EdgeCortix SAKURA-II Edge AI accelerator deliver up to 60 TOPS in an 8W power envelope

SAKURA-II M.2 and PCIe Edge AI accelerators

EdgeCortix has just announced its SAKURA-II Edge AI accelerator with its second-generation Dynamic Neural Accelerator (DNA) architecture delivering up to 60 TOPS (INT8) in an 8Watts power envelope and suitable to run complex generative AI tasks such as Large Language Models (LLMs), Large Vision Models (LVMs), and multi-modal transformer-based applications at the edge. Besides the AI accelerator itself, the company designed a range of M.2 modules and PCIe cards with one or two SAKURA-II chips delivering up to 120 TOPS with INT8, 60 TFLOPS with BF16 to enable generative AI in legacy hardware with a spare M.2 2280 socket or PCIe x8/x16 slot. SAKURA-II Edge AI accelerator SAKURA-II key specifications: Neural Processing Engine – DNA-II second-generation Dynamic Neural Accelerator (DNA) architecture Performance 60 TOPS (INT8) 30 TFLOPS (BF16) DRAM – Dual 64-bit LPDDR4x (8GB,16GB, or 32GB on board) DRAM Bandwidth – 68 GB/sec On-chip SRAM – 20MB Compute Efficiency – […]

MemryX MX3 edge AI accelerator delivers up to 5 TOPS, is offered in die, package, and M.2 and mPCIe modules

MemryX MX3 EVB

Jean-Luc noted the MemryX MX3 edge AI accelerator module while covering the DeGirum ORCA M.2 and USB Edge AI accelerators last month, so today, we’ll have a look at this AI chip and corresponding modules that run computer vision neural networks using common frameworks such as TensorFlow, TensorFlow Lite, ONNX, PyTorch, and Keras. MemryX MX3 Specifications MemryX hasn’t disclosed much performance stats about this chip. All we know is it offers more than 5 TFLOPs. The listed specifications include: Bfloat16 activations Batch = 1 Weights: 4, 8, and 16-bit ~10M parameters stored on-die Host interfaces – PCIe Gen 3 I/O and/or USB 2.0/3.x Power consumption – ~1.0W 1-click compilation for the MX-SDK when mapping neural networks that have multiple layers Under the hood, the MX3 features MemryX Compute Engines (MCE) which are tightly coupled with at-memory computing. This design creates a native, proprietary dataflow architecture that utilizes up to 70% […]

NanoCell V2.1 battery-powered ESP32-C3 IoT board runs ESPHome for Home Assistant integration

nanocell v2 stack

The NanoCell V2.1 is a development board built around the Espressif ESP32-C3 SoC (system-on-a-chip) preloaded with ESPHome firmware for low-power applications and improved Lithium battery management. The development board is a white printed circuit board with gold-plated contacts and a battery fuel-gauge IC, designed by Frapais’ lab in Greece. As the name suggests, the NanoCell V2.1 is the latest in a series of iterations of ESP32-C3-based devices targeted at low-power applications. Compared to earlier versions, it offers a better user experience and improved power efficiency. It features a buck-boost converter that reduces standby current consumption to 66uA (excluding the current consumed by the ESP32 module). The battery management system (BMS) integrated circuit supports accurate capacity measurement and protects connected Lithium batteries from overcharging and other harmful scenarios. Also, two LEDs on the board serve as power and charging indicators to relay the board’s status. It is based on the same […]

M5Stack CoreMP135 – A Linux-powered industrial controller based on STM32MP135 Cortex-A7 MPU

M5Stack CoreMP135 industrial control host

M5Stack CoreMP135 is an industrial control host powered by the STM32MP135DAE7 Arm Cortex-A7 core microprocessor running at 1GHz, equipped with up to 512MB DDR3L SDRAM memory, and loaded with high-performance interfaces such as two Gigabit Ethernet ports, three USB ports, two CAN FD interfaces, two Grove interfaces, and an “HD” video output. An integrated PWR485 communication board bundles a 9V to 24V power input and an RS485 interface. The device also features a microSD card slot for storage, a small IPS capacitive touch screen, and a 1W speaker for human-machine interaction. The CoreMP135 is designed for low-power consumption and uses an Allwinner AXP2101 chip for power management. It supports scheduled wake-up and sleep with an integrated real-time clock (BM8563 module). The device runs Linux and comes with a microSD card loaded with the Debian operating system, simplifying setup and allowing usage out of the box. A DIN rail base plate […]

Quectel BG95-S5 5G NTN + LTE IoT module can seamlessly connect across satellite and cellular networks

