Hardware Requirements
This course supports multiple hardware configurations. Choose the tier that fits your budget and learning goals.
Learning Objectives
By reading this section, you will be able to:
- Select the appropriate hardware tier for your learning goals and budget
- Understand minimum requirements for running NVIDIA Isaac Sim
- Compare cloud vs on-premise development workflows
- Identify recommended robot platforms for the capstone project
Tier 1: High-Performance Workstation (Full Course)
The recommended setup for running NVIDIA Isaac Sim and training RL policies locally.
| Component | Minimum | Recommended |
|---|---|---|
| GPU | RTX 4070 Ti (12GB VRAM) | RTX 4090 (24GB VRAM) |
| CPU | Intel i7-12700 / AMD Ryzen 7 | Intel i9-13900K / Ryzen 9 |
| RAM | 32 GB DDR5 | 64 GB DDR5 |
| Storage | 512 GB NVMe SSD | 2 TB NVMe SSD |
| OS | Ubuntu 22.04 LTS | Ubuntu 22.04 LTS |
Estimated cost: $2,000–$4,000
# Check your GPU compatibility for Isaac Sim
nvidia-smi
# Minimum: CUDA 11.8+, Driver 525+
# Recommended: CUDA 12.x, Driver 545+
# Check VRAM
nvidia-smi --query-gpu=memory.total --format=csv,noheader,nounits
# Should output: 12288 or higher (12 GB minimum)
Tier 2: Economy Student Kit (~$700)
An affordable setup using NVIDIA Jetson for edge AI development.
| Component | Specification | Price (approx.) |
|---|---|---|
| NVIDIA Jetson Orin Nano 8GB | ARM Cortex-A78AE, 1024-core Ampere GPU | ~$250 |
| Intel RealSense D435i | Stereo depth camera + IMU | ~$150 |
| USB Hub + Power | Powered 4-port USB hub | ~$30 |
| MicroSD + NVMe | 256GB storage | ~$50 |
| Cooling + Case | Active cooling kit | ~$40 |
| Peripherals | Monitor/keyboard (optional) | ~$150 |
Total: ~$670–$720
# Check Jetson specs programmatically
import subprocess
result = subprocess.run(['cat', '/proc/device-tree/model'], capture_output=True, text=True)
print(result.stdout) # e.g., "NVIDIA Jetson Orin Nano 8GB Developer Kit"
# Check JetPack version
result = subprocess.run(['cat', '/etc/nv_tegra_release'], capture_output=True, text=True)
print(result.stdout)
# Required: JetPack 6.0+ (based on Ubuntu 22.04, CUDA 12.2)
Tier 3: Robot Platforms
Option A: Unitree G1 (Full Humanoid)
- Type: Full-size bipedal humanoid
- Payload: 3 kg
- Height: 127 cm
- Sensors: Depth cameras, LiDAR, IMU
- ROS 2: Supported via Unitree SDK
- Price: ~$16,000
Option B: Unitree Go2 (Quadruped, Learning Platform)
- Type: Quadruped robot dog
- Payload: 3 kg
- ROS 2: Full support via
unitree_ros2 - Price: ~$1,600 (Air) – $3,200 (Pro)
- Best for: Navigation, SLAM, manipulation arm add-on
Option C: Hiwonder AiArm (Desktop Arm)
- Type: 6-DOF robotic arm
- Reach: 280mm
- Payload: 250g
- ROS 2: Supported
- Price: ~$400
- Best for: Manipulation tasks, pick & place capstone
Cloud vs On-Premise Comparison
| Factor | Cloud (Vast.ai / AWS) | On-Premise (Workstation) |
|---|---|---|
| Initial Cost | $0 (pay per use) | $2,000–$4,000 |
| Cost for Isaac Sim (8h/day, 30 days) | ~$200–$400/month | ~$15–$20/month electricity |
| Latency | High (remote display) | Zero |
| Data Privacy | External servers | Local |
| Scalability | Infinite | Single machine |
| Recommendation | Modules 1–3 training runs | Module 4 capstone demos |
Software Requirements (All Tiers)
# Ubuntu 22.04 LTS (required)
lsb_release -a
# ROS 2 Humble (required for all modules)
sudo apt install ros-humble-desktop
# Python 3.10+ (required)
python3 --version
# NVIDIA Container Toolkit (for Isaac ROS)
sudo apt install nvidia-container-toolkit
# Verify CUDA
nvcc --version