Ryzen AI NUC Demo Plan

In progress · updated 2026-10-30 · embodied-ai-architect/docs/plan/ryzen-ai-nuc-demo.md

Target Platforms

Platform	CPU	GPU	NPU	TDP
Ryzen 7 AI NUC	8-core Zen 4	Radeon 780M (12 CU)	XDNA 10 TOPS	15-28W
Ryzen 9 AI NUC	8-core Zen 4+	Radeon 780M (12 CU)	XDNA 16 TOPS	35-54W

Target Application: PGN&C Pipeline

  Camera → Preprocess → Detect → Track → Scene Graph
    30Hz      GPU         NPU     CPU        CPU
                                    ↓
  IMU ──────────────────────→ State Est. → Planner → Controller
  100Hz                          CPU         CPU       CPU 100Hz

8-Week Implementation Plan

Phase	Weeks	Deliverables
1. Platform Characterization	1-2	Hardware profiles, baseline benchmarks
2. Operator Profiling	2-3	Performance profiles for 15+ operators
3. Reference Architecture	3-4	pgnc_robot_v1 with 4 variants
4. Operator Implementations	4-6	Runnable YOLO (NPU), EKF, PID operators
5. Benchmark Harness	5-6	Real execution with power measurement
6. Demo Workflow	6-7	Interactive agentic design session
7. Documentation	7-8	Report, demo video, polish

Demo Workflow Preview

  User: "Deploy a perception-control pipeline on Ryzen 7 AI NUC, 30fps, 25W"

  Agent: [analyze_architecture] → "48ms latency, only 20fps - below target"
  Agent: [identify_bottleneck] → "Detector on GPU takes 32ms (67%)"
  Agent: [suggest_optimizations] → "Move to NPU: 32ms → 12ms"
  Agent: [benchmark_architecture] → "Measured: 29.3ms, 34fps, 21.3W ✓"

  User: "Compare Ryzen 7 vs Ryzen 9"

  | Metric | Ryzen 7 | Ryzen 9 |
  |--------|---------|---------|
  | Throughput | 34 fps | 45 fps |
  | Power | 21W | 34W |
  | Efficiency | 1.6 fps/W | 1.3 fps/W |

Key Technical Work

1. Hardware catalog entries for both NUCs in embodied-schemas
2. Real operator profiles measured on NPU/GPU/CPU
3. Runnable operators: YOLO via ONNX Runtime + Ryzen AI EP
4. Power monitoring via ryzen_smu or RAPL
5. Architecture runner for real pipeline execution

Success Criteria

• Pipeline runs on both platforms
• Estimates within 15% of measured
• Optimization suggestions actually improve performance
• Clear before/after demonstration