AAEON’s RTC-1010RK is a 10.1-inch (1280 x 800) rugged tablet powered by a Rockchip RK3399 processor and can be equipped with up to 4GB LPDDR4 memory and 128GB eMMC flash. It offers robust connectivity with Wi-Fi, Bluetooth, and optional 4G LTE and Ethernet. It also supports GPS + GLONASS (Default), and BeiDou (Optional) for navigation.
The RTC-1010RK is an IP65-rated tablet with protection against dust and water splashes. Additionally, it has programmable function keys, NFC/smart card readers, f
AAEON’s RTC-1010RK is a 10.1-inch (1280 x 800) rugged tablet powered by a Rockchip RK3399 processor and can be equipped with up to 4GB LPDDR4 memory and 128GB eMMC flash. It offers robust connectivity with Wi-Fi, Bluetooth, and optional 4G LTE and Ethernet. It also supports GPS + GLONASS (Default), and BeiDou (Optional) for navigation.
The RTC-1010RK is an IP65-rated tablet with protection against dust and water splashes. Additionally, it has programmable function keys, NFC/smart card readers, front (8MP) and rear (8MP) cameras, and the option to add a barcode scanner. The hot-swappable battery in the tablet ensures uninterrupted power, making it ideal for demanding field and industrial applications.
Vibration –MIL-STD-810H: 2019, 514.8 Procedure I, Table 514.8C-I Category 4 common carrier
Drop – MIL-STD-810H: 2019, 516.8 Procedure IV Drop height: 122cm Number of Drop: 26 times, for all surfaces, edges, and corners Condition: Based on 2“ plywood over concrete
SHOCK – MIL-STD-810H: 2019, 516.8 Procedure V
ESD – Air charge: +/- 8KV, Contact charge: +/- 4KV
Certifications and Standards – UL62368-1, CE, FCC, IC
The company mentions that the RTC-1010RK will ship with Android 11, and you don’t have to worry about drops, shocks, or vibrations because this tablet is MIL-STD-810H (a military standard that tests a product’s ability to withstand environmental conditions throughout its lifetime) tested for extreme environments.
At the time of writing the AAEON has not declared any pricing information for the tablet, but if you want to get a quote or have any inquiries you can do that from their product page where you’ll also find a datasheet and a manual for download.
ESParagus Media Center is a line of audio streamers from Sonocotta, all powered by an ESP32 microcontroller module. It includes the ESParagus HiFi MediaLink, Loud ESParagus, and the Louder ESParagus.
Louder ESParagus
The ESP32-based audio centers can be used to power old stereo speaker systems that lack streaming capabilities. They are completely open-source, consume little power when not in use, and boot up in seconds.
The ESParagus Media Center products are based on the ESP32-WROVER microcontr
ESParagus Media Center is a line of audio streamers from Sonocotta, all powered by an ESP32 microcontroller module. It includes the ESParagus HiFi MediaLink, Loud ESParagus, and the Louder ESParagus.
Louder ESParagus
The ESP32-based audio centers can be used to power old stereo speaker systems that lack streaming capabilities. They are completely open-source, consume little power when not in use, and boot up in seconds.
The ESParagus Media Center products are based on the ESP32-WROVER microcontroller module with Wi-Fi and Bluetooth connectivity and an onboard PSRAM chip. They are fitted with an external Wi-Fi antenna and the top-end model – the Louder ESParagus – is fitted with a W5500 LAN chip for Ethernet networking.
All three ESParagus Media Centers run squeezelite-esp32 firmware which supports Spotify Connect, Apple AirPlay, and Logitech Media Server. Integrations with Home Assistant are possible and can be useful for multi-room configurations.
The Louder ESParagus is quite similar to the Louder Raspberry Pi media center we covered earlier, but it is powered by an ESP32 microcontroller rather than a single-board computer. Other audio streamers we have taken a look at include the PecanPi audio streamer and the Volumio Motivo.
