cnx-software.com10h ago

Bit-Brick Cluster K1 – A 4-slot RISC-V cluster board for SpacemiT K1-based SSOM-K1 system-on-module

https://web.brid.gy/r/https://www.cnx-software.com/2025/12/28/bit-brick-cluster-k1-a-4-slot-risc-v-cluster-board-for-spacemit-k1-based-ssom-k1-system-on-module/

Bit-Brick Cluster K1 – A 4-slot RISC-V cluster board for SpacemiT K1-based SSOM-K1 system-on-module

Bit-Brick Cluster K1 is a cluster board designed to mount up to four SSOM-K1 system-on-modules powered by a SpacemiT K1 octa-core RISC-V processor. The board targets developers, researchers, and system integrators working on edge computing, AI workloads, and high-performance embedded applications. The Cluster K1 relies on an onboard Gigabit Ethernet switch chip to interconnect four core boards, giving each board an independent network port for fast and stable inter-node communication with automatic IP assignment. Other key features include USB 3.0/2.0 ports, HDMI output (master slot only), a USB Type-C debug port, an M.2 M-Key slot for storage expansion, a 3-pin fan power header, and dedicated power, reset, and download buttons. The SSOM-K1 system-on-module features a K1 octa-core 64-bit RISC-V CPU with vector extension and AI acceleration up to 2 TOPS. It ships with up to 8GB dual-channel LPDDR4X memory, up to 64GB eMMC flash, and offers dual Gigabit Ethernet, HDMI 1.4

CNX Software - Embedded Systems News
MottG1d ago

"UOsaka breatkthrough: World’s fastest and most accurate self-evolving edge AI for real-time forecasting"

New MicroAdapt technology enables real-time learning and forecasting on compact devices, proving up to 100,000 times faster and 60% more accurate than state-of-the-art deep learning methods. The team successfully implemented this self-evolving edge learning mechanism on a Raspberry Pi 4.

https://www.eurekalert.org/news-releases/1103952

#research #AI #edgeAI

Sparknify2d ago

Most hardware startups don’t fail on tech—they fail after the demo.

Prototype works. Pitch lands. Then: costs spike, suppliers misalign, heat shows up, certification drags, runway leaks.

That’s the Prototype Trap—stuck between proof and production.

Read: https://www.sparknify.com/post/20251226-prototype-trap-en

💡 Demos get attention. 🏭 Products survive reality.
#HardwareStartups #PrototypeTrap #DeepTech #EdgeAI #Robotics #ManufacturingMatters #TaiwanTech #ICTGC

Jason2d ago

Traffic demand changes in seconds — lane control should too 🚦

The FCU3501 AI edge computing box brings real-time, autonomous decision-making to intelligent flexible lanes using edge AI, not cloud latency.
https://www.forlinx.net/industrial-news/fcu3501-ai-edge-computing-intelligent-flexible-lanes-762.html

#EdgeAI #SmartTraffic #SmartCity #EdgeComputing

sayzard4d ago

Alex Cheema - e/acc (@alexocheema)

프론티어 LLM을 개인 하드웨어, 특히 Apple Silicon 기반 Mac에서 구동해야 하는 이유를 설명합니다. 메모리 및 메모리 대역폭 단가, 저배치(batch)에서의 성능, 전력 대비 처리량(watt/tok/sec) 등 Apple Silicon이 비용·효율 측면에서 유리하고 Mac은 재판매 가치도 높다고 주장합니다.

https://x.com/alexocheema/status/2003558040503746905

#applesilicon #llm #mac #edgeai

Alex Cheema - e/acc (@alexocheema) on X

Frontier LLMs on your own hardware. Reasons to run on Apple Silicon: - Best memory & memory-bw unit economics, best for low batch sizes - Best watt/gb/sec -> best watt/tok/sec - Macs retain resell value far better than custom GPU rigs - Off-the-shelf hardware, all you need is

X (formerly Twitter)
sayzard4d ago

Q*Satoshi (@AiXsatoshi)

약 1만 달러짜리 GPU가 대거 늘어나고 있다는 관찰과 함께, 600W 전원 5대로 3000W를 넘기는 자작 PC로 '오프그리드' 방식으로 LLM을 운영하려는 시도가 소개됩니다. 고전력 소비를 감수한 개인/소규모 구축을 통해 독립적으로 대형 언어모델을 구동하려는 하드웨어·배포 실험에 대한 기대감을 표현한 게시물입니다.

