Specifications
NPU
ARIES
Performance
80 TOPS
NPU Frequency
1.25GHz
NPU Cores
8
Memory Type
LPDDR4X
Memory Capacity
16 GB / 32 GB
Memory Bandwidth
66.7GB/s
Flash Memory
32 MB
Host Interface
PCI Express Gen4 8-lane
UART
1 Port
LED Indicators
2 (Power, Heartbeat)
Supported OS
Linux, Windows
Supported Frameworks
PyTorch, TensorFlow, TFLite, ONNX, Keras
TDP
25 W
Input Voltage
DC +12 V
Cooling System
Heatsink, Fan cooling
Dimensions (mm)
Type A: 82 (L) × 70 (W)
Weight (g)
110
Where MLA100 MXM Fits
Industrial Embedded Systems
Industrial controllers, inspection systems, and factory automation equipment.
Drones and UAVs
Advanced LLM and vision workload within UAV payload and power constraints.
AIoT and Home Appliances
On-device voice, vision and language processing for smart appliances, home robots and connected devices with no cloud dependency.
Kiosks
Real-time user interaction, product recognition and local language processing embedded directly into kiosk hardware.
80 TOPS in a standard MXM slot.
MLA100 MXM drops into any MXM-compatible carrier board and adds 80 TOPS of AI inference while drawing 25 W power with no external power connector.
With a compact footprint of 82 x 70 mm, MLA100 MXM is built to run complex models in parallel, inside on-device platforms and AI boxes where space and power is limited.
80 TOPS at 25 W
ARIES NPU with 8 cores at 1.25 GHz, 16 GB LPDDR4X onboard.
The same NPU architecture that runs complex parallel workload on MLA100 PCIe Card and MLA400 PCIe Card, in a compact 82 × 70 mm module.
Extended Temperature Range
−35°C to +65°C standard operation, −35°C to +75°C extended.
Designed for deployment in industrial, outdoor and vehicle-mounted environments.
490+ Validated Models, One SDK
490+ model architectures validated across vision, transformer and multimodal workloads.
Pre-compiled open-source models are available in Mobilint Model Zoo.