Hardware Requirements Guide
This guide outlines the hardware needed for the Physical AI & Humanoid Robotics curriculum. We provide three tiers to accommodate different budgets and learning goals.
Learning Outcomes
By reviewing this guide, you will be able to:
- Select appropriate hardware for your learning goals
- Understand the trade-offs between local and cloud development
- Plan a budget for Physical AI development
- Identify essential vs. optional components
Hardware Tiers Overview
| Tier | Purpose | Budget | Best For |
|---|---|---|---|
| Tier 1: Cloud | Remote development | $50-200/month | Students, beginners |
| Tier 2: Workstation | Local development | $3,000-5,000 | Serious learners |
| Tier 3: Full Lab | Complete development | $20,000+ | Research, production |
Tier 1: Cloud Development
For learners without access to high-end hardware, cloud computing provides a cost-effective alternative.
Recommended Cloud Services
| Service | Instance Type | GPU | Cost/Hour | Monthly (40hr/week) |
|---|---|---|---|---|
| AWS | g5.2xlarge | A10G (24GB) | $1.21 | ~$194 |
| Azure | NC6s_v3 | V100 (16GB) | $0.90 | ~$144 |
| Google Cloud | n1-standard-4 + T4 | T4 (16GB) | $0.35 | ~$56 |
| Lambda Labs | gpu_1x_a10 | A10 (24GB) | $0.60 | ~$96 |
Cost Optimization
- Use spot/preemptible instances for 60-80% savings
- Stop instances when not in use
- Use smaller instances for coding, larger for training
Cloud Setup for Isaac Sim
# AWS g5.2xlarge setup
# 1. Launch Ubuntu 22.04 AMI with g5.2xlarge
# 2. Install NVIDIA drivers
sudo apt-get update
sudo apt-get install -y nvidia-driver-535
# 3. Install Omniverse and Isaac Sim
# Follow NVIDIA's cloud deployment guide
Tier 2: Digital Twin Workstation
For serious development, a local workstation provides the best experience.
Minimum Specifications
| Component | Minimum | Recommended | Notes |
|---|---|---|---|
| GPU | RTX 4070 Ti (12GB) | RTX 4090 (24GB) | Isaac Sim requires RTX |
| CPU | Intel i7-12700 / Ryzen 7 5800X | Intel i9-13900K / Ryzen 9 7950X | Physics runs on CPU |
| RAM | 32GB DDR4 | 64GB DDR5 | Complex scenes need more |
| Storage | 500GB NVMe SSD | 2TB NVMe SSD | Isaac Sim is ~30GB |
| OS | Ubuntu 22.04 LTS | Ubuntu 22.04 LTS | Windows not recommended |
Recommended Build (~$3,500)
| Component | Model | Price |
|---|---|---|
| GPU | NVIDIA RTX 4070 Ti Super 16GB | $800 |
| CPU | AMD Ryzen 9 7900X | $400 |
| Motherboard | ASUS ROG Strix X670E-F | $350 |
| RAM | 64GB DDR5-5600 (2x32GB) | $180 |
| Storage | Samsung 990 Pro 2TB NVMe | $180 |
| PSU | Corsair RM850x | $140 |
| Case | Fractal Design Meshify 2 | $150 |
| Cooling | Noctua NH-D15 | $100 |
| Total | ~$2,300 |
Power Requirements
RTX 4090 systems require 850W+ PSU. Ensure adequate cooling—sustained GPU loads generate significant heat.
Software Stack
# Ubuntu 22.04 setup
# 1. Install NVIDIA drivers
sudo apt-get update
sudo apt-get install -y nvidia-driver-535
# 2. Install CUDA Toolkit
wget https://developer.download.nvidia.com/compute/cuda/12.2.0/local_installers/cuda_12.2.0_535.54.03_linux.run
sudo sh cuda_12.2.0_535.54.03_linux.run
# 3. Install ROS 2 Humble
sudo apt install software-properties-common
sudo add-apt-repository universe
sudo apt update && sudo apt install curl -y
sudo curl -sSL https://raw.githubusercontent.com/ros/rosdistro/master/ros.key -o /usr/share/keyrings/ros-archive-keyring.gpg
echo "deb [arch=$(dpkg --print-architecture) signed-by=/usr/share/keyrings/ros-archive-keyring.gpg] http://packages.ros.org/ros2/ubuntu $(. /etc/os-release && echo $UBUNTU_CODENAME) main" | sudo tee /etc/apt/sources.list.d/ros2.list > /dev/null
sudo apt update
sudo apt install ros-humble-desktop
# 4. Install Isaac Sim via Omniverse Launcher
# Download from https://www.nvidia.com/en-us/omniverse/
Tier 3: Physical AI Edge Kit
For deploying to real robots, you need edge computing hardware.