ESParagus Media Centers specifications:
Microcontroller – ESP32-WROVER microcontroller module, based on ESP32 dual-core 32-bit processor @ up to 240 MHz, up to 8MB PSRAM
Flash – 16MB
PSRAM – 8MB
DAC
ESParagus HiFi MediaLink – TI PCM5100A 32-bit Stereo DAC (with -100 dB typical noise level)
ESParagus HiFi MediaLink and Loud ESParagus – 80 x 50 x 20 mm
Louder ESParagus – 100 x 80 x 38 mm
As mentioned earlier, the ESParagurus products are fully open-source, and design files, documentation, and other information can be found in the project’s GitHub repository and Hackaday.
The ESParagus Media Center lineup of ESP32 audio streamers has been launched on CrowdSupply with a $5000 funding goal. The ESParagus HiFi MediaLink and Loud ESParagus audio streamers are both priced at $45 while the Louder ESparagus version costs $69. There is free shipping within the United States and a $12 shipping fee applies to the rest of the world. Orders are expected to ship by November 12, 2024.
WeAct ESP32-C6-Mini is a small development board based on Espressif Systems ESP32-C6 WiFi 6, Bluetooth, and 802.15.4 wireless SoC that’s the cheapest we’ve seen so far selling for just $3.86 on Aliexpress.
WeAct was one of the first companies to launch a third-party ESP32-C6 development board year with the price starting at just $6. I still remember the excitement around the first $5 ESP8266 WiFi 4 module when we first covered it in 2014, and ten years later, we can get a full $4 development boa
WeAct ESP32-C6-Mini is a small development board based on Espressif Systems ESP32-C6 WiFi 6, Bluetooth, and 802.15.4 wireless SoC that’s the cheapest we’ve seen so far selling for just $3.86 on Aliexpress.
WeAct was one of the first companies to launch a third-party ESP32-C6 development board year with the price starting at just $6. I still remember the excitement around the first $5 ESP8266 WiFi 4 module when we first covered it in 2014, and ten years later, we can get a full $4 development board with a more powerful 160 MHz RISC-V microcontroller, WiFi 6, Bluetooth 5.0, and an 802.15.4 radio for Thread, Zigbee, and Matter.
WeAct focuses on designing and manufacturing cheap hardware, but I don’t think I’ve ever seen documentation from the company. They only say the board can be programmed with the ESP-IDF v5.1 and greater framework. However, considering it’s basically a cost and features-down version of the ESP32-C6-DevKitC-1 development kit people can use the official documentation from Espressif, and the Arduino IDE should also be supported as long as you install ESP32 Arduino Core 3.0.0 or greater.
The $3.86 price tag does not include a $0.99 shipping fee, but in my case at least (shipping to Thailand), it’s possible to get free shipping by ordering at least three boards making the order over $10.
Geekworm X1011 is a new expansion board for the Raspberry Pi 5 with four M.2 sockets enabling the insertion of up to four M.2 NVMe SSDs with data pushed through the PCIe Gen2 interface of the popular SBC.
We were already wondering why most people would want to connect two NVMe SSDs to the Raspberry Pi 5 when the Geekworm X1004 HAT+ was launched considering the 5GT/s limitation from the board and the PCIe switch, but the company decided to double the number of drives with the X1011 meaning each d
Geekworm X1011 is a new expansion board for the Raspberry Pi 5 with four M.2 sockets enabling the insertion of up to four M.2 NVMe SSDs with data pushed through the PCIe Gen2 interface of the popular SBC.
We were already wondering why most people would want to connect two NVMe SSDs to the Raspberry Pi 5 when the Geekworm X1004 HAT+ was launched considering the 5GT/s limitation from the board and the PCIe switch, but the company decided to double the number of drives with the X1011 meaning each drive can achieve up to around 100 MB/s (or 400MB/s) when used simultaneously. It does look nice and fairly compact though.
Geekworm X1011 specifications:
Supported SBC – Raspberry Pi 5 and other SBCs with a compatible 16-pin PCIe FPC connector and mounting holes
Chipset – ASMedia ASM1184e PCI express packet switch with 1x PCIe Gen2 x1 upstream port and 4x PCIe Gen2 x1 downstream ports
4x M.2 sockets for up to 16TB storage capacity (4x 4TB) with M.2 NVMe 2280/2260/2242/2230 SSDs (SATA not supported)
Note – NVMe boot is not supported because the Raspberry Pi firmware does not currently support PCIe switches.
