MemryX MX3 M.2 2280 (MX3-2280-M-4) 24-TFLOPS At-Memory AI Card
M.2 2280 M-Key edge AI accelerator module cascading four MemryX MX3 At-Memory Dataflow chips to deliver 24 TFLOPS/TOPS of mixed BF16/INT8 compute at ~8W typical power (~2W per chip) with zero external DRAM required and push-button ONNX/PyTorch compilation.
Editorial Rating Methodology (★ 4.8 / 5.0): This score is an internal editorial evaluation compiled by the RECIWO Labs editorial team—not a user poll or crowd-voted rating. It assesses published silicon compute efficiency (25%), optical/sensor/actuator engineering (25%), battery & thermal envelope (20%), SDK/software openness (15%), and price-to-specification value (15%) based on official manufacturer datasheets and public technical documentation. Learn more about our evaluation rubric →
Key Engineering Highlights
- Eliminates external LPDDR DRAM bottlenecks entirely—storing 42M to 80M model weights directly in on-die SRAM across 4 cascaded MX3 chips for deterministic microsecond latency
- Computes activations in BFloat16 (BF16) with INT4/INT8/INT16 weights, delivering 24 TFLOPS/TOPS within an ~8W typical thermal budget
- MemryX Developer Hub compiles standard PyTorch, ONNX, TensorFlow, and TFLite models in minutes with zero manual quantization calibration or retraining
Full Hardware Specifications
| Processor / Edge SoC | 4× MemryX MX3 At-Memory Spatial Dataflow AI Accelerators (6 TFLOPS/TOPS per Chip) |
|---|---|
| AI Model Stack | 1,000+ Validated Vision, Depth, Pose & ViT Models (42M–80M Params Resident in On-Die SRAM) |
| Display & Optics | Host Pass-Through (Standard M.2 2280 M-Key Module for x86 & ARM64 Motherboards) |
| Camera & Vision | Accelerates Multi-Stream 1080p/4K RTSP, USB & MIPI Vision Pipelines via Host CPU |
| Measured Weight | 13.0g (22mm × 80mm M.2 2280 PCB — Zero External LPDDR Chips Required) |
| Battery & Runtime | ~8.0W Typical Card Power (~2.0W per MX3 Chip / 14W M.2 Slot Peak Limit) |
| Connectivity & I/O | PCIe Gen 3 (2×2-Lane) M.2 2280 M-Key (Linux & Windows C++/Python Driver Support) |
| Sensors & Acoustics | Distributed On-Die At-Memory SRAM Weight Storage + BF16 Activation Dataflow Fabric |
⚖️ Side-by-Side Comparison: MemryX MX3 M.2 2280 (MX3-2280-M-4) 24-TFLOPS At-Memory AI Card vs. Edge Compute & DevKits Peers
Direct specification comparison against leading competing devices in the Edge Compute & DevKits category.
