FearMinerGPU and CPU miner, with a free cockpit to run your whole farm from your phone or your PCNVIDIA (CUDA) · Apple silicon (Metal) · other GPUs (Vulkan) · CPU
Linux · Windows · macOS · HiveOS · mmpOS
Website | Download | Docs | Cockpit demo | Telegram
Mining, made simplerWe built FearMiner to take the chores out of mining. One line installs it and runs it as a service that starts with the machine. You choose what it mines: a wallet and a pool on the command line, or a mining sheet in your cockpit. Then run every rig from one screen, on your phone or your PC.
The cockpit: your farm, from anywhereapp.fearminer.com is free. No account, no email, no password: your fleet is 12 words you write down once.
- The whole farm on one screen: the fleet's hashrate and history, every rig and every card.
- Mining sheets: what each rig mines (algorithm, wallet, pools), like a HiveOS flight sheet. Change the sheet and the rigs follow.
- Pause, resume or restart a rig.
- Benchmark a rig from the cockpit: it measures its own cards, then mines again as it was.
- Update one rig or the whole fleet. The rig checks the release signature itself, installs it, and puts the previous version back on its own if the new one does not come up (Linux services).
End-to-end encrypted. Your browser and your rigs talk through a relay that only passes sealed messages. The relay cannot read them, cannot add a rig to your fleet and cannot sign a command. A new wallet sent from the cockpit applies at once by default; a rig started with
--remote-wallet-delay makes it wait, announced on every cockpit and cancellable.
fearminer remote off on the rig refuses every remote command.
No rig yet? Open the
cockpit demo.
Coming from a classic miner? Nothing changesThe cockpit is optional. FearMiner is a plain command line miner first:
fearminer -a pearl -o stratum+ssl://POOL:PORT -u YOUR_WALLET -w rig1
- Any stratum pool you choose, TLS or plain. Repeat -o for backup pools. FearMiner never picks a pool for you.
- start_<algo>.sh and .bat launchers in every archive: set your wallet, pool and worker, run.
- HiveOS custom miner package (flight sheet below) and mmpOS package (beta).
- A TOML config file, FEARMINER_* environment variables, an HTTP API (/stats, /hive-stats, /api/v1/*).
Already on HiveOS or mmpOS? Add
--enroll fm1_... (from the cockpit's
Add a rig) to the extra arguments and the rig appears in your cockpit. You keep choosing what it mines in your flight sheet.
Algorithms and fees| Algorithm | Coin | -a | Runs on | Dev fee |
| Quantus (QPoW, Poseidon2) | QTC | quantus | NVIDIA Volta to Blackwell, Apple silicon, other GPUs through Vulkan, CPU | 2 % |
| Pearl (pearlhash) | PRL | pearl | NVIDIA RTX 30, 40 and 50 (sm_86, sm_89, sm_120), about 4.7 GB free | 2 % |
| RandomX | XMR | randomx | CPU | 0.85 % |
| TensorCash (proof of inference) | TSC | tensorcash | NVIDIA, about 11 GB free, 16 GB of RAM | 2 % |
| Common Foundry (ProductionV4) | CMFD | commonfoundry | NVIDIA, about 7 GB free, 8 GB of RAM | 3 % |
How the fee is taken. In one-minute rounds, on a separate connection, never on your pool session. The log prints every round (
fee round started (1 min),
fee round ended: N share(s)) and the rate sits in the panel header and in
--list-algorithms. Each rate is a ceiling set by the signed engine index: it can be lowered, never raised.
The chain is read off your wallet address, so
-a is optional, and every address is checksum-verified before anything is sent.
Engines update without a miner update. Each algorithm is a signed engine the miner downloads, checks and self-tests before it mines, and most new algorithms arrive the same way. A new engine that fails its self-test, or mines 5 % slower over ten minutes, is rolled back on its own.
InstallOne line, kept running as a service. Linux and macOS, no sudo:
curl -fsSL https://get.fearminer.com | sh
Windows (PowerShell):
irm https://get.fearminer.com/win | iex
With a cockpit: its
Add a rig button gives you the line with your rig's code, then you apply a mining sheet to the rig.
curl -fsSL https://get.fearminer.com | sh -s -- fm1_...
Without a cockpit, your wallet on your pool:
curl -fsSL https://get.fearminer.com | sh -s -- --wallet YOUR_WALLET --pool stratum+ssl://POOL:PORT --worker rig1
On Windows, set
FEARMINER_WALLET,
FEARMINER_POOL and
FEARMINER_WORKER before the same
irm line.
The service is your account's and starts with the machine (
--system installs the machine's, with sudo).
fearminer watch then follows the miner live, over SSH too. Nothing runs before the download is checked:
SHA256SUMS against the release key (minisign, or OpenSSL 3), or else download.fearminer.com and GitHub must agree, then the archive against
SHA256SUMS.