Power Management
5V/5A DC via FFC & pogo pins (using the USB-C port on the Pi 5)
5V/5A DC via 5.5/2.1mm DC power jack
DC/DC step-down converter delivering a maximum of 10A to power the SSDs
Compatible with the HAT+ standby power state, automatically turning off when the Raspberry Pi 5 shuts down.
Important: Do not power the Raspberry Pi 5 through its USB-C port at the same time if you are using the DC jack.
Dimensions – 109 x 87.2mm
My favorite part of the design are the three pogo pins used to power the expansion board.GEEKOM X1011 and accessories
While the Raspberry Pi 5’s PCIe interface could be configured up to switch to PCIe Gen 3 x1, this won’t help because the PCIe switch only supports PCIe Gen 2 x1. That means we are left with a solution that delivers about the same sequential read/write performance as SATA hard drives, albeit in a much smaller form factor and random I/Os will still be considerably faster. A wiki has been set up with more technical details and instructions to get started.
Jeff Geerling got a sample and confirmed the performance. He also compared it to the Rockchip RK3588-powered FriendlyELEC CM3588 NAS Kit which is about $20 more expensive for a complete system, but with a PCIe Gen3 x4 interface (PCIe Gen3 x1 (8GT/s) per socket) that delivers much better performance and a 2.5GbE port that can push 2.5 times more data to the network than the gigabit Ethernet on the Raspberry Pi 5. None of these support SATA drives, and if you prefer this type of drive, the Radxa Penta SATA HAT provides a viable option to build a NAS for the Raspberry Pi 5.
The Geekworm X1011 expansion board ships with two PCIe FFC cables (one is a spare), various screws, spacers, and nuts. It can be purchased on Geekworm for $51, on Aliexpress for $50, and might soon be found on the company’s Amazon store.
Qingping CGDK2 and Xiaomi LYWSD03MMC Bluetooth LE (BLE) temperature and humidity monitors based on Telink TLSR8258 can be switched to Zigbee thanks to a custom firmware flashed over-the-air.
Telink TLS8258 is a multi-protocol wireless microcontroller supporting Bluetooth LE 5, Bluetooth Mesh, Zigbee, RF4CE, Thread, 6LoWPAN, HomeKit, ANT, and 2.4GHz proprietary and found in boards such as the LILYGO T-Zigbee and various products. Some products only enable one wireless protocol, for example, Bluet
Qingping CGDK2 and Xiaomi LYWSD03MMC Bluetooth LE (BLE) temperature and humidity monitors based on Telink TLSR8258 can be switched to Zigbee thanks to a custom firmware flashed over-the-air.
Telink TLS8258 is a multi-protocol wireless microcontroller supporting Bluetooth LE 5, Bluetooth Mesh, Zigbee, RF4CE, Thread, 6LoWPAN, HomeKit, ANT, and 2.4GHz proprietary and found in boards such as the LILYGO T-Zigbee and various products. Some products only enable one wireless protocol, for example, Bluetooth LE as in the just-mentioned Qingping and Xiaomi monitors, but “SmartHomeScene” has found out that it was possible to switch from BLE to Zigbee on those by simply updating the firmware.
The Telink TLS8258 devices mentioned above will publish data over BLE every 10 minutes (unless humidity or temperature changes, in which case data is transmitted immediately) with the default firmware, but Zigbee can be enabled by using the pvvx custom firmware for the following models:
Xiaomi LYWSD03MMC
Qingping CGDK2
Xiaomi MiaoMiaoCe MHO-C401
Xiaomi MiaMiaoCe MHO-C122
Both custom BLE and Zigbee firmware are available for these devices, and you have to select one of the wireless standards. There’s no need to open the device although flashing through a USB-to-serial adapter is possible, and most people will prefer to go with one of the two OTA firmware update methods, either the “devbis” or “pvvx” method as explained in the aforelinked post by SmartHomeScene along with the instructions to connect the Zigbee devices with Home Assistant through the Zigbee2MQTT integration.
The switch is mostly useful for people preferring to use Zigbee over BLE in their Smart Home setup. Battery life should not be impacted too much, and the pvvx firmware’s features list mentions over one year of battery with either BLE or Zigbee.
