RM1828MC0
M.2 2280 AI accelerator with 20 TOPS NPU and 5GB memory

- Interface Type
- M.2 M-KEY
- RAM
- On-chip 5GB DRAM
- OS
- Linux
- PCIe
- 1 × PCIe2.1 (multiplexed with USB3.0)
M.2 2280 AI accelerator with 20 TOPS NPU and 5GB memory
Maker, Prototyping, Open-Hardware & Developer Boards
M.2 2280 AI accelerator with a 20 TOPS (INT8) NPU and 5GB of DRAM, designed for running vision and language models locally.
Overview. The Firefly RM1828MC0 is an AI accelerator in M.2 2280 format (M-KEY interface) built around the Rockchip RK1828 SoC. It integrates a 20 TOPS (INT8) NPU and 5GB of on-chip DRAM, enabling local execution of vision and language models.
Main characteristics. The processor is a 64-bit tri-core RISC-V with FPU, clocked up to 850 MHz. The NPU supports mixed-precision computation in INT4, INT8, INT16, FP8, FP16 and BF16. The reference operating system is Linux.
Interfaces and connectivity. The module provides one PCIe 2.1 interface (multiplexed with USB 3.0), one USB 3.0 DRD port multiplexed with PCIe, and one USB 2.0 DRD port.
Power and environment. 12 V DC power supply. Operating temperature from -20 ℃ to 60 ℃, storage temperature from -20 ℃ to 70 ℃, storage humidity from 10% to 90% RH non-condensing.
Typical uses. YOLO object-detection algorithms are supported, as is private deployment of large Transformer-based models such as Gemma, ChatGLM, Qwen and Phi. CNN, RNN and LSTM architectures are also supported.
Software integration. RKNN model import and export, compatibility with TensorFlow, TensorFlow Lite, PyTorch, Keras and PaddlePaddle, custom operator development and Docker containerized management.
Dimensions and weight. 22mm × 80mm × 6mm without heat sink, 25mm × 81mm × 23.05mm with heat sink; 11g without heat sink, 53g with heat sink.
No published question about this part yet. Questions are reviewed before publication.
Every characteristic recorded for RM1828MC0, with where it comes from. Use the filters to find one quickly.
| Characteristic | Value |
|---|---|
| CPU | Tripe-core 64-bit RISC-V processor (with FPU), max. clock frequency 850 MHz |
| Dimension | 22mm × 80mm × 6mm (without heat sink); 25mm × 81mm × 23.05mm (with heat sink) |
| Environment | Operating temperature: -20℃ ~ 60℃, Storage temperature: -20℃ ~ 70℃, Storage humidity: 10% ~ 90%RH (non-condensing) |
| Interface Type | M.2 M-KEY |
| NPU | NPU computing power: 20 TOPS (INT8); supports mixed-precision computation of INT4/INT8/INT16/FP8/FP16/BF16 |
| OS | Linux |
| PCIe | 1 × PCIe2.1 (multiplexed with USB3.0) |
| Power | DC 12V |
| RAM | On-chip 5GB DRAM |
| SOC | Rockchip RK1828 |
| Software Support | Supports YOLO object-detection algorithms.; Supports private deployment of large-scale Transformer-based models, such as Gemma, ChatGLM, Qwen, Phi and other LLM series.; Supports classic network architectures including CNN, RNN and LSTM; supports RKNN model import & export; compatible with multiple deep-learning frameworks: TensorFlow, TensorFlow Lite, PyTorch, Keras, PaddlePaddle, etc.; Supports custom operator development and Docker containerized management |
| USB | 1 × USB3.0 DRD (multiplexed with PCIe2.1, PCIe by default), 1 × USB2.0 DRD |
| Weight | 11g (without heat sink), 53g (with heat sink) |
No specification matches.
Last verification : 2026-11-02
Official documents and the technical data recorded for RM1828MC0. For a missing document or a specific revision, our experts send it to you.
M.2 2280 AI accelerator with a 20 TOPS (INT8) NPU and 5GB of DRAM, designed for running vision and language models locally.
| Parameter | Value |
|---|---|
| CPU | Tripe-core 64-bit RISC-V processor (with FPU), max. clock frequency 850 MHz |
| Dimension | 22mm × 80mm × 6mm (without heat sink); 25mm × 81mm × 23.05mm (with heat sink) |
| Environment | Operating temperature: -20℃ ~ 60℃, Storage temperature: -20℃ ~ 70℃, Storage humidity: 10% ~ 90%RH (non-condensing) |
| Interface Type | M.2 M-KEY |
| NPU | NPU computing power: 20 TOPS (INT8); supports mixed-precision computation of INT4/INT8/INT16/FP8/FP16/BF16 |
| OS | Linux |
| PCIe | 1 × PCIe2.1 (multiplexed with USB3.0) |
| Power | DC 12V |
| RAM | On-chip 5GB DRAM |
| SOC | Rockchip RK1828 |
| Software Support | Supports YOLO object-detection algorithms.; Supports private deployment of large-scale Transformer-based models, such as Gemma, ChatGLM, Qwen, Phi and other LLM series.; Supports classic network architectures including CNN, RNN and LSTM; supports RKNN model import & export; compatible with multiple deep-learning frameworks: TensorFlow, TensorFlow Lite, PyTorch, Keras, PaddlePaddle, etc.; Supports custom operator development and Docker containerized management |
| USB | 1 × USB3.0 DRD (multiplexed with PCIe2.1, PCIe by default), 1 × USB2.0 DRD |
| Weight | 11g (without heat sink), 53g (with heat sink) |
Answers to the questions that the recorded characteristics of RM1828MC0 can answer. A question is not asked when nothing is recorded to answer it.
Yes — recorded under "USB": 1 × USB3.0 DRD (multiplexed with PCIe2.1, PCIe by default), 1 × USB2.0 DRD.
Similar parts to answer your needs. Compare the main characteristics. A recorded replacement is a documented relation; a similar product is a similarity only, never a compatibility claim.
M.2 2280 AI accelerator with 20 TOPS NPU and 5GB memory

Sensor Adapter for Click, QWIIC and Grove Modules

RS485 Module 3PEAK TP8485E

4-20mA Module STMicroelectronics LM2902

Describe your application and constraints in a request.
What is recorded about the regulatory status and the life of this part.
No compliance statement is recorded for this part. Ask the supplier for the RoHS and REACH declarations and, for a radio product, the EU declaration of conformity.
| Current status | Status not stated |
| Key dates | Not recorded |
| Manufacturer part number | RM1828MC0 |
| Manufacturer | Firefly |
Maker, Prototyping, Open-Hardware & Developer Boards
M12 fisheye lens for camera module · Ref. : M12_FISHEYE
Price on request
Maker, Prototyping, Open-Hardware & Developer Boards
RK3568 System-on-Module with 1 TOPS NPU
Price on request
Maker, Prototyping, Open-Hardware & Developer Boards
RK3568 Android tablet mainboard with DTV tuner · Ref. : K3-3568
Price on request
Maker, Prototyping, Open-Hardware & Developer Boards
Boost converter module 3.3 V to 12 V 50 mA · Ref. : 110155
Price on request
No advertising or analytics cookies. A single, strictly necessary cookie is set only if you sign in to your account. Your quote basket and comparison stay in your browser. Details