Just the binary, nothing installed. The engines download at the first start:
wget -O fearminer https://download.fearminer.com/latest/fearminer-linux-x86_64
chmod u+x fearminer
./fearminer -o stratum+ssl://POOL:PORT -u YOUR_WALLET -w rig1
macOS:
fearminer-macos-arm64. Windows:
fearminer-windows-x86_64.exe. The archives with their
start_<algo> launchers, the HiveOS and mmpOS packages and the signed
SHA256SUMS are on
GitHub and
download.fearminer.com.
HiveOS flight sheetNothing to run in the rig's shell: the flight sheet installs and runs FearMiner. Pick the coin, the wallet and the pool as usual, then the miner
Custom and
Setup Miner Config:
Miner name fearminer_custom (filled in from the URL)
Installation URL https://github.com/fearminer/fearminer/releases/download/v1.15.0/fearminer_custom-1.15.0.tar.gz
Hash algorithm qpow | pearlhash | randomx | tensorcash | commonfoundry
(empty: the miner reads the chain off the wallet)
Wallet and worker %WAL%.%WORKER_NAME%
Pool URL stratum+ssl://POOL:PORT
Pass x
Extra config args --enroll fm1_... (optional: adds the rig to your cockpit)
- Keep the versioned URL as it is: HiveOS reads the miner name and version off the file name. A new version is a new URL.
- TensorCash pools are wss://HOST/v1/ws, Common Foundry pools cmfd+tls://IPv4:PORT?pin=.... The HiveOS guide has every case.
- Your overclock stays in the flight sheet: without an OC option the miner touches no clock, power limit or fan.
- Stats go to the HiveOS dashboard as usual. A crashed or hung card restarts the mining process; --no-supervisor leaves every restart to HiveOS.
- Needs a HiveOS image on Ubuntu 22.04 (glibc 2.35). Driver 550 runs the CUDA 12.4 engines, a newer driver the CUDA 12.9 ones.
mmpOS (beta):
fearminer_mmpos-1.15.0.tar.gz as a custom miner,
--enroll fm1_... after its arguments. See the
mmpOS guide.
Measure it on your own cardYou will not find a hashrate table here. Those numbers go stale with every release and every network fork. Measure your own card instead, no pool and no wallet needed:
fearminer -a pearl --benchmark --duration 60
fearminer -a all --benchmark
It reports the rate, the power measured through NVML (and RAPL for the CPU), the efficiency, and the gap with your previous run. If FearMiner already runs as a service, stop it first (
fearminer service stop) or start the benchmark from the cockpit.
Share your bench! Paste the summary in a reply with your card, driver and FearMiner version. Every result helps the next miner pick a card and helps us tune the kernels. Release posts below will give the gains per card generation.
Built to be checked- Every release ships SHA256SUMS signed with minisign. fearminer verify SHA256SUMS checks it offline; without our binary: minisign -Vm SHA256SUMS -P RWQR3owV+nRhfgNJeUBqcRK868S52NFG2BOrIzKpkQ5ey6lCEfM9+3Eg
- A CycloneDX SBOM (sbom.cdx.json) lists every component inside the binary.
- At start, one egress line names every host the miner will talk to: your pool, download.fearminer.com (the signed engine index), telemetry.fearminer.com, and relay.fearminer.com only once the rig joins a cockpit. Nothing else.
- Telemetry is anonymous and never carries your wallet, worker, password or machine name. fearminer telemetry show prints exactly what is sent, --no-telemetry turns it off, and no feature depends on it.
- Every card answers a known-answer test before its first share. GPU shares are re-checked on the CPU before they are sent (one in ten when checking all of them would cost more than 1 % of a core).
- The miner refuses to run as root unless you ask for it. RandomX's huge pages and MSR tweaks go through a small separate helper, and the original values are put back at exit and after a crash.
FearMiner is closed source, and the binaries are not code-signed yet. Windows SmartScreen and macOS Gatekeeper stop them at the first run, and antivirus engines file any miner under HackTool or CoinMiner. Check the download first, then follow the
steps per platform.
Also inside- Backup pools, TLS certificate pinning, SOCKS5 proxy (Tor works), DNS over HTTPS.
- Supervisor and watchdog: hang, hashrate, share and reject checks, thermal cut-off on NVIDIA.
- NVIDIA overclocking: core and memory clocks, locked clocks, power limit, fan. Read back after every set, put back at exit and after a crash.
- Configuration reloaded without stopping the mining.
- Alerts on Telegram, Discord or your own URL, plus a heartbeat.
- Each rig keeps its own history (90 days by default), read offline with fearminer history.
- Coded errors: fearminer explain E324 tells you what happened and what to do.