ADLINK’s MXM GPU Module Based on NVIDIA Turing Architecture


NVIDIA Quadro GPUs are an ideal fit for mission-critical edge computing and AI applications

Edge systems integrated with GPUs can perform much better with respect to applications, such as medical imaging, manufacturing defect inspection, traffic-flow analysis in smart cities and many other embedded segments. By using Tensor Cores, an integer data path and unified architecture for shared memory, that enables 50 per cent improvement in performance per CUDA core, as compared with the NVIDIA Pascal architecture, can be achieved. Thus, GPU-accelerated computing is quite popular nowadays.

ADLINK Technology’s MXM GPU modules can deliver high-performance computing, computer graphics and artificial intelligence (AI) by tapping the power of the Turing architecture and addressing the size, weight, and power (SWaP) challenges facing embedded applications, making them an ideal fit in mission-critical edge computing and AI applications. The NVIDIA CUDA platform enables the GPU to be continually upgraded for additional performance gains with minimal software updates.


“In the example of medical imaging, image construction involves massive parallel data processing. Integrating required computing capabilities and meeting stringent SWaP requirements of mobile medical imaging instruments can be challenging, which is why NVIDIA Turing architecture-based MXM GPU modules are needed. ADLINK has seen the demand for GPU-accelerated computing springing up in the healthcare industry in China and Europe,” said Zane Tsai, Director of ADLINK’s Platform Product Center at the Embedded Platform & Module Business Unit. “Big data and AI are being applied to screening and detection and lesion diagnosis in preventive and precision medicine. Turing architecture-based MXM GPU modules with Tensor Cores can provide high-performance efficiency for these mission-critical tasks.”

World-first Neuromorphic Chip For AI

High performing GPUs with respect to autonomous machines

Another growing application where GPU performance stands out is autonomous machines, such as self-driving vehicles, robots, and drones, which involve the processing of large volumes of high-dimensional data. “With the ability to process high volumes of data in parallel to produce high-precision results in the SWaP-constrained mobile environment, the Turing architecture-based MXM GPU modules deliver performance enhancements for autonomous machines to perform complicated manoeuvres,” said Tsai.

“Powerful performance and enterprise reliability are essential to a rapidly expanding ecosystem of graphics-intensive applications that need the capabilities of NVIDIA Quadro GPUs, but operate in systems that do not support a standard PCI Express graphics card,” said Scott Fitzpatrick, vice president of Quadro Product Marketing, NVIDIA. “The ADLINK MXM GPU embedded modules deliver NVIDIA Quadro RTX performance and features in a custom, upgradeable form factor with long-term availability to bring the power of AI and big data to an even wider range of systems across healthcare, robotics, manufacturing and smart cities.”

Issues and ADLINK’s solution for edge computing

Edge computing and AI applications tend to have long development cycles and require a large investment associated with development, function validation and acquiring certifications.

As one of the providers of edge computing solutions, ADLINK’s GPU-enabled products come in a variety of form factors with longevity support, custom firmware and optimised computing platforms to enable developers to meet the requirements of edge applications.

ADLINK will showcase the MXM GPU modules powered by NVIDIA Quadro RTX embedded GPUs featuring the NVIDIA Turing architecture at Embedded World 2020, Germany from 25th to 27th February 2020.



Please enter your comment!
Please enter your name here

Are you human? *