Secure embedded controllers integrate AI/ML acceleration

Built to power the next generation, always-connected IoT products, Alif Semiconductor’s Ensemble and Crescendo families are power-efficient devices which integrate AI/ML acceleration, multi-layered security, LTE Cat-M1 and NB-IoT connectivity, GNSS positioning, and memory.

The Ensemble family scales from single Arm Cortex®-M55 microcontrollers to so-called fusion processors which blend up to two Cortex-M55 cores, up to two Cortex-A32  microprocessor cores capable of running high-level operating systems, and up to  two Arm Ethos-U55 microNPUs for artificial intelligence (AI) and machine learning (ML)  acceleration.

The Ensemble family devices contain an advanced secure enclave that provides device integrity protection, secure identity, strong  root-of-trust and secure lifecycle management. They also have large on-chip SRAM and non-volatile memory, accelerated graphics, imaging, making them suitable for smart home automation, appliances, point of sale (PoS) terminals and robotics applications, said the company.

The Crescendo family offers the same functionality as the Ensemble family, with the addition of LTE Cat-M1 and NB-IoT cellular connectivity, optional iSIM for simplified subscriber management, integrated RF, power amplifiers, and a concurrent GNSS receiver. These capabilities make them well-suited for smart city, connected infrastructure, asset tracking, healthcare devices and wearable devices said the company, using a single chip to minimise size and weight.

To address the reliance on battery power in IoT devices, which can be challenging when there is a high requirement for local processing, AI/ML, and wireless communication, Alif Semiconductor has also introduced Autonomous Intelligent Power Management (aiPM) technology that allows fine-grained control of when resources  in the chip are being powered. This results in low power operation, enabling intelligent devices to last longer on smaller batteries.

“The solution that Alif delivers fills a significant gap in the market,” said Jerome Schang, head of Microsoft Azure Edge Silicon devices strategy. “We are always on the lookout for the most efficient technology platforms for our Edge experiences, and the Ensemble and Crescendo families are very well aligned with our customers’ needs.”

Ensemble and Crescendo devices are sampling now to lead customers. They are supported by Alif Semiconductor’s software, development tools and kits. Production qualification will be complete in 1Q22.

https://www.alifsemi.com/

> Read More

SOM-5883 COM is engineered for graphics performance

Equipped with 11th Generation Intel Core processors, the SOM-5883 computer on module (COM) is designed by Advantech for high-performance edge computing applications.

It is a COM Express Basic Type6 Module and offers octa-core computing performance under thermal dynamic performance (TDP) 45W. It also has Intel Iris Xe graphics and an Edge AI Suite software toolkit, which make the COM suitable for medical imaging and AI applications.

To resolve the heavy workload of image processing applications, SOM-5883 has the capability to support up to 128Gbyte DDR4 3200 SODIMM, with high capacity and high frequency memory technology. The COM includes high speed I/O interfaces including PCIe Gen4 and USB3.2 Gen2 (10Gbits per second) and 2.5Gbase-T for enhanced system expansion. There is also optional NVMe SSD and TPM2.0 onboard design. Other features are -40 to +85 degrees C operating temperature and 8.5 to 20V power input. As a result, says Advantech, the SOM-5883 is reliable and secure for edge computing requirements.

The SOM-5883 offers a 1.7X computing performance growth and 1.5X 3D graphics upgrade. Advantech provides Edge AI Suite containing Intel Open VINO and more than a hundred AI models.

There are four independent displays up to 4K over three DisplayPort 1.4 / HDMI 2.1, optional eDP or LVDS, and VGAl, and the COM is configurable up to two ports 8K HDR outputs. Graphics-driven applications such as ultrasound can achieve much more accurate diagnosis with the combination of native graphics engine and external expansion for system configuration, said Advantech. At the same time, SOM-5883 supports multiple cutting-edge I/O technologies. It enables high speed data transmission with improved bandwidth of 2.5G LAN and an NVMe SSD onboard, with the capability to transfer huge data between AIoT applications. The 2.5G LAN equipped with time sensitive networking (TSN) is suitable for automation or edge applications requiring low-latency. This technology fulfils the demand of real-time traffic, prioritising critical data in the communication process, such as transferring command, actions or safety information on time, said the company. It also supplies extraordinary USB4, which can be configured as DisplayPort, Thunderbolt, USB 3.2, and USB 2.0 in a slim USB-Type-C connector. Advantech also offers USB4 reference design documents and SOM-MZ10 development board, to help partners accelerate the implementation of new I/O technology.

