Description
LLM-8850 Kit is a high-performance AI accelerator kit designed for edge AI and embedded computing scenarios. It consists of the LLM-8850 Card AI accelerator card and the LLM-8850 PiHat adapter board. The core accelerator card LLM-8850 Card is based on the Axera AX8850 SoC, adopting the M.2 M-Key 2242 form factor. In a compact size, it delivers up to 24 TOPS @ INT8 computing power, efficiently extending AI application capabilities such as multimodal large model inference and video analysis. The card integrates an active cooling system with a micro turbine fan + aluminum alloy CNC finned heatsink, intelligently controlled by an onboard EC based on a temperature-current curve. Even under long-term full load or in a sealed enclosure, it maintains a low temperature steady state, preventing performance degradation due to thermal throttling. The accompanying LLM-8850 PiHat adapter board provides a 2.54-40P bus and a USB Type-C PD power input, and connects to the Raspberry Pi 5 via a PCIe expansion FPC adapter, allowing the Raspberry Pi platform to easily integrate high-computing-power AI acceleration and quickly build a compact, efficient edge intelligent system.
Features
- Ultra-compact form factor: M.2 M-Key 2242 size
- High-performance NPU: 24 TOPS @ INT8, octa-core Cortex‑A55 @ 1.7 GHz CPU
- Intelligent cooling/power: onboard turbine fan + CNC aluminum alloy integrated heatsink, EC monitors temperature-current-speed closed loop
- Hardware video engine: 8 K @ 30 fps H.264/H.265 encoding, 8 K @ 30 fps decoding, supports 16-channel 1080P parallel decoding
- Secure boot & encryption: AES/DES/3DES/SHA‑256 hardware security module
- Native AXCL: one-click running of full-stack models including CNN, Transformer, CLIP, Whisper, Llama3.2, Qwen3, InternVL3, supports simultaneous H.264/H.265 encoding/decoding and transcoding
- Memory: 4GB 64‑bit LPDDR4x, 4266 Mbps (2+2)
- USB Type-C PD power input:
- USB PD3.0 100W protocol
- DC 24V overvoltage protection
- Power supply requirement > 9V@3A (27W)
Includes
LLM-8850 Kit(4G) (SKU:AI-002-4G)
- 1x LLM-8850 Card(4G)
- 1x LLM-8850 PiHat
- 1x 2.54-40P Female Header
- 1x Screw and Copper Standoff Kit
Applications
- Edge AI inference deployment
- Embedded video analysis
- Raspberry Pi AI computing power expansion
- Multimodal large model deployment
Specifications
| Specification | Parameter |
|---|---|
| SoC | Axera AX8850 |
| CPU | Octa-core Cortex‑A55 1.7 GHz |
| NPU | 24 TOPS @ INT8 |
| Video Encoder | 8 K @ 30 fps H.264/H.265 encoding, supports scaling/cropping |
| Video Decoder | 8 K @ 30 fps H.264/H.265 decoding, supports 16-channel 1080P parallel decoding, supports scaling/cropping |
| Memory | Optional: 4GB 64‑bit LPDDR4x, 4266 Mbps (2+2) |
| Storage | 32Mbits QSPI NOR Flash (Bootloader only) |
| Form Factor | M.2 M-Key 2242, PCIe 2.0 ×2 |
| Cooling | Micro turbine fan + aluminum alloy CNC integrated heatsink, EC intelligent temperature control |
| Operating Ambient Temperature | 0 ~ 60 °C |
| Full Load Temperature at Room Ambient | 70 °C |
| Adapter Board PD Power Input | USB PD3.0 100W protocol (DC 9V/12V/20V) |
| Adapter Board Load Capability | Output to Raspberry Pi 5: DC 5V@4A, Output to M.2 accelerator card: DC 3.3V@6A |
| LLM-8850 Card Power Consumption | 7W @ 3.3V |
| Power Supply Requirement | > 9V@3A (27W) |
| Product Size | LLM‑8850 Card: 42.6 x 24.0 x 9.7mm LLM-8850 PiHat Adapter Board: 65.0 x 58.0 x 12.7mm (including LLM‑8850 Card) |
| Product Weight | AI-002-4G: 30.8g |
| Package Size | 165.0 x 90.0 x 17.5mm |
| Gross Weight | AI-002-4G: 81.7g |
Learn
System Compatibility
| Operating System | Supported |
|---|---|
| Ubuntu20.04 | ✅ |
| Ubuntu22.04 | ✅ |
| Ubuntu24.04 | ✅ |
| Debian12 | ✅ |
| Debian13 | ✅ |
| Windows10 | ✅ |
| Windows11 | ✅ |
| macOS | ❌ |
| WSL | ❌ |
| VMware | ❌ |
| VBox | ❌ |
Schematics
Model Size
Datasheets
Softwares
Quick Start
Video
- LLM‑8850 Kit Product Introduction and Feature Demonstration