NVIDIA Jetson Options
| Model | GPU Cores | RAM | AI Performance | Price |
|---|---|---|---|---|
| Jetson Orin Nano | 1024 | 8GB | 40 TOPS | $499 |
| Jetson Orin NX 8GB | 1024 | 8GB | 70 TOPS | $699 |
| Jetson Orin NX 16GB | 1024 | 16GB | 100 TOPS | $899 |
| Jetson AGX Orin 32GB | 2048 | 32GB | 200 TOPS | $1,599 |
| Jetson AGX Orin 64GB | 2048 | 64GB | 275 TOPS | $1,999 |
Jetson Student Kit (~$700)
Complete edge development kit for students:
| Component | Model | Price |
|---|---|---|
| Compute | Jetson Orin Nano Developer Kit | $499 |
| Camera | Intel RealSense D435i | $350 |
| Storage | Samsung 256GB NVMe | $40 |
| Power | 65W USB-C Power Supply | $30 |
| Cables | USB-C, Ethernet, HDMI | $30 |
| Case | 3D Printed Enclosure | $20 |
| Total | ~$970 |
Student Discounts
NVIDIA offers educational discounts on Jetson hardware. Check with your institution for academic pricing.
Sensor Options
| Sensor | Purpose | Price | Notes |
|---|---|---|---|
| Intel RealSense D435i | RGB-D + IMU | $350 | Best value for indoor |
| Intel RealSense D455 | RGB-D + IMU | $450 | Longer range |
| Stereolabs ZED 2i | Stereo + IMU | $499 | Better outdoor |
| Velodyne VLP-16 | 3D LiDAR | $4,000 | Outdoor navigation |
| RPLidar A1 | 2D LiDAR | $100 | Budget indoor |
Tier 4: Robot Lab (Optional)
For hands-on robot development:
Humanoid Robot Options
| Robot | DOF | Height | Features | Price |
|---|---|---|---|---|
| Unitree G1 | 23 | 1.3m | Walking, manipulation | ~$16,000 |
| Unitree H1 | 19 | 1.8m | Fast walking | ~$90,000 |
| Agility Digit | 16 | 1.6m | Industrial use | ~$250,000 |
Alternative Platforms
For learning without full humanoids:
| Platform | Type | Price | Best For |
|---|---|---|---|
| Unitree Go2 | Quadruped | $1,600 | Locomotion basics |
| Franka Emika | Arm | $20,000 | Manipulation |
| TurtleBot 4 | Mobile base | $1,200 | Navigation |
| Open Manipulator X | Arm | $500 | Budget manipulation |
Cloud Cost Calculator
Estimate your monthly cloud costs:
| Activity | Hours/Week | Instance | Cost/Hour | Monthly |
|---|---|---|---|---|
| Coding/Testing | 20 | t3.medium | $0.04 | $3.20 |
| Simulation | 10 | g5.2xlarge | $1.21 | $48.40 |
| Training | 5 | g5.4xlarge | $2.42 | $48.40 |
| Total | 35 | ~$100 |
Spot Instances
Using spot instances can reduce costs by 60-80%:
- g5.2xlarge spot: ~$0.40/hr (vs $1.21 on-demand)
- Best for interruptible workloads like training
Recommended Learning Path
Phase 1: Foundations (Weeks 1-5)
- Hardware: Cloud (g5.2xlarge) or laptop with GPU
- Cost: ~$50-100/month cloud or existing hardware
Phase 2: Simulation (Weeks 6-10)
- Hardware: Workstation with RTX 4070 Ti+ or cloud
- Cost: ~$100-200/month cloud or one-time workstation
Phase 3: Edge Deployment (Weeks 11-13)
- Hardware: Jetson Orin Nano + RealSense camera
- Cost: ~$850 one-time
Phase 4: Robot Integration (Capstone)
- Hardware: Access to robot platform (lab, rental, or purchase)
- Cost: Varies widely
Assessment
Recall
- What is the minimum GPU VRAM required for Isaac Sim?
- What are the advantages of Jetson Orin over cloud deployment?
- Why is Ubuntu recommended over Windows for robotics development?
Apply
- Design a hardware setup for a $2,000 budget that can run Isaac Sim.
- Calculate the 6-month cost of cloud development vs. purchasing a workstation.
- Select sensors for a humanoid robot that needs to navigate indoors and manipulate objects.
Analyze
- Compare the trade-offs between Jetson Orin Nano and Jetson AGX Orin for a mobile robot.
- When would cloud development be more cost-effective than local hardware?
- Design a phased hardware acquisition plan for a robotics lab with a $50,000 annual budget.