Intel NPU (AI Boost)
Vendor: Intel
Category: NPU Neural Processor
Architecture: Gaudi NPU Core / Movidius
Introduction
Intel Core Ultra processor built-in NPU (AI Boost) for low-power AI inference acceleration. Supports OpenVINO framework, suitable for AI workloads on thin-and-light laptops and edge devices.
Specifications
| Model | Compute | Memory | Interface | TDP | Process |
|---|---|---|---|---|---|
| Core Ultra 200V | 48 TOPS (INT8) | LPDDR5X (shared) | Integrated SoC | 17-45W | Intel 4 |
| Core Ultra (Meteor Lake) | 12 TOPS (INT8) | LPDDR5X (shared) | Integrated SoC | 28W | Intel 4 |
| Movidius Myriad X | 4 TOPS (INT8) | 512MB LPDDR4 | USB 3.0 | 1.5W | 16nm |
Official Website
Driver Downloads
Windows
Linux
Related Documentation
OS Support
| Windows | Linux | macOS | Android |
|---|---|---|---|
| ✅ | ✅ | ❌ | ❌ |
Version History
| Version | Release Date | Description |
|---|---|---|
| OpenVINO 2024.4 | 2024 | Lunar Lake NPU support |
| OpenVINO 2024.0 | 2024-Q1 | LLM inference optimization |
Performance Benchmarks
| Model | Task | Performance Metric |
|---|---|---|
| Core Ultra 200V | NPU TOPS | 48 TOPS |
| Core Ultra (Meteor Lake) | NPU TOPS | 11 TOPS |
| Movidius Myriad X | Vision Inference | 4 TOPS, 1.5W |
Pricing
| Model | Reference Price | Notes |
|---|---|---|
| Core Ultra 200V | Provided with laptop | AI PC laptop |
| Movidius Myriad X | ~$69-99 | USB accelerator stick |
Quick Installation
Windows 11
Intel NPU is automatically used through the OpenVINO runtime with the latest drivers installed.
# Install OpenVINO
pip install openvino
# Verify NPU availability
python -c "from openvino import Core; core = Core(); print(core.available_devices)"
Code Examples
Python (OpenVINO NPU)
from openvino import Core
from openvino.runtime import Model
# Detect NPU device
core = Core()
print(f"Available devices: {core.available_devices}") # Contains 'NPU'
# Load and compile model to NPU
model = core.read_model("model.xml")
compiled = core.compile_model(model, "NPU")
# Inference
output = compiled(input_data)
Architecture Highlights
- Intel NPU Core: Low-power NPU core based on Gaudi technology, oriented toward daily AI tasks on AI PCs
- Movidius VPU: Independent VPU for edge vision inference, suitable for cameras/robots
- OpenVINO: Intel unified inference framework, automatically selecting the best CPU/GPU/NPU backend
Model Compatibility
| Model/Framework | Support | Notes |
|---|---|---|
| OpenVINO | ✅ Native | Best support |
| ONNX | ✅ Convert | OpenVINO direct load |
| PyTorch | ✅ Export | Convert ONNX → OpenVINO |
| Small LLMs | ✅ | OpenVINO + NPU acceleration |
| Speech/Vision | ✅ | Movidius vision optimized |
Related Products
If you're evaluating alternatives, the following products may also fit your scenario:
- AMD Ryzen AI NPU — AMD (NPU Neural Processor)
- Apple Neural Engine — Apple (NPU Neural Processor)
- Intel Gaudi 3 — Intel Habana (ASIC Dedicated Accelerator)
- Qualcomm Hexagon NPU — Qualcomm (NPU Neural Processor)
- Hailo-8 / Hailo-15 — Hailo (NPU Neural Processor)
- Huawei Ascend — Huawei (NPU Neural Processor)
- NVIDIA GPU / CUDA — NVIDIA (GPU Graphics Processor)