Aetina has unveiled a new line of industrial MXM accelerator modules built on NVIDIA’s Blackwell architecture, aiming to bring data centre-level AI performance to autonomous robotics, manufacturing systems, and medical devices deployed at the Edge.
The Taiwan-based Edge AI specialist said the modules, part of its CoreEdge MXM family, deliver up to three times the AI performance of the previous generation thanks to NVIDIA RTX PRO Blackwell Embedded GPUs. The series adheres to the MXM 3.1 industrial standard, targeting space- and power-constrained environments such as autonomous mobile robots, unmanned vehicles, smart-factory equipment, and surgical assistance systems.
The flagship MX5000B-XA, powered by the NVIDIA RTX PRO 5000 Blackwell GPU, features 10,496 CUDA cores, 80 RT cores, and 320 Tensor cores, with a stated 40.62 TFLOPS FP32 performance. Aetina said validation tests showed twice-as-fast large-language-model inference on Google’s Gemma-2-9b model and an 80% uplift in Geekbench AI benchmarks. Support for FP4 precision, DLSS 4 multi-frame generation, and GDDR7 memory—offering bandwidth up to 896GB/s—is designed to enable more complex models, including LLMs and vision transformers, to run efficiently at the edge.
Aetina emphasised industrial durability as a core differentiator. The modules occupy roughly one-fifth of the space of a conventional graphics card yet are rated for operation between –40°C and +85°C. Conformal coating protects against dust and moisture, while adjustable TGP settings and reinforced mounting points are intended to support deployments in high-vibration environments. The company offers a five-year product lifecycle guarantee.
The modules undergo extensive thermal cycling and cold-boot testing, with every unit chamber-tested before shipment, according to Aetina. Evaluation kits with PCIe Gen4 x16 interfaces and HDMI outputs are also available to accelerate prototyping and system integration.
Jackal Chen, Director of AI Accelerator Card & GPU Products at Aetina, said the Blackwell-powered MXM range is designed to prepare customers for “the age of AI agents, multimodal applications, and real-time AI-accelerated computing”, adding that the company provides engineering services to reduce time-to-market for Edge deployments.
The MX5000B-XA and MX500B-QA are available immediately. The MX4000B and MX2000B models are expected to enter mass production in the first quarter of 2026.
There’s plenty of other editorial on our sister site, Electronic Specifier! Or you can always join in the conversation by commenting below or visiting our LinkedIn page.
