Are You Ready For AI?
IEI offers you choices for your AIoT Ready Systems
Introducing Google TPU-Based Mustang-T100-T5 Accelerator
Google TPU-based accelerator card featuring:
  • 5x Google Coral Edge TPU’s
  • 4 Tera-operations per second (TOPS) each, 20 TOPS total
  • 0.5 Watts each
  • Lower cost AI computing with enhanced efficiency (2 TOPS per watt)
  • Can process in parallel or have individual video streams attached to each
  • Supported by well-developed Tensor Lite Community
  • PCIe card (Gen 2x4), low profile, total power about 15 Watts

Intel® Movidius™ Myriad™ X-based VPU (Video Processing Unit) Accelerators
Featuring:
  • Variety of scale: 1, 2, 4 or 8 Myriad X VPU’s (MA2485)
  • Small form factor options: M.2, Mini-PCIe, low profile PCIe card
  • 1 TOPS processing per VPU, low-cost accelerators
  • Can process in parallel or have individual video streams attached to each
  • Supported by Intel® OpenVINO Toolkit for rapid development
  • Rugged fanless enclosed platform available (ITG-100AI)

Applications include:
Transportation, Factory, Medical, Retail, Networking
 
--A variety of AI-Ready Development, Training and Inference platforms available 
Intel® Stratix® 10 NX FPGA Optimized for AI Applications
In recent years, compute-intensive artificial intelligence (AI) applications have used graphics processing units (GPUs) to increase the AI compute throughput. Fully utilizing the GPU’s architectural parallelism can be challenging as the AI workloads evolve. FPGAs offer a different approach by configuring the FPGA’s resources to match the desired AI algorithms achieving substantially higher application performance.

The Intel® Stratix® 10 NX FPGA takes the AI compute performance to the next level by integrating AI-optimized Tensor blocks which enables up to 15X the INT8 throughput when compared to the Intel® Stratix® 10 MX FPGA for AI workloads. High-level Neural Processing Unit (NPU) software toolchains abstracts the FPGA-specific logic complexity to enable the rapid development of sophisticated AI models and applications. High-speed transceiver interfaces connected to over two million FPGA Logic Elements (LE) can be customized to efficiently transport data in and out of the FPGA’s low latency HBM2 memories eliminating connectivity bottlenecks thereby maximizing AI application performance.
Key Features
  • 3960 high-performance AI Tensor blocks
  • Near-compute memory with integrated 3D stacked HBM2 high-bandwidth DRAM
  • High-speed networking interfaces with 57.8G PAM4 transceivers
  • 2.1M LE Hyperflex connected FPGA fabric
  • NPU development toolchains
Target Applications
  • Security
  • Real-time Video Analytics
  • Native Language Processing
With Macnica's SENSPIDER edge computing solution with analog sensor interfaces, customers can collect multiple types of sensor data simultaneously and build AI-powered Condition-based Maintenance (CBM), preventing sudden machine outages and downtime in Industrial IOT equipment.

Advanced sensor data collection
SENSPIDER connects to various types for high-precision analog sensors with no amplifier required, collecting data at a high sampling rate (max 48KHz) from up to eight channels, all in the same time series.

Flexible support for sensing technology
By switching interface cards, you can collect electric current, temperature, and other types of sensor data in addition to vibration data. This helps you find correlations between different types of sensing data which is useful for improving AI model performance.

All-in-one embeddable form factor
Sensor power supply, sensor amplifier, analog-digital converter, computing, and data logger are all provided in one box. It is so compact it fits neatly into a control panel.
Empowering innovation of next-gen intelligent solutions, including advanced robotics, AI-enabled drones, AI processing gateways and more
The Open-Q™ 5165RB SOM is an ultra-compact (50mm x 29mm) production-ready computing module based on the powerful Qualcomm® QRB5165 System-on-Chip. The QRB5165 utilizes Qualcomm Technologies’ heterogenous compute expertise to provide an SoC with multiple specialized processing cores such as the 5th generation AI Engine, hardware video analytics engine, Qualcomm Spectra™ ISP, Qualcomm Adreno™ GPU, and Qualcomm Hexagon™ DSP. Coupled with the latest Wi-Fi 6 connectivity, advanced camera features, and high-speed interfaces, the 5165RB SOM creates the perfect computing core for a variety of leading-edge robotics applications including advanced robotics for consumer, industrial, and defense use. The 5165RB SOM is also useful for the development of AI-enabled drones, UAVs, machine vision platforms, AI processing gateways, face detection and recognition systems, deep learning and neural networks processing and autonomous systems.
Key Features
  • SOM with powerful specialized processing cores
  • On-device Qualcomm® AI Engine™ for machine vision, neural networks, deep learning
  • Power-efficient edge AI computing solution pushing 15 TOPS
  • Ubuntu 18.04 Linux with support for Robot Operating System 2.0
  • Container and accelerator APIs - Docker, TensorFlow Lite, NNAPI and others
Ambarella Acquires Oculii
Deployable on Ambarella’s existing CVflow® SoCs, the Oculii adaptive AI software algorithms enable radar perception, provide enhanced resolution and accuracy with a range of up to 400m with a wide FoV – increasing the safety and reliability of everything from advanced driver-assistance systems (ADAS) and autonomous vehicles to robotics and security. The fusion of Ambarella’s camera technology with the Oculii radar software stack provides an all-weather, low-cost and scalable perception solution, enabling higher levels of autonomy for manufacturers.
Macnica - Empowering Efficient Innovation
About Macnica Americas
Macnica Americas, Inc. is the North America division of Macnica Fuji Electronics Holdings, Inc., a $5.2B global leader in technology product distribution, technical support, and design services.
To sign up for our monthly newsletter, please click here . 
For other inquiries, contact us here .