ADLINK launches first embedded MXM graphics modules based on NVIDIA Ampere architecture for edge computing and AI

Joyce Siow, DIGITIMES Asia, Taipei 0

Credit: ADLINK

ADLINK EGX-MXM-A1000, EGX-MXM-A2000 and EGX-MXM-A4500 are the first modules to use NVIDIA's embedded GPUs based on NVIDIA Ampere architecture. ADLINK embedded MXM graphics modules offers high performance GPU acceleration in the compact, power-efficient MXM form factor, bringing edge computing and embedded AI to numerous vertical markets in healthcare, manufacturing, transportation, and more. Rugged design built for severe temperature extremes, shock and vibration, and corrosion in harsh conditions.

ADLINK Technology Inc., a global leader in edge computing, today introduced the industry's first embedded MXM graphics modules based on the NVIDIA Ampere architecture, built for accelerated computing and AI workloads at the edge. The new embedded graphics modules deliver real-time ray tracing, AI-accelerated graphics and energy-efficient AI inference acceleration in the compact mobile PCI express (MXM) form factor. These modules enhance responsiveness, precision and reliability for mission-critical, time-sensitive applications in healthcare, manufacturing, transportation, gaming and other sectors.

"Computing and AI workloads are shifting from the cloud to the edge to shorten response time, enhance security and lower communication costs. With more and more data being processed at the edge, the performance requirements are getting higher while the power budget remains nearly unchanged. The NVIDIA Ampere architecture delivers a major boost in performance and power efficiency, which enables general computation, image processing and reconstruction, and AI inference at the edge to advance to the next level," said Zane Tsai, director of platform product center, ADLINK.

"The NVIDIA Ampere architecture delivers breakthrough performance and features, combining the latest generation RT Cores, Tensor Cores, CUDA Cores, PCIe Gen 4, and NVIDIA video codecs," said Scott Fitzpatrick, vice president of product marketing at NVIDIA. "As rendering and simulation become ubiquitous across industries, NVIDIA's latest embedded solutions offer up to 2x rendering performance, 2x FP32 throughput, as well as hardware-accelerated video encoding and decoding for significant increases in both graphics and compute workloads."

Based on the NVIDIA Ampere architecture, ADLINK's embedded MXM GPU modules offer up to 5,120 CUDA Cores, 160 Tensor Cores, and 40 RT Cores with support for PCIe Gen 4 and up to 16GB GDDR6 memory at up to 115 watts of TGP. These modules can satisfy compute-intensive, graphics-demanding, memory-hungry applications. They are one-fifth the size of full-height, full-length PCI Express graphics cards, and are hardened to operate under severe temperature extremes, shock and vibration, and corrosion resistance for use in size, weight and power-constrained edge environments. These embedded graphics modules offer longevity support for five years. Developers, solution architects and system integrators can innovate new solutions with confidence that the supply is steady through a product life cycle.

Applications of the embedded MXM graphics modules include:

Healthcare: Accelerated image reconstruction for mobile X-ray, ultrasound, and endoscopic systems
Transportation: Real-time object detection on railways or airport runways to enhance transport safety
Retail and Logistics: Navigation and route planning for autonomous drones and mobile robots (AMRs) to assist with last-mile delivery
Aerospace and Defense: Time-sensitive and mission-critical command, control, communications, computers (C4) applications; intelligence, surveillance and reconnaissance (ISR) applications
Gaming: Immersive and stunning visuals experience for multi-display gaming machines

For more information, visit the product page here.

ADLINK launches first embedded MXM graphics modules

ADLINK launches first embedded MXM graphics modules based on NVIDIA Ampere architecture for edge computing and AI
Photo: Company