Options can be passed as command-line flags (--flag value) or as environment variables (FORGE_FLAG) — handy for HiveOS and .bat files.
--algorithm FORGE_ALGO pearlhash, xelis, kawpow (RVN/QUAI/XNA),
cfx (aka conflux), quantus, iron
--coin FORGE_COIN KawPow coin: rvn, quai or xna (auto-detected
from the pool host if omitted; set it
explicitly for Neurai/Vipor pools)
--pool FORGE_POOL pool host:port; TLS is detected on its own (prefix
ssl:// to force it); comma-separate several pools
for fail-over
--wallet FORGE_WALLET your payout wallet address
--worker FORGE_WORKER worker / rig name shown on the pool
--password FORGE_PASS pool password (usually x)
--gpu FORGE_GPU mine only these GPU indices, e.g. --gpu 0,1,2,6
(default: all GPUs; NVIDIA first in nvidia-smi
order, then AMD in PCI bus order)
FORGE_AMD 1 = mine on the AMD cards only, 0 = on the
NVIDIA cards only (default: both)
--kernel FORGE_KERNEL Pearl on RTX 50: 1 (default) is the fastest, 2 draws
about 2% less power for ~1% hashrate; other cards ignore it
--proxy FORGE_PROXY route every pool connection, dev fee included,
through socks5://[user:pass@]host:port
or http://[user:pass@]host:port
--tls FORGE_TLS force TLS on (--tls) or off (--tls false) for
every pool; usually not needed
FORGE_LOWVRAM low-VRAM mode for 8 GB cards on Pearl (auto by default)
Temperature protection:
--temp-limit FORGE_TEMP_LIMIT pause a card once its core reaches this temperature, °C --temp-resume FORGE_TEMP_RESUME resume a paused card once its core cools to this temperature, °C (default: limit − 8) --mem-temp-limit FORGE_MEM_TEMP_LIMIT the same for the memory temperature, on cards that report it (Windows; GDDR6X/GDDR7) --mem-temp-resume FORGE_MEM_TEMP_RESUME resume once the memory cools to this temperature, °C (default: limit − 8)
Overclocking and fans (each flag takes one value applied to all GPUs, or a comma list mapped to --gpu; requires root/Administrator):
--cclk FORGE_CCLK lock GPU core clock, MHz
--coff FORGE_COFF core clock offset, MHz (+/-)
--mclk FORGE_MCLK lock memory clock, MHz
810 on GTX 10/16, RTX 20 and CMP holds the memory at 810 MHz;
405 also on RTX 20–50 (Linux, driver 595+); 5000/810/405 on CMP 50HX/70HX/90HX
--moff FORGE_MOFF memory clock offset, MHz (+/-)
--plimit FORGE_PLIMIT power limit, watts (Windows with AMD: % of the card's limit)
--vddoff - core voltage offset, mV (+/-): AMD RX 6000/7000/9000, NVIDIA RTX 20–50 on Linux
--vddc - core voltage, mV: AMD RX 5000/6000, NVIDIA RTX 20–50 on Linux
--voltage-info show the voltage range a card offers (with --gpu LIST)
--voltage-recover put a card's voltage back after an interrupted run
--fan FORGE_FAN fixed fan speed, %
--fan-curve FORGE_FANCURVE temp:percent points, e.g. 45:30,60:55,70:75,80:100 (NVIDIA; on AMD use --fan); one curve per card, split by /
--oc-delay FORGE_OC_DELAY apply the overclock N seconds after startup
--oc-keep FORGE_OC_KEEP=1 leave the overclock on when the miner exits
--oc-reset - reset the cards to stock and exit (e.g. after a crash)
Example — per-GPU overclock on a mixed rig:
--gpu 0,1,2,6 --coff 300,250,300,200 --plimit 280,280,300,260
Put * in a list to leave that card untouched — here GPU 2 keeps its own clocks and automatic fan:
--gpu 0,1,2,6 --cclk 1500,1500,*,1500 --fan 70,70,*,70
Monitoring API, off by default and read-only: web dashboard at /, JSON at /summary, Prometheus at /metrics. On HiveOS set FORGE_API=127.0.0.1:7777 in Extra config.
--api - dashboard and API on 127.0.0.1:7777 (this machine only) --api-bind FORGE_API the same on the given address, e.g. 0.0.0.0:7777 for the LAN
Logging and display:
--log FORGE_LOG also write the output to a file (appends) --no-color, --plain FORGE_NO_COLOR ASCII output without colours (automatic without a console) --v1-tui - start in the classic scrolling view (L switches views)
CMP hardware unlock (Linux; CMP 40HX, 50HX, 70HX and 90HX; needs NVIDIA driver 610.43.03; run with sudo):
--cmp-install install the unlock helper and reboot (10 s to cancel) --no-reboot with --cmp-install: install only, reboot yourself --cmp-verify show the unlock status of each CMP card --cmp-rollback restore the previous driver from backup and reboot
--cmp-install can stay in the flight sheet for good: if the unlock is already active or the rig has no CMP cards, the miner prints one line and starts mining.
