Section 1
Minimum Requirements
Operating System
Ubuntu 22.04 LTS
Required. Not Windows, not macOS.
GPU
NVIDIA RTX (Ampere+)
Compute capability 8.0 or higher
VRAM
8 GB minimum
System RAM
64 GB minimum
CUDA
12.1+
Storage
50 GB software + 10 GB assets
Version Control
Git + Git LFS
Required for large file management
Section 2
Recommended: Demo Station
Dell Alienware 18 Area-51
Portable workstation for development and demonstrations
CPU
Intel Core Ultra 9 275HX
GPU
NVIDIA RTX 5090 (24 GB GDDR7)
RAM
64 GB DDR5-6400
Storage
2 TB NVMe
Display
18" WQXGA 300Hz
Price
~$4,000 - $5,000
- Portable -- take to demos and conferences
- Full Omniverse GUI with display
- x86 architecture -- zero compatibility issues
- Loud under sustained GPU load
- Heavy at 4+ kg
Best for: Demos, development, single-vehicle to small fleet simulation
Section 3
Research Hardware
DGX Spark
Compact desktop AI workstation
GPU
GB10 Blackwell Superchip
Memory
128 GB unified LPDDR5x
Architecture
ARM (aarch64)
Price
~$4,999
- Quiet, compact desktop form factor
- 128 GB unified memory for large datasets
- ARM architecture -- compatibility unknowns
- Headless only -- no display output
- PhysX GPU issues have been reported
Best for: Headless batch simulation, medium-scale experiments
DGX GB300 NVL72
Enterprise-scale AI supercomputer (available at NPS)
GPUs
72 Blackwell Ultra GPUs
GPU Memory
20+ TB total (288 GB HBM3e each)
Performance
1+ exaFLOP (FP4)
Power
~120 kW
Cooling
Direct liquid cooling required
Best for: Fleet-scale simulation (100+ vehicles), Monte Carlo analysis, RL training at scale, multi-student shared resource
Section 4
Decision Matrix
Which hardware for which task.
| Task | Recommended Hardware |
|---|---|
| Quick demo for visitors | Alienware Laptop |
| Development and testing | Alienware Laptop |
| Single thesis research | Alienware Laptop or DGX Spark |
| Fleet-scale simulation | DGX GB300 |
| Monte Carlo / RL training | DGX GB300 |
| Shared department resource | DGX GB300 |