Quectel BG95 S5 NTN satellite communication module

The Quectel BG95-S5 is a “multi-mode” 5G NTN satellite + LTE IoT communication module designed for seamless connectivity in remote areas. It supports 3GPP Release 17 IoT-NTN (S and L band frequencies) and offers fallback options with LTE Cat M1, Cat NB2, eGPRS, and integrated GNSS. These features make the module suitable for asset management, wireless POS, heavy industry, and wearables applications. In January this year, Quectel announced their CC660D-LS IoT NTN module in partnership with Skylo which focuses solely on NTN connectivity, whereas the “multi-mode” BG95-S5 seamlessly switches between different connectivity technologies (NTN, LTE Cat M1, Cat NB2, eGPRS). Quectel BG95-S5 IoT-NTN satellite module specifications: Module Type – NTN satellite communication module Compliance – Supports 3GPP Rel-17 IoT-NTN/ LTE Cat M1/ Cat NB2/ EGPRS module with Ultra-low power consumption Frequency Bands IoT-NTN S-Band – B23/ B256 L- Band – B255 (Under development/ planning) LTE-FDD Cat M1 – B1/ 2/ […]

STMicro LSM6DSV32X Edge AI motion sensor aims to extend battery life in wearables, trackers, and activity monitors

STMicro LSM6DSV32X AI motion sensor

STMicro LSM6DSV32X is a new low-power 6-axis inertial module with the company’s machine-learning core (MLC) with AI algorithms based on decision trees, a 3-axis accelerometer with a full-scale range of 32g, and a 4000 degrees-per-second gyroscope designed to measure intensive movements, impacts, and freefall height (estimation). The features of the sensors enable it to lower the power consumption for functions such as gym-activity recognition to under 6µA, while 3D orientation tracking consumes just 30µA when using STMicro’s Sensor Fusion Low-Power (SFLP) algorithm. It will be found in a range of battery-powered devices such as consumer wearables, asset trackers, and impact and fall alarms for workers or the elderly. STMicro LSM6DSV32X specifications: Triple-channel architecture for UI (user interface), EIS, and OIS data processing Data storage – Smart FIFO up to 4.5 KB Sensors Accelerometer – ±4/±8/±16/±32 g full scale Gyroscope – ±125/±250/±500/±1000/±2000/±4000 dps full scale Embedded temperature sensor Host interfaces SPI, […]

Silicon Labs BG22E, MG22E, FG22E wireless MCUs target energy harvesting, battery-free IoT devices

BG22E MG22E FG22E energy harvesting wireless microcontroller

Silicon Labs xG22E is a family of wireless SoCs consisting of the BG22E, MG22E, and FG22E and designed to operate within an ultra-low power envelope required for battery-free, energy harvesting applications such as electronic shelf labels, Smart Home sensors, remote controls, and so on. Like the just-announced Silicon Labs MG26, BG26, and PG26, the new xG22E family features a Cortex-M33 core clocked up to 76.8MHz. As an evolution of the low-power BG22, the BG22E offers Bluetooth 5.4 LE connectivity with “Energy Conservation” to further enhance battery longevity and even support designs that eliminate the need for batteries. The MG22E is similar but embeds an 802.15.4 radio for Zigbee connectivity, and the FG22E provides a 2.4GHz proprietary plus the Cortex-M33 is clocked up to 38.4MHz. Silicon Labs BG22E, MG22E, FG22E specifications: MCU – Arm Cortex-M33 @ 76.8 MHz with DSP and FPU (38.4 MHz for FG22E) Memory – Up to 32 […]

SB Components LoRaWAN gateways and nodes are made for Raspberry Pi and ESP32 boards (Crowdfunding)

lorawan product series

UK-based hardware developer, SB Components, has designed a new LoRaWAN product series (gateways and nodes) for the Raspberry Pi SBCs, Raspberry Pi Pico, ESP32, and other hardware, based on RAKWireless RAK5146 and RAK3172 modules. The products are available in up to five variants (plus two relay boards) and are built to cater to hobbyists with different needs. They support several LoRaWAN server platforms including The Things Stack, Chirpstack, and Helium, with adaptive spreading factors, coding rates, and bandwidth configurations. The LoRaWAN products include: Gateways – RAK5146 LoRaWAN Gateway HAT and RAK3172 LoRaWAN Gateway HAT for the Raspberry Pi SBCs Nodes – RAK3172 LoRaWAN Module (Powered by Raspberry Pi Pico), Raspberry Pi RP2040 USB Dongle, RAK3172 LoRaWAN Module (powered by ESP32), LoRaWAN Breakout, GatePi LoRaWAN 4-Ch Relay, GatePi LoRaWAN 8-Ch Relay LoRaWAN (long-range wide access network) uses the LoRa modulation technique to transmit data over large distances. In a LoRAWAN network, […]