Advantech’s QFCS (Quadra flow cooling system) thermal technology enables the SOM-5883 to release 100 per cent CPU power. The technology is not only efficient in heat dissipation but also thin, quiet and light, sufficiently so to be integrated into a 1U height slim chassis, added Advantech.

The SOM-5883 features error code correction (ECC) memory, TPM2.0 chip onboard to prevent cyber-threats, and also supports BIOS storage protection, security boot, BIOS power-management, plus WISE-DeviceOn for remote hardware monitoring and over-the-air (OTA) software update to prevent system malfunction.

http://www.advantech.eu

> Read More

Advantech bases edge AI PCs on NXP’s i.MX 8M Plus

Four edge artificial intelligence (AI) products based on NXP’s i.MX 8M Plus have been launched by Advantech.

The early access partner of NXP has released the RSB-3720 2.5-inch PICO-ITX single board computer (SBC), the EPC-R3720 compact edge AI box PC system, and the ROM-5722 SMARC 2.1 computer on module (COM).

The RSB-3720 and EPC-R3720 have UIO40-Express, modular application-oriented I/O expansion boards, and AIM-Linux to Vision AI applications.

All three are powered with up to four core 1.8GHz Cortex-A53 processors supported by a Cortex-M7 co-processor for handling real-time tasks. There are two image signal processors (ISPs) and a dedicated neural network processor operating at up to 2.3T operations per second. This is sufficient for image and AI inferencing capabilities and all three consume less power than when doing the calculations by the core CPUs alone. This allows for AI applications like object detection, recognition, classification and pose estimation.

Advantech AIM-Linux modular Linux software services integrate many types of application add-ons and Edge AI inference engines/libraries mainly based on the NXP eIQ toolkit. Python-based demo programs empower AI inference using mainstream inference engines, including Arm NN, TensorFlow Lite and ONNX.

To meet the demands of edge AI applications such as facial recognition and product assembly optimisation, all of which require premium camera inputs, the ROM-5722, RSB-3720, and EPC-R3720 provide 2 x MIPI-CSI camera inputs for connection with 2 x built-in ISPs — enabling intelligent vision-based systems. They accommodate multiple displays via 4Kp30 HDMI, one dual channel LVDS (configurable as two single channel LVDS) and one four-lane MIPI-DSI. They also provide I/O for advanced network and peripheral connection connectivity like two GbE LANs. Expansion for WLAN and WWAN are available.

The RSB-3720 features an UIO40-Express mechanical design for connection with application-oriented I/O expansion boards. Developers can select I/O extension boards for more serial ports (RS-232, RS-485), isolated DI/DOs, USB 2.0, and four port GbE Ethernet hub. AIM-Linux includes embedded Linux board support package (BSP) based on the long-term-support version and peripheral drivers for diverse software requirements.

All three modules and the EPC-R3720 support a wide operating temperature range of -40 to +85 degrees C and have high vibration resistance (3.5Grms for board and 3.5Grms for box).

Advantech’s WISE-DeviceOn provides fast and simple set up for projects using multiple IoT devices. It provides real-time device management, firmware over the air (OTA) maintenance, failure analysis, and operation efficiency optimisation on the same dashboard.

http://www.advantech.eu  

> Read More

COMs set bandwidth benchmark for IoT and edge computing

Claiming to set a new benchmark of up to 20 PCIe Gen 4.0 lanes, the conga-HPC/cTLH COM-HPC Client Size B modules (120 x 120mm) and conga-TS570 COM Express Basic Type 6 modules (125 x 95mm) will be available from congatec. They will be based on scalable 11th Gen Intel Core vPro, Intel Xeon W-11000E, and Intel Celeron processors, with selected variants for extreme temperatures ranging from -40 to +85 degrees C.

