DEEPX AI Chip Line-up
Powering the Future of Edge AI

DX-M Series
DX-M1

DX-M Series
DX-M1M
(Coming Soon)

DX-V3
(Coming Soon)

DX-M Series
DX-M2
(Coming Soon)

DX-M1
Key Features
Delivering breakthrough performance while consuming significantly less power than competitors, enabling longer battery life for mobile devices and reduced energy costs for data centers. (Butter-Proof)
Purpose-built dedicated DRAM enables simultaneous multi-model processing, allowing your applications to run multiple AI models concurrently without performance degradation.
Engineered for seamless integration with any host CPU architecture—zero compatibility issues means faster deployment, simplified development, and protection for your existing infrastructure investments.
Our innovative design minimizes expensive on-chip SRAM requirements, delivering substantial cost advantages without compromising performance—making advanced AI accessible at scale.
AI Performance
AI Model | Input Resolution | Accuracy (mAP)
FP32 GPU | Accuracy (mAP) *IQ8 DX-M1 | FPS DX-M1 | NPU Power (W) DX-M1 | FPS/W DX-M1 |
---|---|---|---|---|---|---|
YOLOv5m6 | 640×640 | 45.08 | 45.07 | 40.42 | 2.20 | 18.37 |
YOLOv7-e6 | 640×640 | 55.58 | 55.45 | 19.49 | 2.50 | 7.81 |
YOLOv8x | 640×640 | 53.63 | 53.12 | 49.03 | 3.17 | 15.48 |
YOLOv8l | 640×640 | 52.57 | 52.01 | 91.13 | 3.60 | 25.32 |
YOLOv8m | 640×640 | 50.11 | 49.56 | 133.60 | 2.76 | 48.32 |
YOLOv9c | 640×640 | 52.86 | 52.36 | 47.89 | 2.83 | 16.95 |
- *This is the test result of Pro compilation among Pro and Master of DEEPX compiler.
- *IQ8: High-accuracy AI processing using efficient INT8 operations.
Successful Use Cases


Storage Server

Robotics Platform

Computing System

DX-M1M
Ultra-Compact AI Processing with Integrated Memory. The DX-M1M builds on the exceptional foundation of our flagship DX-M1 AI accelerator by integrating high-performance LPDDR memory directly onto the same die. This design delivers the same industry-leading AI performance while dramatically reducing the overall footprint required for deployment.
TBU
TBU

DX-V3
DX-V3 is a powerful AI vision processor offering exceptional computational efficiency and versatile capabilities. This all-in-one solution integrates a CPU, ISP, DSP, video codec, and high-performance NPU, delivering 13 TOPS of AI power with industry-leading efficiency. Optimized for next-gen vision systems, DX-V3 processes imagery, video streams, and 3D point-cloud data with ease.

DX-M2
DX-M2 is a specialized semiconductor designed specifically for efficient processing of GenAI & language models with up to 20 billion parameters. Delivering impressive performance of 25-30 TPS (Tokens Per Second), this processor enables deployment of sophisticated language models in a variety of edge and enterprise applications without sacrificing response time or accuracy.