AISiMa.ai Modalix

Modalix SoM

System-on-Module

[SiMa.ai]

The SiMa.ai Modalix System-on-Module puts the 50 TOPS Modalix MLSoC, LPDDR5 memory and power on one compact module. It is pin-compatible with NVIDIA Jetson Orin NX and Nano modules, so it can take their place on existing carrier boards.

System Overview
  • Form Factor

    System-on-Module

  • AI Performance

    50 TOPS

  • Memory

    8 / 16 / 32 GB LPDDR5

  • Processor

    8x Arm Cortex-A65

Product Information
  • Model

    Modalix SoM

  • Vendor

    SiMa.ai

  • Category

    On-Premise AI

  • Product Group

    SiMa.ai Modalix

  • Type

    System-on-Module

AceTec Support
  • Consulting

    Requirements analysis and custom system design

  • Implementation

    Integration guidance and lifecycle technical support

  • Verification

    Testing against specifications and industry standards

Features

  • Modalix MLSoC: 50 TOPS (BF16, INT8, INT16)
  • Multimodal transformers, LLMs, LMMs and GenAI, as well as CNNs and traditional computer vision
  • 8x Arm Cortex-A65 @ 1.4 GHz, 32k Dhrystone MIPS
  • 4-core Synopsys EV74 computer vision DSP @ 1 GHz, 720 16-bit GOPS
  • Arm Mali-C71AE low-latency image signal processor @ 1.2 GHz
  • 4K60 video decode (H.264, H.265, AV1) and encode (H.264, H.265)
  • 8, 16 or 32 GB of on-board 128-bit LPDDR5 at 6400 Mbps
  • Pin-compatible with NVIDIA Jetson Orin NX and Nano modules
  • Secure boot with boot code authentication and encryption; secure key storage
  • 69.6 x 45 mm, 260-pin SO-DIMM connector, 5–20 V power input
  • Commercial and industrial temperature grades; -40 to +85 °C

Overview

Modalix on a module

The Modalix SoM carries SiMa.ai's Modalix MLSoC with its LPDDR5 memory and power, and plugs into a carrier board through a 260-pin SO-DIMM connector. It is pin-compatible with NVIDIA Jetson Orin NX and Nano modules, so it can replace them on existing carrier boards.

Built for Physical AI

The MLSoC combines a 50 TOPS machine-learning accelerator with Arm application cores, a computer vision DSP, an image signal processor and 4K video codecs. It runs multimodal transformers, large language and multimodal models and generative AI, as well as convolutional networks and traditional computer vision. Example applications named by SiMa.ai: smart vision, conversational interfaces, drones, robotics, Industry 4.0, automotive, smart retail and healthcare, aerospace and defense.

Software

Applications are built with SiMa.ai's Palette software suite and its Palette Neat agentic development environment. ONNX models and popular frameworks such as OpenCV are supported.

Specifications

Processing

Item

Content

ML Accelerator

50 TOPS (BF16, INT8, INT16); LLMs, neural networks and GenAI

Application Processor

8x Arm Cortex-A65 @ 1.4 GHz, 32k Dhrystone MIPS

Computer Vision Unit

4-core Synopsys EV74 DSP @ 1 GHz, 720 16-bit GOPS

Image Signal Processor

Arm Mali-C71AE @ 1.2 GHz, low latency

Video Decode

H.264 / H.265 / AV1, 4K60

Video Encode

H.264 / H.265, 4K60

Memory and Storage

Item

Content

Memory

8 / 16 / 32 GB on-board 128-bit LPDDR5, 6400 Mbps

Storage

16 GB eMMC flash, 64 MB QSPI flash, 128 Kb EEPROM

External NVMe through PCIe x4; external SSD through USB 3.0

I/O

Item

Content

Display

HDMI 1.4, 4K resolution

Camera

4x2 MIPI CSI-2

PCIe

4x PCIe Gen5, RC & EP

Ethernet

1x 1 Gb Ethernet PHY

USB

3x USB 3.0

Other

4x I2C, 3x UART, 14x GPIO, 2x SPI

Security

Item

Content

Boot and Security

Secure boot with boot code authentication and encryption

Secure key storage and management; security engine for user code

Module

Item

Content

Size

69.6 x 45 mm

Connector

260-pin SO-DIMM

Power Input

5–20 V

Temperature

-40 to +85 °C; commercial and industrial grades

Compatibility

Pin-compatible with NVIDIA Jetson Orin NX and Nano modules

Versions

Item

Content

8 GB

Standard product

16 GB, with or without HDMI

Standard product

32 GB, with or without HDMI

Special order

Resources