What is your goal?
Choose Your Path
Explore
Explore Applications
Explore AI applications across Smart Infra, Machine Vision, and Smart Retail verticals with ready-to-deploy models.
Explore
Explore Models
Find the right AI model for your use case with our interactive model comparison and selection tool.
Coming Soon
Sensor-to-AI Data Streaming Pipeline
Reference design demonstrating deterministic, real-time sensor data streaming to iGPU processing.
Coming Soon
Convert CUDA to ROCm
Migrate your CUDA applications to ROCm with AI-assisted code conversion tools and guided workflows.
Supported Hardware
Device Specifications
Purpose-built Ryzen AI processors for high-performance edge inference, supported via Vitis, MIGraphX, ROCm & Vulkan runtimes.
Strix Point / P100
Ryzen AI 300 Series
NPU Performance50 TOPS
CPU CoresUp
to 12C / 24T
iGPURadeon
800M
Best ForMainstream Edge AI
Strix Halo / X100
Ryzen AI Max Series
NPU Performance50 TOPS
CPU CoresUp
to 16C / 32T
iGPURadeon
8060S
Best ForHigh-perf Workstation AI
R8000
Embedded Series
NPU Performance16 TOPS
CPU CoresUp
to 8C / 16T
iGPURadeon
740M
Best ForEmbedded & IoT Edge
Understanding the Hardware
iGPU vs NPU
Ryzen AI devices feature both integrated GPU and a dedicated Neural Processing Unit. Here's how they compare.
iGPU
Integrated Graphics Processing Unit
- Massively parallel compute for graphics & general-purpose AI workloads
- Supports ONNX and Vulkan-based inference
- Higher throughput for large batch sizes and vision models
- Shared memory with CPU for zero-copy data transfer
NPU
Neural Processing Unit (XDNA)
- Dedicated AI accelerator with ultra-low power consumption
- Up to 50 TOPS for sustained, always-on AI inference
- Optimized for INT8/INT4 quantized models and transformers
- Frees GPU & CPU for other tasks during AI workloads