Skip to content
TensorRigs

Operating System Installation

Choosing Your Operating System

Section titled “Choosing Your Operating System”

For deep learning workstations, you have three main options:

Section titled “Ubuntu Linux (Recommended)”

Best for: Most users, especially those new to deep learning

Pros:

  • Best NVIDIA driver support
  • Largest community for ML/DL
  • Most tutorials assume Ubuntu
  • Native Docker support
  • Package managers optimized for ML tools

Cons:

  • Learning curve if new to Linux
  • Some commercial software unavailable
  • Gaming performance worse than Windows

Recommended version: Ubuntu 22.04 LTS or 24.04 LTS

Windows 11

Section titled “Windows 11”

Best for: Users who need Windows software or prefer Windows workflow

Pros:

  • Familiar interface
  • Better for gaming/general use
  • WSL2 provides Linux compatibility
  • Easy NVIDIA driver installation

Cons:

  • WSL2 has performance overhead (5-15%)
  • Some ML tools Linux-only
  • More resource overhead
  • Updates can interrupt training

Best for: Dual-use systems (work + personal)

Dual Boot

Section titled “Dual Boot”

Best for: Users who want both

Pros:

  • Native performance on both OSes
  • Use Windows for gaming, Ubuntu for training
  • No VM/WSL overhead

Cons:

  • More complex setup
  • Need to reboot to switch
  • Requires partitioning

Installation Guide

Section titled “Installation Guide”

Ubuntu 22.04 LTS Installation

Section titled “Ubuntu 22.04 LTS Installation”

1. Download Ubuntu

Section titled “1. Download Ubuntu”

Visit ubuntu.com/download and download:

  • Ubuntu 22.04.3 LTS (most stable)
  • Ubuntu 24.04 LTS (latest, if you want newer packages)

2. Create Bootable USB

Section titled “2. Create Bootable USB”

On Windows:

# Download Rufus from rufus.ie
# Select your USB drive
# Select Ubuntu ISO
# Click Start

On Linux:

# Find your USB device
lsblk

# Write ISO to USB (replace sdX with your device)
sudo dd if=ubuntu-22.04.3-desktop-amd64.iso of=/dev/sdX bs=4M status=progress
sudo sync

3. Boot from USB

Section titled “3. Boot from USB”
  1. Insert USB drive
  2. Restart computer
  3. Press Boot Menu key (usually F12, F11, or Esc)
  4. Select USB drive

4. Installation Process

Section titled “4. Installation Process”

Choose Installation Type:

○ Try Ubuntu (test without installing)
● Install Ubuntu (proceed with installation)

Keyboard Layout:

  • Select your keyboard layout
  • Test in the text box to verify

Updates and Other Software:

☑ Normal installation
☑ Download updates while installing Ubuntu
☑ Install third-party software for graphics and Wi-Fi hardware

Installation Type:

For ML-only system:

● Erase disk and install Ubuntu
  (Uses entire disk - simplest option)

For dual boot with Windows:

● Install Ubuntu alongside Windows
  (Follow partition wizard)

Manual partitioning (advanced users):

  • / (root): 50-100GB (ext4)
  • /home: Remaining space (ext4)
  • swap: 16-32GB (or equal to RAM)
  • Optional: Separate /data partition for datasets

5. Post-Installation

Section titled “5. Post-Installation”

Update system:

sudo apt update
sudo apt upgrade -y

Install essential tools:

sudo apt install -y \
  build-essential \
  git \
  curl \
  wget \
  vim \
  tmux \
  htop \
  net-tools

Partitioning Recommendations

Section titled “Partitioning Recommendations”

For ML Workloads

Section titled “For ML Workloads”

Minimum (Ubuntu):

  • /: 50GB
  • /home: 100GB+
  • swap: Equal to RAM

Recommended:

  • /: 100GB (OS and software)
  • /home: 200GB+ (configs, small datasets)
  • /data: Remaining (large datasets)
  • swap: 16-32GB

Why separate /data?

  • Easy to reinstall OS without losing datasets
  • Can use different filesystem (ext4 vs xfs)
  • Better organization

Storage Configuration

Section titled “Storage Configuration”

For single NVMe:

/dev/nvme0n1p1  512MB   EFI
/dev/nvme0n1p2  100GB   / (root)
/dev/nvme0n1p3  Remaining  /home

For multiple drives:

NVMe (fast):    OS and code
SATA SSD:       Datasets
HDD:            Archives and backups

Performance Optimizations

Section titled “Performance Optimizations”

Ubuntu Tweaks

Section titled “Ubuntu Tweaks”

Install nvidia-prime for laptop GPUs:

sudo apt install nvidia-prime
sudo prime-select nvidia

Disable unnecessary services:

# Disable Bluetooth (if not needed)
sudo systemctl disable bluetooth

# Disable CUPS (printing)
sudo systemctl disable cups

Install minimal desktop (optional):

# For server-like setup with remote access
sudo apt install ubuntu-server
# Then SSH in from another machine

Windows Tweaks

Section titled “Windows Tweaks”

Disable Windows Search indexing on data drives:

Right-click drive → Properties → Uncheck "Allow files on this drive to have contents indexed"

Set power plan:

Control Panel → Power Options → High Performance

Disable Game Bar:

Settings → Gaming → Game Bar → Off

Verification

Section titled “Verification”

Check Installation Success

Section titled “Check Installation Success”
# Check OS version
lsb_release -a

# Check kernel version
uname -r

# Check disk space
df -h

# Check memory
free -h

# List PCI devices (GPUs should appear)
lspci | grep -i nvidia

Common Issues

Section titled “Common Issues”

GPU Not Detected

Section titled “GPU Not Detected”

Boot Order Issues

Section titled “Boot Order Issues”
  • Access BIOS boot menu
  • Set proper boot priority

Dual Boot - Windows Missing

Section titled “Dual Boot - Windows Missing”
# Update GRUB
sudo update-grub

# Should detect Windows

Ubuntu Stuck at Login Loop

Section titled “Ubuntu Stuck at Login Loop”
  • Usually a driver issue
  • Boot to recovery mode
  • Reinstall display drivers

Next Steps

Section titled “Next Steps”

After OS installation:

  1. Install NVIDIA drivers and CUDA
  2. Set up Python environment
  3. Configure remote access (see OS documentation)