Skip to content

MMX Node v0.10.3 (with compressed plot support)
Compare
Choose a tag to compare
@madMAx43v3r madMAx43v3r released this 13 May 11:43
· 19 commits to master since this release
v0.10.3
a9cf825

New features in this version:

  • Remote compute server, to offload compute from harvesters to another set of compute machines.
  • Remote compute proxy, to distribute farming compute to multiple compute machines in a centralized manner.

Remote_Compute_Drawings drawio

To use the remote compute feature:

  • Start chia_recompute_server on the machine that is doing the compute (available here: https://github.com/madMAx43v3r/chia-gigahorse/tree/master/chiapos/windows).
  • export CHIAPOS_RECOMPUTE_HOST=... on the harvester (replace ... with the IP address or host name of the compute machine)
  • On Windows you need to set CHIAPOS_RECOMPUTE_HOST variable via system settings.
  • CHIAPOS_RECOMPUTE_HOST can be a list of recompute servers, such as CHIAPOS_RECOMPUTE_HOST=192.168.0.11,192.168.0.12. A non-standard port can be specified via HOST:PORT syntax, such as localhost:12345. Multiple servers are load balanced in a fault tolerant way.
  • CHIAPOS_RECOMPUTE_PORT can be set to specify a custom default port for chia_recompute_server (default = 11989).
  • See chia_recompute_server --help for available options.

To use the remote compute proxy:

  • Start chia_recompute_proxy -n B -n C ... on a machine A. (B, C, etc are running chia_recompute_server)
  • Set CHIAPOS_RECOMPUTE_HOST on your harvester(s) to machine A.
  • chia_recompute_proxy can be run on a central machine, or on each harvester itself, in which case A = localhost.
  • See chia_recompute_proxy --help for available options.

Same as the open source MMX node but with support for farming compressed plots. Regular plots are supported as well.

Note: Newer AMD GPUs starting with RX 6000 series do not work due to a driver bug, we have to wait for AMD to fix it. Vega series is known to still work when using an older driver with amdgpu-install --opencl=pal,legacy. RX 5000 series is unknown at this point.

Note: See limit-gpu-usage on how to select an OpenCL platform if you have multiple (excluding Nvidia).

Assets 3
Loading