [10501] — with the current one in brackets. “Why is memory at 810” is the question that sends someone to this page.

Logs and log levels

--log-level info          # network | debug | info (default) | warn | error
-v, -vv                   # one step more verbose each;  -q, -qq quieter
--logfile rig1.log        # name/path of the file log (default bzminer.worker.log)
--logfile-level debug     # pin the FILE's level; empty = follow the screen, live
--logfile-mode timestamp  # append (default) | overwrite | timestamp, one file per run

There is always a file log, whichever screen you are watching, and it holds the same lines and the periodic tables. --log-level info --logfile-level debug is the useful pair: a readable screen and a file worth sending to support. --log-level network prints the raw stratum traffic, which is the fastest way to settle an argument with a pool about what was actually sent.