| Specification | MemryX MX3 M.2 2280 (MX3-2280-M-4) 24-TFLOPS At-Memory AI Card (Current) | NVIDIA Jetson Orin NX Super Edge AI DevKit → | Xiaozhi AI Voice Cube Pro (ESP32-S3 Open DevKit) → | Apple Mac mini (M5 Pro 64GB) Private AI Node → |
|---|---|---|---|---|
| Brand | MemryX | NVIDIA | Open Hardware Community | Apple |
| Retail Price | $149 | $499 | $12 | $1,799 |
| Editorial Rating | ||||
| Edge SoC | 4× MemryX MX3 At-Memory Spatial Dataflow AI Accelerators (6 TFLOPS/TOPS per Chip) | NVIDIA Ampere GPU (1024 CUDA + 32 Tensor Cores) | Espressif ESP32-S3R8 (Dual-Core LX7 @ 240MHz + 8MB PSRAM) | Apple M5 Pro (14c CPU + 20c GPU + 16c Neural Engine) |
| AI Model Stack | 1,000+ Validated Vision, Depth, Pose & ViT Models (42M–80M Params Resident in On-Die SRAM) | Runs Llama-3.2 8B / Qwen2.5-VL 7B / YOLO11 Locally | Edge ESP-SR Wake + Realtime Cloud Voice LLMs | MLX Accelerated Qwen3-32B / Llama-3.3-70B |
| Display / Optics | Host Pass-Through (Standard M.2 2280 M-Key Module for x86 & ARM64 Motherboards) | HDMI 2.1 / DP 1.4 (Up to 8K) | 1.28-inch Circular IPS LCD (240×240) / 0.96" OLED | Supports up to three 6K displays |
| Camera | Accelerates Multi-Stream 1080p/4K RTSP, USB & MIPI Vision Pipelines via Host CPU | Dual MIPI CSI-2 Camera Lanes + 4× USB 3.2 | Optional OV2640 / GC2145 Camera Module | External USB-C / Studio Display Camera |
| Weight | 13.0g (22mm × 80mm M.2 2280 PCB — Zero External LPDDR Chips Required) | 450g (including active heatsink & carrier board) | 62g (with 3D-printed enclosure & speaker cavity) | 730g |
| Battery | ~8.0W Typical Card Power (~2.0W per MX3 Chip / 14W M.2 Slot Peak Limit) | 19V DC Input (15W–40W configurable TDP) | 800mAh LiPo (5h continuous dialogue) | Internal 155W PSU (4W idle draw) |
📚 Related Engineering Teardowns & Reviews
Browse All 113 Articles →MemryX MX3 vs Kinara Ara-2 (NXP) vs Axelera Metis M.2 Comparison: DRAM-Free At-Memory Dataflow vs 16GB 40-TOPS Edge GenAI NPU
A technical silicon architecture comparison of the MemryX MX3 M.2 2280 ($149, 4-chip 24-TFLOPS BF16/INT8 At-Memory Dataflow accelerator requiring zero external DRAM at ~8W), Kinara Ara-2 / NXP Ara-240 M.2 ($188, 40-TOPS polymorphic Edge GenAI NPU with 16GB LPDDR4X for local LLMs and VLMs), and Axelera Metis M.2 Max ($189).
HP Z2 Mini G1a (Strix Halo 128GB) vs ASUS Ascent GX10 (GB10) vs NVIDIA DGX Spark Comparison: 96GB UMA VRAM vs 1-PFLOP FP4
A technical architecture and specification comparison of the HP Z2 Mini G1a ($2,299, 2.9L AMD Ryzen AI Max+ PRO 395 workstation with 128GB LPDDR5X-8533 and 96GB allocatable UMA VRAM), ASUS Ascent GX10 ($3,499, NVIDIA GB10 Grace Blackwell Superchip with 1,000 FP4 TOPS, 128GB unified RAM, and 200GbE ConnectX-7), Framework Desktop ($1,999), and NVIDIA DGX Spark ($3,999).
Astribot S1 vs MagicLab MagicBot Gen1 vs Galbot G1 Comparison: 10 m/s Cable-Driven VLA Arms vs 20kg Factory Bimanual Humanoid
A technical specification and actuation comparison of the Astribot S1 ($69,000, 170cm omnidirectional wheeled bimanual robot with ≥10 m/s cable-driven end-effector velocity, 100 m/s² acceleration, ±0.1mm repeatability, and 5kg/arm full-reach payload), MagicLab MagicBot Gen1 ($42,000, 42-DoF bipedal industrial humanoid with 350 N·m actuators and 20kg dual-arm payload), and Galbot G1 ($38,000).
Captify Pro vs Chamelo Music Shield vs Vuzix Z100 Comparison: 37g Binocular Live-Captioning HUD vs 0.1s Electrochromic Audio Glasses
A technical specification and optical architecture comparison of the Captify Pro ($699, 37g binocular 1,500-nit MicroLED waveguide smart glasses with acoustic beamforming for deaf/hard-of-hearing live captioning and 40+ language translation), Chamelo Music Shield ($249, 38g 0.1-second Eclipse liquid-crystal electrochromic tint-shifting AI audio sports glasses), and Vuzix Z100 ($499).
