Aetina Corporation has unveiled its latest MXM AI Accelerator Module series, part of the CoreEdge MXM AI Module family, featuring the NVIDIA RTX PRO Blackwell Embedded GPUs. This new series delivers up to three times the AI performance of earlier models, setting a fresh standard for edge AI computing.
Designed to be compact, efficient, and robust, the modules adhere to the MXM 3.1 industrial standard. They cater to high-precision and high-density applications, including autonomous mobile robots, unmanned vehicles, smart manufacturing, and medical systems.
NVIDIA Blackwell architecture: Enhancing AI at the edge
With the transition of AI applications from data centers to edge environments, Aetina's CoreEdge embedded MXM AI modules, featuring NVIDIA RTX PRO Blackwell GPUs, offer compact and reliable solutions. These modules incorporate groundbreaking AI, ray tracing, and neural graphics technologies.
The flagship MX5000B-XA model, powered by the NVIDIA RTX PRO 5000 Blackwell GPU, features 10,496 CUDA cores, 80 RT cores, and 320 Tensor cores, achieving 40.62 TFLOPS FP32 performance. Tests indicate it provides twice the LLM inference speed on the Gemma-2-9b model and 80% faster AI inference in Geekbench AI benchmarks.
The series supports FP4 precision, DLSS 4 multi-frame generation technology, and GDDR7 memory with up to 896 GB/s bandwidth, enhancing AI computing throughput significantly. This enables the deployment of advanced AI models like large language models and vision transformers in constrained environments, providing low-latency and high-throughput computing for demanding applications such as physical AI.
Reliability for critical applications
The modules adapt elastic TGP power settings and offer robust mounting for high-vibration domains
Aetina’s CoreEdge MXM AI modules are designed to use only one-fifth of the space required by usual graphics cards, offering workstation-class performance. Engineered surpassing industrial standards, the design supports temperatures from -40°C to +85°C and features protective coatings against dust and moisture.
The modules accommodate adjustable TGP power settings and offer robust mounting for high-vibration environments. Aetina ensures a five-year product lifecycle, ensuring stability of supply.
This reliability is guaranteed through rigorous engineering practices, including wide-temperature components, extensive thermal cycling, and 100% chamber testing before delivery. The modules aim to provide confidence for deployment in critical settings like autonomous robots and medical equipment. Additionally, Aetina offers evaluation kits with PCIe Gen4 x16 interfaces and HDMI outputs to facilitate faster development and simpler integration.
Advancing generative and physical AI
Aetina's Director of AI Accelerator Card & GPU Products Division, Jackal Chen, stated, "Aetina's CoreEdge embedded MXM AI accelerator module series demonstrates our commitment to delivering exceptional performance and reliability. With NVIDIA Blackwell technology and our engineering expertise, we achieve high-performance AI computing in demanding environments. We are prepared for the age of AI agents, multimodal applications, and real-time AI-accelerated computing, providing value-added engineering services to reduce time-to-market and harness edge generative and physical AI's full potential."
The MX5000B-XA and MX500B-QA modules are currently available, while the MX4000B and MX2000B modules are expected to enter mass production in the first quarter of 2026.
Aetina Corporation, a pioneering provider of Edge AI solutions, has announced the launch of its next-generation MXM AI Accelerator Module series accelerated by NVIDIA RTX PRO Blackwell Embedded GPUs, which is part of its CoreEdge MXM AI Module family. This new series is accelerated by the NVIDIA Blackwell GPU platform to deliver up to 3X the AI performance uplift of its predecessor, establishing a new benchmark for advanced AI computing at the edge
These modules are designed to be compact, efficient, and rugged, adhering to the MXM 3.1 industrial standard. They are purpose-designed for high-precision, high-density workloads in critical applications such as autonomous mobile robots (AMRs), unmanned vehicles (UAVs/UGVs), smart manufacturing equipment, and medical and surgical assistance systems. This series brings the computing power of generative AI and physical AI closer to the edge, facilitating a comprehensive upgrade of smart applications.
NVIDIA Blackwell architecture: Proven exceptional performance
As AI applications shift from data centres to edge environments, Aetina's CoreEdge embedded MXM AI modules offer compact and industrially reliable designs featuring NVIDIA RTX PRO Blackwell GPUs, integrating breakthrough AI, ray tracing, and neural graphics technologies.
Aetina’s flagship MX5000B-XA model, accelerated by NVIDIA RTX PRO 5000 Blackwell GPU, boasts 10,496 CUDA cores, 80 RT cores, and 320 Tensor cores, delivers an impressive 40.62 TFLOPS FP32 performance. Validation testing demonstrates 2x faster LLM inference on the Gemma-2-9b model and 80% faster AI inference performance in Geekbench AI benchmarks.
By supporting FP4 precision, DLSS 4 multi-frame generation technology, and GDDR7 memory with bandwidth up to 896 GB/s, the NVIDIA Blackwell MXM series boosts AI computing throughput by up to 3X. These performance breakthroughs enable deployment of advanced AI models such as large language models (LLMs) and vision transformers in space- and power-constrained edge environments, providing lower latency and higher throughput computing for demanding applications such as physical AI.
Industrial-grade reliability: Engineered for critical environments
Aetina's CoreEdge embedded MXM AI modules occupy only one-fifth of the board space required by traditional graphics cards while delivering performance comparable to workstation-class computing. Unlike commercial-grade products, Aetina's series design powered by the NVIDIA Blackwell platform surpasses industrial standards. It supports a wide temperature range of -40°C to +85°C and features dust- and moisture-resistant conformal coating protection.
Additionally, it offers adjustable TGP power settings and high-vibration resistant mounting holes on MXM Type B modules, ensuring secure installation in high-vibration environments. Aetina also provides a 5-year product lifecycle support, guaranteeing long-term supply stability.
This comprehensive reliability stems from Aetina's rigorous engineering methodology, including the use of wide-temperature components with high-reliability packaging, extensive thermal cycling testing, cold boot testing, and 100% chamber testing before shipment. This ensures system integrators and developers can deploy with absolute confidence in critical applications from autonomous robots, unmanned vehicles, smart manufacturing and medical equipment. Additionally, Aetina provides evaluation kits equipped with PCIe Gen4 x16 interfaces and HDMI outputs to accelerate development processes and simplify integration.
Unleashing the generative and physical AI at the edge
"Aetina's CoreEdge embedded MXM AI accelerator module series demonstrates our commitment to delivering exceptional performance and reliability," said Jackal Chen, Director of AI Accelerator Card & GPU Products Division at Aetina.
"By incorporating NVIDIA Blackwell technology alongside Aetina's extensive engineering expertise, we achieve high-performance AI computing for even the most demanding environments, Additionally, we are prepared for the age of AI agents, multimodal applications, and real-time AI-accelerated computing. We offer comprehensive value-added engineering services to streamline system integration processes, helping customers significantly reduce their time-to-market and fully harness the potential of edge generative AI and physical AI."
The MX5000B-XA and MX500B-QA are available now. The MX4000B and MX2000B modules are expected to enter mass production in Q1 2026.
