NVIDIA container runtime for Wind River Linux

Training and using AI models are tasks that demand significant computational power. Current trends are pointing more to deep neural networks, which include thousands, if not millions of operations per iteration. In the past year, more and more researchers have sounded the alarm on the exploding costs of deep learning. The computing power needed to do AI is now rising seven times faster than ever before [1]. These new needs are making hardware companies create hardware accelerators like Neural processing units, CPUs, and GPUs.

Embedded systems are not an exception to this transformation. We see every day intelligent traffic lights, autonomous vehicles, intelligent IoT devices, and more. The current direction is to have accelerators inside these embedded devices, Systems On-Chip mainly. Hardware developers have embedded small accelerators like GPUs, FPGAs, and more into SOCs, SOMs, and other systems. We call these modern systems: heterogeneous computing architectures.

The use of GPUs on Linux is not something new; we have been able to do so for many years. However, it would be great to accelerate the development and deployment of HPC applications. Containers enable portability, stability, and many other characteristics when deploying an application. For this reason, companies are investing so much in these technologies. For instance, NVIDIA recently started a project that enables CUDA on Docker [2].

One concern when dealing with containers is the loss of performance. However, when comparing the performance of the GPU with and without the containers environment, researchers found that no additional overhead is caused [3]. The consistency in the performance is one of the principal benefits of containers over virtual machines; accessing the GPU is done seamlessly as the kernel stays the constant.

Spherical Quadruped Arduino Robot

[Greg06] started learning electronics the same way most of us did: buy a few kits, read a few tutorials, and try your hardest to put a few things together. Sound familiar? After a while, you noticed your skills started increasing, and your comfort level with different projects improved as well. Eventually, you try your hand at making your own custom projects and publishing your own tutorials.

96Boards CE Extended SBC taps the quad -A7 Quectel SC20 4G module

Shiratech’s “SRT-96B-MAIN-SC20-E/A” is a 96Boards CE Extended SBC that runs Android on a Quectel SC20 module equipped with a quad -A7 Snapdragon 210, 4G LTE, GNSS, and WiFi/BT. The SBC features a CSI cam, a MEMs mic, accelerometer, and photometric sensor.

Last week we reported on Shiratech’s Linux-driven Stinger96 96Boards IoT Edition SBC with a dual Cortex-A8 STM32MP157 SoC and a Quectel BG96 LTE CAT-M1 radio for NB-IoT communications. We now see that Shiratech has posted specs for a SRT-96B-MAIN-SC20-E/A SBC, AKA the 96Board Main with SC-20. The board uses the 96Boards Consumer Edition Extended form factor and offers 4G LTE via a Quectel SC20 module that integrates a 1.1GHz, quad -A7 Snapdragon 210.

Advantech Edge AI Computer and PoE NVR Feature NVIDIA Jetson Xavier NX SoM

Following the launch of NVIDIA Jetson Xavier NX SoM last year, we noted several third-party carrier boards and embedded PCs had been announced, and we expected more to come soon. Advantech has now unveiled two systems based on NVIDIA latest module with respectively MIC-710AIX edge AI computer and MIC-710IVX NVR system equipped with eight PoE ports.

Librem 5 Update: Fresh Dogwood Pictures

As we mentioned in our Dogwood update post, we have been busy testing the significant changes that have gone into the Dogwood batch. In the previous post we just showed a few pictures of the board with some testing wires attached but we figured you’d like to see more. Now that we are about ready to wrap up testing we wanted to share some additional Dogwood pictures. Like with previous batches Dogwood is a small batch process. We’ll see mass-production processes with the next batch Evergreen.

MARK AI Robot Kit Aims to Teach AI & Robotics to 12+ Years Old (Crowdfunding)

Tinkergen offers pre-trained model to recognized objects like humans, books, pens, or smartphones, as well as traffic signs, numbers from 1 to 9, as well as domestic and wild animals. As such, the company also provides signs, as well as optional illustrated and numbered cards together with magnets and card stands to facilitate various AI recognition applications.

Ameba RTL8722DM Cortex-M33/M23 IoT Development Board Now Available for $23.90

Earlier this year, we noticed some RealTek RTK8720DN dual-band WiFi and Bluetooth 5.0 IoT modules, and quickly mentioned Ameba RTL8722DM development board available for under $50.

[...]

It ships with one IPEX antenna, as well as two Arduino female headers that you’d need to solder if you plan on using Arduino shields. Seeed Studio also explains the final version of the board will have all the GPIO pins bent at an angle of 90° from the board to allow convenient access to the Arduino pins.

Fanless edge server has 24-core Arm SoC and 3-TOPS-per-Watt NPU

Foxconn, Socionext, and Hailo are launching a fanless “BOXiedge” AI edge server with Socionext’s Linux-driven SynQuacer SC2A11 SoC with 24x Cortex-A53 cores. The system also features an up to 26-TOPS Hailo-8 NPU for analyzing 20 streaming cam feeds in real time.

In January, Foxconn, Socionext, and Network Optix announced a BOXiedge AI video analysis solution that combined a Foxconn edge computer equipped with an unnamed AI acceleration card and Socionext’s SC2A11 Arm server SoC. The BOXiedge also incorporated Network Optix’s Ubuntu compatible Nx Witness v4.0 video processing and analytics software. Now Foxconn and Socionext have followed up with a new BOXiedge model that omits the Nx Witness v4.0 stack and replaces the AI card with an up to 26 TOPS Hailo-8 deep learning processor.

Make it rain chocolate with a Raspberry Pi-powered dispenser

$31 Honor Router 3 WiFi 6+ Router Supports 3000 Mbps Data Rates

I wrote about the “Most Affordable WiFi 6 Router Yet“, namely Xiaomi Mi Router AX1800, just two days ago, as it sells for around $60 on Aliexpress and 329 CNY ($46) in China. But the Xiaomi router did not hold the top spot for long, as Honor Router 3 WiFi 6 AX3000 router was just launched for 219 CNY (about $31 US) in China.

Coffee Lake Box PC offers PCIe and PCI expansion

Ibase’s semi-rugged “AMS210” box PC runs on 8th or 9th Gen Coffee Lake CPUs with features including dual SATA bays, 4x GbE, 6x USB, 3x DP, and PCI and PCIe expansion slots.

Ibase announced an industrial embedded box PC and “automatic control system” for factory automation, machine vision, digital signage, and industrial IoT applications. The AMS210 has a lot of in common with last month’s AMI230 series, including the 8th and 9th Gen Coffee Lake CPUs, dual SATA bays, 4x GbE ports, and PCIe expansion slots. No OS support is listed, but we imagine that like the AMI230 series, the AMS210 supports Linux and Windows.

Olimex ESP32-S2-Devkit-LiPo WiFi Board Consumes as Little as 2uA in Sleep Mode

When we covered ESP32 powered TTGO T-Watch-2000 smartwatch this week-end, people noted that with a 350 mAh battery, the watch would last about 3.65 days considering a 4mA drain with the screen always off, WiFI and Bluetooth off, and around 65mA when the screen is on good for about 5 hours of continuous use without Bluetooth nor WiFi.

[...]

Despite the added battery charger and circuitry, ESP32-32-Devkit-LiPo will have the same size as ESP-32-Saola-1R, and use the same ESP32-S2-WROVER module with 2MB RAM and 4MB Flash.

Coolest Projects goes online and everyone is welcome!