hashwarp --help-all lists them. In the configuration file the key is the long name without the dashes: --temp-stop 85 is "temp-stop": [85].
- Value: what the option takes. A comma separated list takes several values:
--pool a.example.com:3333,b.example.com:3333. Tuning, temperature and overclocking lists hold one value per GPU, or one for all. - Options without a value are flags: give them to turn the feature on.
- Several names on one row are the same option. The second ones are the names of Hashwarp 1.x and ethminer, kept so that older command lines keep working.
Pool
| Option | Value | Default | Description |
|---|---|---|---|
-o-P--pool | text, comma separated list | Pool as host:port or URL (stratum+tcp://, stratum+ssl://, http:// for getwork). Repeat or comma separate for failover pools | |
-u--wallet | text (or HASHWARP_WALLET) | Your Parallax address (0x…) | |
-w--worker | text | Worker (rig) name shown by the pool | |
-p--password | text | x | Pool password |
--tls | Use TLS for pools given as host:port | ||
--esm | auto, 0, 1, 2, 3 | auto | Stratum flavour: auto, 0 = EthProxy, 1/2 = EthereumStratum/1.0.0 (NiceHash), 3 = EthereumStratum/2.0.0 |
--solo | text | Solo mine on a Parallax node (default http://127.0.0.1:8545; ws://host:8546 to have the node push new work). Rewards go to the node’s —miner.coinbase | |
-R--report-hashrate--report-hr | Report the hashrate to the pool or node | ||
--pool-retries--farm-retries | number, 0 to 99999 | 3 | Connection attempts before switching to the next pool (a single pool is retried forever) |
--retry-delay | number, 1 to 999 | 5 | Seconds before reconnecting, doubles on each failed attempt (up to 120) |
--no-pool-check | Don’t warn when the pool finds most of the network’s blocks (hashwarp reads the pools’ shares from status.parallaxprotocol.org every 10 minutes) | ||
--failover-timeout | number, 0 to 999 | 0 | Minutes on a failover pool before trying the first one again (0 = never) |
--work-timeout | number, 600 to 99999 | 7200 | Reconnect if no new work arrives for this many seconds |
--response-timeout | number, 2 to 999 | 2 | Reconnect if the pool doesn’t answer within this many seconds |
--getwork-interval--farm-recheck | number, 100 to 99999 | 1000 | Milliseconds between getwork polls (solo / http pools) |
Devices
| Option | Value | Default | Description |
|---|---|---|---|
--list-devices | List the GPUs and exit | ||
--backend | auto, cuda, opencl | auto | auto (CUDA on NVIDIA, OpenCL on the others), cuda or opencl |
-U--cuda | Same as —backend cuda | ||
-G--opencl | Same as —backend opencl | ||
--gpu-id--devices | numbers, comma separated | GPUs to mine on, as listed by —list-devices (default: all) | |
--disable-gpu-nvidia | Don’t mine on NVIDIA GPUs | ||
--disable-gpu-amd | Don’t mine on AMD GPUs | ||
--disable-gpu-intel | Don’t mine on Intel GPUs |
Tuning
Per GPU lists: one value per GPU, or one for all.| Option | Value | Default | Description |
|---|---|---|---|
--cuda-grid-size--cu-grid-size | numbers, comma separated | CUDA grid size [1 .. 131072] (default 8192) | |
--cuda-block-size--cu-block-size | numbers, comma separated | CUDA block size: 32, 64, 128 or 256 (default 128) | |
--cuda-streams--cu-streams | numbers, comma separated | CUDA streams [1 .. 99] (default 2) | |
--cuda-schedule--cu-schedule | auto, spin, yield, sync | sync | How the CPU waits for CUDA kernels: auto, spin, yield or sync (all GPUs) |
--cuda-kernel | auto, 0, 1 | auto | CUDA search kernel: auto (time both on each GPU at start), 0 (best when memory bandwidth limits the hashrate) or 1 (best when compute does) |
--cl-global-work | numbers, comma separated | OpenCL global work size multiplier (default 65536) | |
--cl-local-work | numbers, comma separated | OpenCL work group size: 64, 128 or 256 (default 128) | |
--cl-noexit | OpenCL: don’t use the fast exit kernel (all GPUs) | ||
-L--dag-load-mode | number, 0 to 1 | 0 | 0 = build DAGs on all GPUs at once, 1 = one GPU after the other |
--noeval | Submit solutions without checking them on the CPU first (not recommended) | ||
--ergodicity | number, 0 to 2 | 0 | Pick a new random nonce range 0 = once, 1 = per connection, 2 = per job |
Hardware
| Option | Value | Default | Description |
|---|---|---|---|
--no-hwmon | Don’t read temperatures, fans and power | ||
--temp-stop--tstop | numbers, comma separated | Pause a GPU at this core temperature in C [30 .. 100] (default off) | |
--temp-resume--tstart | numbers, comma separated | Resume a GPU paused by —temp-stop at this temperature (default stop - 15) | |
