GOMA ELETTRONICA SpA
Scheda modulo AI-MXM-H84A con chip e connettore a pettine dorato sul bordo inferiore

Mezzanine Modules

AI-MXM-H84AMXM Type B mezzanine module with graphics function and four Hailo-8 processors

Manufacturer

AI-MXM-H84A is an inference accelerator in MXM 3.1 Type B mezzanine format, intended for embedded systems already fitted with an MXM slot. Aetina positions it on AGVs, virtual fencing systems, autonomous machines and computer vision applications, where neural networks and deep learning run on board the machine with low latency.

The distinguishing axis is SWaP reduction: four Hailo-8 processors of 26 TOPS each fit on a module of 82.0 x 105.0 mm and 0.05 kg, so inference capability goes inside the existing enclosure, not around an add-in card. Aetina also provides a software package with AI development tools and a fine-tuning service for integrating inference models.

The catalogue configuration is AI-MXM-H84A (MXM3.1 Type B, 4× Hailo-8, 0°C to +70°C); the PCIe carrier board 92-6MXM4H-1000 with 4× HDMI is available as an optional accessory.

At a glance

  • Typical power consumption 25 W
  • CE and FCC certifications
  • TensorFlow and ONNX supported
  • Windows 10/11 and Linux 64-bit
Request informationFull specifications ↓

Technical specifications

FPGA / processing4× Hailo-8™ AI inference processors
Graphics4× Hailo-8 AI processor (26 TOPS ciascuno, 104 TOPS totali)

MemoryN/A

Architecture standardMXM
Board functionGPU / Graphics
Bus interfacePCIe Gen3 x16

Power consumption25 W

Dimensions82.0 x 105.0 mm (3.22” x 4.13”)
Weight0.05 kg

Operating temperature0–70 °C
Storage temperature-40–85 °C

CertificationsCEFCC

Operating systemsLinuxWindows

Ordering variantsMXM3.1 Type B, 4× Hailo-8, 0°C to +70°C

Form factorMXM Type B

Data extracted from the manufacturers' technical sheets: the official product spec sheet remains the reference for final confirmation.

Documentation

DatasheetPDF · 447 KB

In the same category

See the full category →
M3A1000-PP

Aetina

M3A1000-PP

MXM Type A graphics module with NVIDIA RTX A1000 GPU, Ampere architecture

M3A2000-VY

Aetina

M3A2000-VY

MXM Type A module with NVIDIA RTX A2000 GPU on Ampere architecture

M3A4500-WP

Aetina

M3A4500-WP

MXM Type B graphics module with NVIDIA RTX A4500 Embedded, Ampere architecture

M3A500-PP

Aetina

M3A500-PP

MXM Type A mezzanine module with NVIDIA RTX A500 Embedded GPU, Ampere architecture

GOMA's Turin headquarters at dusk: the red canopy over the entrance, the glazed façade and the GOMA ELETTRONICA sign

Let's talk about your project

Phone and email go straight to our technical sales team in Turin: the first answer is already a technical one.

Request a quoteContact support

Subscribe to our newsletter

New references in the catalogue, qualifications obtained and trade show dates. One email a month, no spam.