Golem GPU Provider: Installation and Configuration Guide

Ready to earn GLM tokens by sharing your GPU's computing power with the Golem Network? This comprehensive guide provides step-by-step instructions for setting up and configuring a dedicated GPU provider node.

The gpu-golem-live project allows you to contribute your GPU resources to the Golem Network. These powerful resources will enable demanding tasks such as artificial intelligence (AI) processing, contributing to a decentralized network for advanced computations. As a provider, you will be rewarded with GLM tokens for your contribution.

Before starting, check if your hardware meets these requirements:

• IOMMU Support: A computer with a motherboard and CPU that supports IOMMU (Input-Output Memory Management Unit) virtualization.
• Nvidia GPU: An Nvidia graphics card, at least a 30xx series, with 8 GB or more VRAM.
• RAM: At least the same amount of RAM as your GPU's VRAM, with double the amount recommended.
• SSD: An SSD with a minimum capacity of 64 GB, connected via a fast USB interface (USB 3.2 recommended). A high-speed USB flash drive can work but is not recommended.
• Network: A stable internet connection with a minimum speed of 100 Mbps (both upload and download), preferably via an RJ45 Ethernet cable.
• Ethereum Wallet: An Ethereum wallet to receive your GLM earnings.
• BIOS Settings: Enable VT-d (for Intel CPUs) or SVM (for AMD CPUs) in your computer's BIOS settings.

⚠️ Important Considerations:

• Network Optimization: For better peer-to-peer connections, consider opening UDP port 11500 on your router.
• GPU Detection Issues: If you've enabled virtualization in BIOS but the setup doesn't detect your GPU, try moving it to a different PCIe slot.
• GPU Usage: Your GPU might not be in use all the time. It will be active when tasks are assigned.
• Single GPU Configuration: The current setup supports one provider node per GPU. Multi-GPU support will be added later.

1. Download the gpu-golem-live provider image: Download image

https://gpu-provider.dev.golem.network/golem-gpu-live-latest.img.xz

You can also download it using wget:

wget https://gpu-provider.dev.golem.network/golem-gpu-live-latest.img.xz

The file is about 2GB.

2. Extract the downloaded .xz archive using an archiving tool (7-Zip, The Unarchiver, or your system's default tool) or use these terminal commands:

xz -d golem-gpu-live-latest.img.xz

The uncompressed image file (golem-gpu-live-latest.img) should be around 8.6 GB.

1. Connect the SSD: Connect your SSD to your computer.
2. Delete Existing Partitions: Delete all existing partitions on the SSD.

1. Linux: Run sudo fdisk -l to find your SSD (identified by its size) and its path (e.g., /dev/sdX).
2. macOS: Use diskutil list to locate your SSD and its identifier (e.g., /dev/diskY).

Warning: Double-check the SSD path before using dd, as incorrect usage can cause data loss!

If your extracted image is in your Downloads folder and the SSD is /dev/sdX:

sudo dd if=Downloads/golem-gpu-live-latest.img of=/dev/sdX bs=4M status=progress

• Replace /dev/sdX with the actual path.
• bs=4M is for faster transfer.
• status=progress shows the flashing progress.

1. Select Image: In balenaEtcher, click "Flash from file" and choose golem-gpu-live-latest.img.

   • 

2. Select Target: Click "Select target" and choose your SSD. Double-check you've chosen the right drive!

   • 
   • 

3. Flash: Click "Flash!" to begin. Confirm with "Yes, I'm sure" if prompted.

   • 
   • 

4. Progress: balenaEtcher will show a progress bar. This can take 5-10 minutes.

   • 
   • 
   • 
   • 
   • 

1. Connect the SSD: Make sure the SSD is connected to your computer.

2. Reboot: Restart your computer.

3. Boot Menu: Enter your BIOS/UEFI boot menu (usually F12, F2, Esc, or Del).

4. Select SSD: Select the SSD with gpu-golem-live. Make sure to choose UEFI OS mode.

   • 

To always boot from the SSD when connected:

1. Access BIOS: Enter BIOS settings during startup (key varies by motherboard, look for F2, Del, Esc).

2. Boot Order: Find the "Boot" or "Boot Order" section.

3. Prioritize SSD: Make the SSD the first boot option.

4. Save and Exit: Save your changes and exit BIOS.

   • 

1. Select OS: Press Enter to choose GOLEM GPU Live.

   • 

2. Welcome: Click "OK."

   • 

3. Terms of Use: Read and accept by selecting "YES".

   • 

4. Persistent Storage: Select a partition on your SSD for storing provider data.

   • 

5. Password: A random password will be generated. Write it down securely for future SSH access.

   • 

6. Network Configuration: Write down the IP address shown (needed for SSH).

   • 

7. Ethereum Wallet: Enter your wallet address to receive your GLM earnings.

   • 

8. Set Price: Enter your desired hourly rate in GLM.

   • 

9. GPU Selection: Choose the GPU you want to use with Golem.

   • 
   • 

10. Confirmation: Click "OK" to finalize. Your computer will restart, and the provider will start running (your monitor might go blank, this is normal).

    • 
    • 
    • 

Your provider is now active on the Golem Network! To check its status and earnings, visit the Golem Network Stats page: https://stats.golem.network/network/providers/online

Search for your provider using its randomly generated name ("Node name" in the configuration summary).

Connect via SSH:

ssh golem@<provider_ip_address>

Use the IP from step 6 and the password from step 5.

1. Connect to your provider via SSH.
2. Run golemsp settings show.
3. Look for the "node name" - that's your provider's name on the Golem Network Stats page.

1. Connect to your provider via SSH.

2. Run:

   golemsp settings set --node-name '<new_node_name>'

1. Connect to your provider via SSH.

2. Run:

   golemsp settings set --account <new_wallet_address>

1. Connect to your provider via SSH.

2. Convert your desired hourly rate to a rate per second (divide by 3600). For example, for 0.25 GLM per hour:

   0.25 GLM/h / 3600 seconds/hour = 0.00006944444444444444 GLM per second

3. Run:

   ya-provider preset update --name vm-nvidia --price duration=0.00006944444444444444 --no-interactive

4. Activate the changes:

   ya-provider preset activate vm-nvidia

1. Open your local terminal (not via SSH).

2. Run:

   scp golem@<your_provider's_ip>:/home/golem/golemwz.log golemwz.log
   scp golem@<your_provider's_ip>:/home/golem/.local/share/yagna/yagna_rCURRENT.log yagna_rCURRENT.log

   Enter your SSH password when asked. The logs will be downloaded to your current directory.

Need more help? Join us on Discord!