--mem-temp-stop | numbers, comma separated | Pause a GPU at this memory temperature in C (hotspot on GPUs without memory sensor) [40 .. 120] (default off) | |
--mem-temp-resume | numbers, comma separated | Resume a GPU paused by —mem-temp-stop at this temperature (default stop - 10) | |
--shutdown-temp | number, 0 to 125 | 0 | Stop hashwarp (exit code 6) if any GPU temperature reaches this in C (default off) |
--watchdog-timeout | number, 0 to 3600 | 120 | Seconds a GPU may give no sign of life before hashwarp exits (code 5) so it can be restarted (0 = no watchdog) |
--watchdog-restarts | number, 0 to 100 | 3 | Times per hour a failed GPU is restarted before hashwarp exits (code 5) |
--watchdog-script | text | Run this program (with the GPU index and the reason) when the watchdog gives up, e.g. to reboot the rig |
Overclocking
Per GPU lists, x leaves a GPU alone; needs root or administrator rights.| Option | Value | Default | Description |
|---|---|---|---|
--oc-profile | text, comma separated list | Overclocking profile: stock, efficient (the tuned core clock, or 80% power), tuned (the —oc-tune results) or one of oc-profiles in the configuration file | |
--oc-core-clock | text, comma separated list | Core clock in MHz: locked on NVIDIA, highest state on AMD (0 = default) | |
--oc-core-offset | text, comma separated list | Core clock offset in MHz (NVIDIA) | |
--oc-mem-offset | text, comma separated list | Memory clock offset in MHz (NVIDIA), in the units of nvidia-settings and HiveOS: twice MSI Afterburner’s | |
--oc-mem-clock | text, comma separated list | Memory clock in MHz of the highest state (AMD, 0 = default) | |
--oc-power | text, comma separated list | Power limit in W, or in percent of the default (75%) | |
--oc-fan | text, comma separated list | Fan speed in percent, or auto | |
--oc-profiles | text | Named overclocking profiles, a JSON object (the oc-profiles object of the configuration file) | |
--oc-tune--oc-tune=core | all | Find the best memory clock and the most efficient core clock of each GPU, save them for —oc-profile tuned and exit. —oc-tune=core only tunes the core clock, which is safe | |
--oc-reset | Set the stock clocks, power limits and fan control of the GPUs and exit, e.g. after hashwarp was killed while overclocking | ||
--oc-tune-file | text | Tuning results (default hashwarp-oc.json next to the configuration file) |
Logging and display
| Option | Value | Default | Description |
|---|---|---|---|
--log-file | text | Also write the log to this file | |
--log-max-size | number | 20 | Rotate the log file when it reaches this many MB (0 = never) |
--log-keep | number, 0 to 99 | 3 | Rotated log files to keep |
--log-level | error, warn, info, debug, trace | info | error, warn, info, debug or trace |
--log-format | text, json | text | text, or json: one JSON object per line, on the console and in the log file |
--log-debug | json, per-gpu, connect, switch, submit, flow, comma separated list | Extra debug output (implies —log-level debug): json, per-gpu, connect, switch, submit, flow | |
--no-color--nocolor | Don’t use colors | ||
--syslog | Log without timestamps and colors (for syslog) | ||
--stdout | Log to stdout instead of stderr | ||
--display-interval | number, 1 to 1800 | 10 | Seconds between hashrate lines |
--no-hotkeys | Don’t read keys from the terminal (h for help) | ||
--stats-interval | number, 0 to 86400 | 60 | Seconds between per GPU statistics tables (0 = never) |
API
| Option | Value | Default | Description |
|---|---|---|---|
--api-bind | text | Serve the API on address:port, e.g. 127.0.0.1:21550 (default off) | |
--api-token--api-password | text | Secret required by the API (required to allow control on non loopback addresses) | |
--api-allow-control | Allow API clients to pause, restart and change pools | ||
--rig | text, comma separated list | Show another rig on the dashboard: [name=][token@]host[:port] (port 21550 by default). Repeat it, or separate rigs with commas | |
--monitor | Don’t mine: only serve the dashboard of the —rig rigs (needs —api-bind) |
Other
| Option | Value | Default | Description |
|---|---|---|---|
-c--config | Read options from a JSON file (keys are option names) | ||
--save-config | text | Write the options given to this JSON file and exit | |
-M-Z--benchmark--simulation | number | Benchmark offline on the DAG of this block number | |
--selftest | number | 30 | Check every GPU’s results on the CPU offline for up to this many seconds (default 30) and exit with 0 if all are correct |
--exit | Exit when a GPU fails instead of carrying on |