GPU Plotter

Autonomys utilizes your drive storage, specifically SSD or NVMe drives, to store plots. After the plotting process is finished, these plots are then farmed using your CPU. Farming is not particularly demanding on the CPU, enabling most modern processors to manage a substantial farm size. However, the plot creation process is highly resource intensive, which makes CPU plotting the main bottleneck.

Utilizing GPU plotting allows you to harness the power of compatible GPUs for plot generation and replotting, either in conjunction with or as a substitute for CPU processing. While many modern CPUs can complete the plotting of a sector in less than two minutes, a single high performance GPU can accomplish the same task in under five seconds, greatly improving efficiency and speed.

Although GPU plotting is not mandatory, it provides enhanced energy efficiency and speed compared to relying solely on a CPU.

Platform Compatibility

Platform	🐧 Linux	🪟 Windows	Nvidia	AMD	Intel
Advanced CLI	✅	✅	✅	🛠️	🔮
Space Acres	✅	✅	✅	🔜	🔮

🛠️ Limited AMD Support for Linux only is available in recent test builds. The most recent test builds are linked on the forum

See Discord farmer-chat channel for limited support.

Supported GPUs

Nvidia

Series/Model	Supported
RTX 20xx and Newer	✅
GTX 16 Series	✅

Nvidia drivers version 550 or later are required. Installing the CUDA Toolkit is not required.

AMD

There are many challenges to overcome regarding AMD ROCm support. There is much more information on this topic on the forum.

You must be using the latest test build for AMD support.

Series/Model	Ubuntu	Windows
RX 7900 XTX	✅	❌
RX 7600 XT	✅	❌
RX 7600	✅	❌
RX 6800	✅	❌
RX 6700 XT	❔	❌
RX 6600 XT	✅	❌
RX 6600	✅	❌
RX 5700 XT	❌	❌
RX 5700	❌	❌
RX 5600	❌	❌
AMD BC-250	❌	❌

🐧 Ubuntu

The subspace-farmer-rocm-* binaries provide ROCm support, with corresponding CLI options similar to CUDA and prefixed with --rocm.

1. In order to install necessary libraries go to Ubuntu native installation — ROCm installation (Linux) and follow these steps for your Ubuntu version:

   • Package signing key
   • Register ROCm packages You don't need a custom driver or full ROCm toolchain to use already compiled application, so skip all other steps.

2. Next install a single package with ROCm runtime:

   sudo apt-get install --no-install-recommends hip-runtime-amd

   And register the library so that farmer and other apps can find it:

   echo "/opt/rocm/lib" | sudo tee /etc/ld.so.conf.d/rocm.conf > /dev/null
   sudo ldconfig

3. Then, to allow the user to access GPU for compute purposes, you need to add your user to render group (please never run farmer as root with sudo 🙏 ):

   sudo usermod -a -G render $LOGNAME

4. Now you'll need to log out of your user profile and log back in or simply reboot for group changes to take effect and you're ready to go.

🪟 Windows

AMD GPUs on Windows (including through WSL) are not currently supported. We hope to add support in the future.

🐳 Docker

Cross-compilation no longer requires separate Dockerfiles (which is why they are removed), both native and cross-compilation is supported with a single file

Container image is now built as multi-platform, meaning no -aarch64 suffix and no awkwardness related to that

Aarch64 farmer container image is now compiled with CUDA support (while regular executable isn't)

Container image now ships with a second executable /subspace-farmer-rocm (see explanation above as to why second binary is needed). Here is how it can be used with Docker Compose, you can use this as an inspiration for other use cases:

services:
  famer:
    image: ghcr.io/autonomys/farmer:TAG
    # Beginning of ROCm-specific options
    devices:
      - /dev/kfd:/dev/kfd
      # This will make all GPUs accessible, you can also limit this
      # to individual devices like `/dev/dri/renderD128`
      - /dev/dri:/dev/dri
    security_opt:
      - seccomp:unconfined
    # `root` user can be replaced with ID of the user on the host system
    # that is in `render` group, needed to access `/dev/kfd` device
    user: root
    entrypoint: /subspace-farmer-rocm
    # End of ROCm-specific options
    ...the rest of typical options you'd normally use

Intel

Intel Arc GPUs may be supported in the future, but specific compatibility details have not been announced yet.

Common Plotting Parameters

Enable CPU Plotting

When a compatible GPU is detected, CPU plotting is automatically disabled by default, but can be re-enabled if needed by specifying number of concurrently encoded sectors: --cpu-sector-encoding-concurrency <sectors>

--cpu-sector-encoding-concurrency 2

Disable GPU Plotting

Linux

  --cuda-gpus ""

Windows

--cuda-gpus 99

Specify specific GPUs

Specify particular GPUs for plotting rather than using all available GPUs (the default configuration employs all compatible GPUs): --cuda-gpus <gpu_ids>

--cuda-gpus 0,1,3

Farming Cluster

When utilizing Farming Cluster, particularly with multiple or fast GPUs, you might encounter limitations due to your network's bandwidth. High performance GPUs can easily surpass the capacity of a 1G connection. While this won't cause the process to fail, it may result in your GPU idling as it waits for data to transfer. To optimize performance in such scenarios, consider upgrading to networking solutions of 2.5G, 10G, or higher.