DEV Community

crow
crow

Posted on

Earn From Your Spare GPU: Step‑by‑Step Guide to Setting Up Golem on a Linux Workstation

#golem #gpu #ubuntu #passiveincome

SEO Title:

Earn From Your Spare GPU: Step‑by‑Step Guide to Setting Up Golem on a Linux Workstation

Intro – The Budget Problem (Hook)

You’ve got a spare GPU humming in the back of your office or home lab, and you’re constantly hearing about “cloud AI services” that bill by the hour. What if you could flip that idle card into a small income stream without paying a cent to Amazon Web Services or Google Cloud?

I built an entire AI‑ready workstation for $800 last year: a mid‑range CPU, 16 GB of RAM, a fast NVMe SSD, and a single NVIDIA RTX 3060. The whole kit (including the GPU) was under $300 – perfect for a hobbyist or a small lab that can’t justify a dedicated data center.

In this video/script we’ll walk through every step from selecting the right parts to installing Ubuntu, drivers, CUDA, LM Studio, and finally Golem. By the end you’ll be running your own GPU‑powered node and earning back your hardware costs in just a few months.

1️⃣ Pick Your Hardware (5–7 Minutes)

Component Recommendation Why it Works
CPU AMD Ryzen 5 5600X or Intel i5‑13400F 6 cores, great single‑thread performance for training small models.
Motherboard B550 (AMD) / B660 (Intel) with PCIe 4.0 support Enough lanes for GPU and future expansion.
RAM 16 GB DDR4/DDR5 (3200 MHz or faster) Sufficient for most AI experiments; upgradeable later.
Storage NVMe SSD, 500 GB Fast read/write for datasets and model checkpoints.
GPU NVIDIA RTX 3060 CUDA‑capable, 12 GB VRAM, affordable.
Power Supply 650W 80+ Gold Reliable power with headroom for GPU + future upgrades.
Case & Cooling Mid‑tower with good airflow Keeps temperatures low during long training runs.

Affiliate Placeholder:

• [AFF: Amazon RTX 3060 (~$300)] – Great price‑to‑performance ratio.

• [AFF: NVMe SSD] – Fast storage for your datasets.

Tip: If you’re building a homelab, consider a case with a built‑in fan controller so you can tweak airflow without opening the box every time.

2️⃣ Install Ubuntu (10–15 Minutes)

  1. Download ISO

    • Go to the official Ubuntu website and grab the latest LTS release (22.04 or newer).
    • Burn it to a USB stick with Rufus or balenaEtcher.

  2. Boot from USB

    • Reboot, press F12/Esc to enter boot menu → choose your USB.

  3. Installation Wizard

    • Select “Install Ubuntu”.
    • When asked about installation type, choose Erase disk and install Ubuntu (or use a custom partition if you already have Windows).
    • Set your time zone, keyboard layout, username, password.

  4. Post‑install Updates

   sudo apt update && sudo apt upgrade -y
Enter fullscreen mode Exit fullscreen mode
  1. Optional: Install an Ubuntu Reference Book
    • A handy guide for beginners can be found on Amazon or the official Ubuntu documentation.
    • [AFF: Ubuntu book] – Great for troubleshooting.

3️⃣ NVIDIA Drivers & CUDA Toolkit (15–20 Minutes)

  1. Add Graphics Drivers PPA 
   sudo add-apt-repository ppa:graphics-drivers/ppa
   sudo apt update
Enter fullscreen mode Exit fullscreen mode
  1. Install Latest Driver
    • Replace 470 with the recommended driver for RTX 3060. 
   sudo ubuntu-drivers autoinstall
Enter fullscreen mode Exit fullscreen mode
  1. Verify Installation 
   nvidia-smi
Enter fullscreen mode Exit fullscreen mode

You should see your GPU listed with its driver version.

  1. Install CUDA Toolkit (Optional but Recommended)

    • Download the installer from NVIDIA’s site (choose the one that matches your Ubuntu version).
    • Run it following the on‑screen instructions – just accept defaults unless you have a custom setup.

