Windows NVIDIA Setup

Run GPUFlow on Windows with NVIDIA GPUs using Windows Subsystem for Linux 2 (WSL2). This setup allows your Windows PC to act as a GPU provider while maintaining normal Windows functionality.

Prerequisites

Before starting, verify your system meets these requirements:

Windows 10 version 2004+ or Windows 11 (any version)
NVIDIA GPU with 8GB+ VRAM (GTX 1060 or newer)
16GB+ system RAM (32GB recommended)
Administrator access to install software
Stable internet connection (25+ Mbps upload)

Step 1: Check Windows version

Open PowerShell as Administrator and check your Windows version:

# Check Windows version
winver

# Check Windows build number
[System.Environment]::OSVersion.Version

You need Windows 10 build 19041 or newer, or any Windows 11 version.

Step 2: Enable WSL2

WSL2 is required for GPU passthrough to Linux containers.

Install WSL2:

# Run as Administrator
wsl --install

# Restart your computer when prompted

After restart, verify WSL2 installation:

wsl --version

You should see WSL version 1.0.0 or newer.

Step 3: Install Ubuntu on WSL2

Install Ubuntu as your Linux distribution:

# Install Ubuntu 22.04 LTS
wsl --install -d Ubuntu-22.04

# Set Ubuntu as default
wsl --set-default Ubuntu-22.04

Set up Ubuntu:

Launch Ubuntu from Start menu
Create a username and password when prompted
Wait for initial setup to complete

Step 4: Install NVIDIA drivers for Windows

Install NVIDIA drivers that support WSL2:

Visit NVIDIA’s WSL2 driver page
Download NVIDIA Drivers for CUDA on WSL2
Run the installer as Administrator
Restart your computer

Step 5: Configure WSL2 with GPU support

Verify GPU access in WSL2:

# Launch WSL2 Ubuntu
wsl

# Check if GPU is visible
lspci | grep -i nvidia

Update Ubuntu packages:

# Update package lists
sudo apt update && sudo apt upgrade -y

# Install essential packages
sudo apt install curl wget software-properties-common

Step 6: Install NVIDIA Container Toolkit in WSL2

Install the container toolkit inside your WSL2 Ubuntu environment:

# Add NVIDIA repository
distribution=$(. /etc/os-release;echo $ID$VERSION_ID) \
    && curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg \
    && curl -s -L https://nvidia.github.io/libnvidia-container/$distribution/libnvidia-container.list | \
        sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | \
        sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list

# Install toolkit
sudo apt update
sudo apt install nvidia-container-toolkit

Step 7: Install Docker in WSL2

Install Docker inside your WSL2 Ubuntu environment:

# Install Docker
curl -fsSL https://get.docker.com -o get-docker.sh
sudo sh get-docker.sh

# Add your user to docker group
sudo usermod -aG docker $USER

# Start Docker service
sudo service docker start

Configure Docker for NVIDIA GPUs:

# Configure NVIDIA runtime
sudo nvidia-ctk runtime configure --runtime=docker

# Restart Docker
sudo service docker restart

Step 8: Test GPU access in containers

Verify that containers can access your NVIDIA GPU:

# Test GPU access
docker run --rm --gpus all nvidia/cuda:12.0-base-ubuntu22.04 nvidia-smi

You should see your GPU information displayed, confirming GPU passthrough works.

Step 9: Deploy GPUFlow provider

Create data directory:

sudo mkdir -p /opt/gpuflow
sudo chown $USER:$USER /opt/gpuflow

Run GPUFlow provider:

docker run -d \
  --name gpuflow-provider \
  --restart=unless-stopped \
  --gpus=all \
  --network=host \
  -v /opt/gpuflow:/data \
  -e GPUFLOW_API_KEY="get-from-dashboard" \
  ghcr.io/gpuflow/provider:latest

Verify provider is running:

docker ps
docker logs gpuflow-provider

Look for these success messages in the logs:

GPU detection successful
Network connectivity established
Provider registered with GPUFlow

Step 10: Auto-start services

Configure services to start automatically when Windows boots:

Create startup script:

# Create startup script
cat > ~/start-gpuflow.sh << 'EOF'
#!/bin/bash
sudo service docker start
sleep 5
docker start gpuflow-provider
EOF

chmod +x ~/start-gpuflow.sh

Add to Windows startup:

Press Win + R, type shell:startup, press Enter
Create a new file called GPUFlow.bat with this content:

@echo off
wsl -d Ubuntu-22.04 -u your-username /home/your-username/start-gpuflow.sh

Replace your-username with your actual WSL2 username

Managing your provider

Common management commands (run in WSL2):

# Check provider status
docker ps

# View logs
docker logs gpuflow-provider -f

# Restart provider
docker restart gpuflow-provider

# Stop provider (to use GPU for gaming)
docker stop gpuflow-provider

# Start provider again
docker start gpuflow-provider

Performance considerations

Resource allocation:

WSL2 uses about 1-2GB RAM overhead
GPU switching between Windows and containers is seamless
Negligible performance impact on Windows gaming

Gaming while providing:

Stop the provider before launching games: docker stop gpuflow-provider
Start provider after gaming: docker start gpuflow-provider
Consider scheduling based on your gaming hours

Next steps

Your Windows NVIDIA provider is now running:

Monitoring and optimization

Check system resource usage:

# In WSL2, monitor container resources
docker stats gpuflow-provider

Windows Task Manager:

GPU usage appears under “GPU Engine” tabs
WSL2 appears as “Vmmem” process in memory usage
Docker containers don’t appear directly in Windows processes