🚀 CUDA Toolkit Installation Guide

🎯 Prerequisites

Hardware Requirements

Component	Minimum	Recommended
GPU	NVIDIA Kepler+	NVIDIA Ampere/Hopper
RAM	4GB	16GB+
Disk Space	2.5GB	10GB
CPU	x86_64	Multi-core x86_64

Software Requirements

graph LR
    A[Operating System] --> B[Linux/Windows]
    B --> C[Compatible Driver]
    C --> D[Development Tools]

Loading

• 🖥️ Operating System
  • Linux: Kernel 3.10+
  • Windows: 10/11
• 🔧 Development Tools
  • GCC 7+ (Linux)
  • MSVC 2019+ (Windows)
• 🎮 NVIDIA Driver: 525.0.0+

---

📦 Installation

Linux Installation

Quick Install (Recommended)

# Download CUDA Toolkit
wget https://developer.download.nvidia.com/compute/cuda/12.6.0/local_installers/cuda_12.6.0_535.54.03_linux.run

# Install
sudo sh cuda_12.6.0_535.54.03_linux.run

Package Manager Install

# For Ubuntu/Debian
sudo apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/3bf863cc.pub
sudo add-apt-repository "deb https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/ /"
sudo apt update
sudo apt install cuda-toolkit-12-6

Windows Installation

Network Installer

1. Download the CUDA Network Installer
2. Run the installer
3. Follow the wizard 🧙‍♂️

---

🔧 Environment Setup

Add these to your shell configuration file (~/.bashrc, ~/.zshrc, etc.):

# CUDA Toolkit Path
export CUDA_HOME=/usr/local/cuda-12.6
export PATH="$CUDA_HOME/bin:$PATH"
export LD_LIBRARY_PATH="$CUDA_HOME/lib64:$LD_LIBRARY_PATH"

📝 Optional Environment Variables

# Optional: CUDA device order
export CUDA_DEVICE_ORDER=PCI_BUS_ID

# Optional: Default device
export CUDA_VISIBLE_DEVICES=0

# Optional: CUDA cache path
export CUDA_CACHE_PATH="$HOME/.cuda-cache"

---

✅ Verification

# Check CUDA compiler
nvcc --version

# Check NVIDIA driver
nvidia-smi

# Run sample test
cd $CUDA_HOME/samples/1_Utilities/deviceQuery
make
./deviceQuery

Expected Output:

📊 CUDA Device Query Results:  CUDA Version: 12.6
  Driver Version: 535.54.03  CUDA Capability: x.y  ...

---

🔍 Troubleshooting

Common Issues

Issue	Solution
nvcc: command not found	Check PATH variable
Driver version mismatch	Update NVIDIA driver
Installation fails	Check system requirements
CUDA not found	Verify environment variables

Diagnostic Commands

# Check CUDA installation
ls -l /usr/local/cuda

# Check driver status
systemctl status nvidia-driver

# Check GPU detection
lspci | grep -i nvidia

---

🚀 Advanced Configuration

Multi-GPU Setup

# List all GPUs
nvidia-smi -L

# Set specific GPUs
export CUDA_VISIBLE_DEVICES=0,1

# Check GPU utilization
nvidia-smi -l 1

Performance Optimization

• Enable persistence mode

sudo nvidia-smi -pm 1

• Set GPU clock speeds

sudo nvidia-smi -ac 5001,1590

---

📚 Additional Resources

• 📖 CUDA Documentation
• 🎓 CUDA Training
• 💻 Sample Projects
• 🗣️ Developer Forums

---