  2. Set Environment Variables

   echo 'export PATH=/usr/local/cuda/bin:$PATH' >> ~/.bashrc
   echo 'export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH' >> ~/.bashrc
   source ~/.bashrc
Enter fullscreen mode Exit fullscreen mode
  1. Test CUDA
    • Compile a simple sample program or run nvcc --version.

4️⃣ Install LM Studio (10–15 Minutes)

LM Studio is a lightweight, open‑source framework that lets you develop and deploy machine learning models locally.

  1. Download the Latest Release 
   wget https://github.com/LM-Studio/LM-Studio/releases/download/vX.Y.Z/LMStudio-linux-x86_64.AppImage
   chmod +x LMStudio-linux-x86_64.AppImage
Enter fullscreen mode Exit fullscreen mode
  1. Run it 
   ./LMStudio-linux-x86_64.AppImage
Enter fullscreen mode Exit fullscreen mode
  • On first run, it will download the necessary dependencies.
  1. Create a Test Model
    • In LM Studio, click “New Project” → choose a simple model (e.g., GPT‑Neo).
    • Train or fine‑tune on a small dataset to confirm that the GPU is being used (nvidia-smi will show activity).

5️⃣ Set Up Golem (20–30 Minutes)

Golem is a decentralized network where you can rent out your GPU compute. Here’s how to join as a node.

  1. Install Go 
   sudo apt install golang-go -y
Enter fullscreen mode Exit fullscreen mode
  1. Download Golem CLI 
   go install github.com/golemfactory/golem-cli@latest
   export PATH=$PATH:$HOME/go/bin
Enter fullscreen mode Exit fullscreen mode
  1. Create a Wallet
    • Golem uses its own cryptocurrency (GNT). 
   gollem wallet create
Enter fullscreen mode Exit fullscreen mode
  • Save the mnemonic securely; you’ll need it to recover your wallet.
  1. Register Your Node 
   gollem node start --gpu 0
Enter fullscreen mode Exit fullscreen mode
  • The --gpu 0 flag tells Golem which GPU index to expose (use nvidia-smi to confirm).
  • If you have multiple GPUs, add more flags (--gpu 1, etc.).

  1. Configure Node Settings

    • Edit the config file (~/.golem/node/config.yaml) to set your desired price per hour and other parameters.
    • Example snippet: 
     gpu:
   price_per_hour: 0.10 # USD

  2. Start Listening for Tasks

   gollem node listen
Enter fullscreen mode Exit fullscreen mode
  • The node will now advertise its availability to the Golem network.
  1. Monitor Earnings
    • Use the CLI or the web dashboard (http://localhost:8080) to see active tasks and payouts.
    • Payouts are automatically sent to your GNT wallet; you can later convert them to fiat via exchanges that support GNT.

6️⃣ Optimize & Maintain (5–10 Minutes)

Tip Action
Keep Drivers Updated sudo apt update && sudo apt upgrade regularly.
Monitor Temperature Install nvtop or nvidia-smi --query-gpu=temperature.gpu --format=csv.
Backup Configs Store your Golem wallet mnemonic and node config in a password manager.
Scale Up When you need more compute, add another GPU and run gollem node start --gpu 1 etc.

Outro – Call to Action

That’s it! You’ve built an AI‑ready Linux workstation for under $800, installed all the necessary software, and now your spare RTX 3060 is earning you real money on the Golem network.

What’s next?

  • Try running a small training job in LM Studio while your node is live – you’ll see how the GPU usage balances between local tasks and rented compute.
  • Experiment with different models or datasets to maximize earnings per hour.

👇 Drop a comment below with your own build, any questions, or tips that helped you get started. Don’t forget to hit Subscribe for more beginner‑friendly guides on Linux, AI, and homelabs. Until next time—happy mining!andrew@echo-X570-Taichi:~/Echo$

Top comments (0)