The modules target the most demanding IoT gateway and edge computing applications, says congatec. They are built on Intel’s 10nm SuperFin technology in a two-package design with dedicated CPU and platform controller hub (PCH) to set the  new bandwidth benchmark of up to 20 PCIe Gen 4.0 lanes for massive connected real-time IIoT gateway and intelligent edge computing workloads.

To process enormous workloads, the modules have up to 128Gbyte DDR4 SO‑DIMM RAM, integrated AI accelerators and up to eight high performance CPU cores that achieve up to 65 per cent gain in multi-thread performance and up to 32 per cent gain in single-thread performance. Visualisation, auditory and graphics intensive workloads are boosted up to 70 per cent compared to predecessors, says congatec, to enhance immersive experiences. Coupled with support for 8K HDR videos for diagnostic, the modules can be used in surgery, medical imaging and e-health edge applications. Combined with the platform’s AI capabilities and the comprehensive Intel OpenVINO toolkit, doctors can gain easy access and insights into deep learning based diagnostic data, advised congatec.

The integrated Intel UHD graphics supports up to four 4K displays in parallel. It can process and analyse up to 40 HD 1080p/30 frames per second video streams in parallel for 360 degree views in all directions. AI infused massive vision capabilities will find applications in factory automation, machine vision for quality inspection in manufacturing, safe spaces and cities, as well as collaborative robotics and autonomous vehicles in logistics, agriculture, construction and public transport.

Both form factors support up to 128Gbyte DDR4 SO-DIMM memory with 3200 Mtransfers per second and optional error code correction (ECC). To connect peripherals with massive bandwidth the COM-HPC modules support 20 PCIe Gen 4 lanes (x16 and x4), and the COM Express versions support 16 PCIe lanes. Designers can leverage 20 PCIe Gen 3 lanes with COM-HPC, and eight PCIe Gen 3 lanes on COM Express.

To support ultra-fast NVMe SSD, the COM-HPC module provides 1x PCIe x4 interface to the carrier board. The COM Express board has NVMe SSD onboard to use all native Gen 4 lanes supported by the new processor. Storage media can be connected via 2x SATA Gen 3 on COM-HPC, and 4x SATA on COM Express.

The COM-HPC module offers latest 2x USB 4.0, 2x USB 3.2 Gen 2, and 8x USB 2.0 and the COM Express module offers 4x USB 3.2 Gen 2 and 8x USB 2.0 in compliance to the PICMG specification. For networking, the COM-HPC module offers two 2.5 GbE and the COM Express module executes one GbE.

Sound is provided via I2S and SoundWire in the COM-HPC version, and HDA on the COM Express modules.AI and deep learning inference algorithms can seamlessly run either massively parallel on the integrated GPU, or on the CPU with built-in Intel Deep Learning Boost that combines three instructions into one, accelerating inference processing and situational awareness.

The COM-HPC Client and COM Express Type 6 platforms have integrated safety functions for the fail-safe operation of many mobile vehicles and robots and stationary machinery. Real-time support is provided by RTOSes such as Real Time Linux and Wind River VxWorks with native support from Real-Time Systems’ hypervisor technology, which is also officially supported by Intel.

Further real-time capabilities include Intel Time Coordinated Computing (Intel TCC) and Time Sensitive Networking (TSN) for real-time connected IIoT/industry 4.0 gateways and edge computing devices.

https://www.congatec.com

> Read More

About Smart Cities

This news story is brought to you by smartcitieselectronics.com, the specialist site dedicated to delivering information about what’s new in the Smart City Electronics industry, with daily news updates, new products and industry news. To stay up-to-date, register to receive our weekly newsletters and keep yourself informed on the latest technology news and new products from around the globe. Simply click this link to register here: Smart Cities Registration