https://x.com/AiXsatoshi/status/2003584584089477178

#gpu #llm #hardware #edgeai

Q*Satoshi⏩ (@AiXsatoshi) on X

約1万ドルのGPUがめちゃ増えてる…しかも600W x 5台で3000Wオーバーの自作PCでオフグリッドLLM運用とか…ヤバすぎますね、超期待してます

X (formerly Twitter)
ArmSoM_Official4d ago
👍When hardware says "unsupported", hack it! A brilliant reverse-engineering of the RK3588 NPU to run Vision Transformers 15x faster. A masterclass in unlocking edge AI potential. #RK3588 #NPU #EdgeAI
https://www.reddit.com/r/LocalLLaMA/comments/1pkhzf0/reverseengineering_the_rk3588_npu_hacking_memory/
TechLİfe4d ago

AI Training vs Inference: Why 2025 Changes Everything for Real-Time Applications

https://techlife.blog/posts/ai-training-vs-inference-why-2025-changes-everything-for-real-time-apps/

#ArtificialIntelligence #AIInference #MachineLearning #AITraining #RealTimeAI #EdgeAI #AI #LLM

AI Training vs Inference: Why 2025 Changes Everything for Real-Time Applications

Discover why AI inference is overtaking training as the dominant workload in 2025. Learn the key differences, cost dynamics, and infrastructure shifts reshaping the AI industry.

TechLife
cnx-software.com6d ago

ESP32-S3 AIoT Basic – A learning and prototyping kit with camera, audio, LCD, and sensors

https://web.brid.gy/r/https://www.cnx-software.com/2025/12/22/esp32-s3-aiot-basic-a-learning-and-prototyping-kit-with-camera-audio-lcd-and-sensors/

ESP32-S3 AIoT Basic – A learning and prototyping kit with camera, audio, LCD, and sensors

The ESP32-S3 AIoT Basic is a low-cost, learning, and prototyping kit for the ESP32-S3. The board integrates common AIoT peripherals directly onto a single PCB, making the design part easy for beginners, classrooms, and rapid prototyping. Built around an ESP32-S3 board, the development platform integrates nine commonly used modules directly on the PCB, including a button, buzzer, LED indicator, light sensor, LCD, digital microphone, SD card slot, audio amplifier, and a camera. Most AI and IoT demos can be run without breadboards or jumper wires, while expansion is supported through standard pin headers and Grove connectors. The board supports 5V power via USB-C, and 6–12 V power input via Vin for driving additional devices. With various tutorials and sample projects, it is suitable for AIoT learning, STEM education, voice and vision demos, sensor-based projects, and quick proof-of-concept development. ESP32-S3 AIoT Basic Specifications: Main board - ESP32-S3 Core Board SoC –

CNX Software - Embedded Systems News
cnx-software.com6d ago

Forlinx FCU3011 – An NVIDIA Jetson Orin Nano fanless industrial computer with 4x GbE, optional 4G/5G and Wi-Fi connectivity

https://web.brid.gy/r/https://www.cnx-software.com/2025/12/22/forlinx-fcu3011-an-nvidia-jetson-orin-nano-fanless-industrial-computer-with-4x-gbe-optional-4g-5g-and-wi-fi-connectivity/

Forlinx FCU3011 – An NVIDIA Jetson Orin Nano fanless industrial computer with 4x GbE, optional 4G/5G and Wi-Fi connectivity

Forlinx Embedded has recently released the FCU3011, a compact, fanless industrial AI edge computer built around the NVIDIA Jetson Orin Nano, designed for 24/7 operations in manufacturing, smart cities, robotics, and machine vision systems, where real-time processing is needed. The fanless system supports NVIDIA Jetson Orin Nano 4GB (34 TOPS) or 8GB (up to 67 TOPS) configurations, with 4GB/8GB LPDDR5 memory and a 128GB PCIe x4 NVMe SSD. Connectivity options include up to four Gigabit Ethernet ports, USB 3.0/2.0, HDMI 2.0 (4K), an SD card slot, optional 4G/5G and dual-band Wi-Fi via M.2 modules, along with industrial interfaces such as isolated RS-485, CAN, opto-isolated inputs, relay outputs, and an RTC. The system takes a wide 9–24V DC power input, features ESD-protected interfaces, and can be used for AGVs, visual inspection, smart factories, intelligent traffic analysis, medical devices, and small commercial robots. Forlinx FCU3011 specifications: SoM options NVIDIA Jetson Orin Nano

CNX Software - Embedded Systems News