The Waveshare UGV Rover is a 6-wheel robot platform based on Raspberry Pi 4 or 5 as well as an ESP32 module and built for remote exploration, object recognition, and autonomous navigation. Since the source code for the platform will be open-sourced it can also be used for educational purposes, programming, robotics, AI experimentation, and many other applications.
This Unmanned Ground Vehicle (UGV) rover features a 2mm thick aluminum body, six 80mm shock-absorbing tires, and a four-wheel drive s
The Waveshare UGV Rover is a 6-wheel robot platform based on Raspberry Pi 4 or 5 as well as an ESP32 module and built for remote exploration, object recognition, and autonomous navigation. Since the source code for the platform will be open-sourced it can also be used for educational purposes, programming, robotics, AI experimentation, and many other applications.
This Unmanned Ground Vehicle (UGV) rover features a 2mm thick aluminum body, six 80mm shock-absorbing tires, and a four-wheel drive system controlled by an ESP32 sub-controller. The sub-controller also handles sensors, LiDAR, cameras, and more. The brain or the main controller of the rover is a Raspberry Pi SBC – either a Pi 4B or Pi 5 – which notably handles computer vision and machine learning operations.
Gigabit Ethernet, WiFI 5, and Bluetooth 5 on Raspberry Pi 4/5 SBC
2.4 GHz WiFi and Bluetooth on ESP32 including ESP-NOW support
Optional 4G LTE/5G module
Expansion GPIO – 40-pin GPIO extended header
Rover Construction
Made with 2mm thick aluminum plates
High-torque encoder motor with 80mm shock-absorbing tire
6-wheel with 4-wheel drive setup for enhanced maneuverability
High-brightness LED light for clear images in low-light conditions
Additional Pan-Tilt module features 360° horizontal and 120° vertical rotation with a 5MP camera and vertical stabilization for better camera control
Additional mounting plate for installing D500 / STL27L LiDAR, battery, or other peripherals
Optional 21mm wide rail and 30 KG/CM high precision & high-torque bus servo for tactical extension
Rover Speed – 0.5m/s max
Misc
Comes with a wireless gamepad
Automatically creates hotspot when there is no network available
Power Options
3x 18650-based UPS module with support for charging and discharging at the same time
Optional battery set with XH2.54 interface
Dimensions – 230.42mm x 252.40mm x 254.53mm
Another cool feature of this rover is that it features ESP-NOW protocol that enables inter-rover communication with minimum latency while allowing 4G/5G module expansion for communication.
UGV Rover Motor and Audio Driver Control Boards
The ESP32 powred sub-controller supports different communication protocols, including serial port, HTTP request, and ESP-NOW through which provides some basic control of the robot with a web interface, it also performs basic tasks for the robot, such as high-frequency PID controller, high-frequency inverse kinematics calculations for robotic arm linkages, position interpolation calculation, Pan-Tilt angle control, and OLED screen control, etc. Additionally, it is responsible for reading data from IMU (9-axis attitude sensor) and INA219 (battery voltage sensor), and can automatically achieve the camera vertical stabilization function.
UGV Rover Web Application
You can control the robot through a web application, compatible with standard browsers, which provides for high-frame-rate video transmission and AI-enhanced computer vision functionalities.
OpenCV based recognition, tracking, and targeting
Additional AI features include object recognition, gesture recognition, face detection, motion detection, vision line tracking, color recognition, auto-targeting, and multi-threaded computer vision.
The company mentions that they will provide tutorials for Ngrok, but won’t provide any Ngrok accounts or servers. They also explain that the eRover is compatible with Debian Bookworm and ROS2-HUMBLE-LTS, as well as JupyterLab. We should eventually expect a guide and tutorials for all the software features, but at the time of writing that is not available on their Wiki page.
Waveshare gives us custom options for the UGV rover. If you already own a Raspberry Pi, you can order the rover without the Pi and there is also an option to include the Pan-Tilt Module in your order. The company also provides options for US, EU, and UK power plugs. This type of Raspberry Pi 4/5-based robot can be a lot of fun and we recently reviewed the SunFounder PiCar-X 2.0 which looks to have better documentation than the Waveshare model at the time of writing.
The UGV Rover PT PI5 AI Kit or UGV Rover PT PI4B AI Kit, including the Pan-Tilt Module, can be found on Amazon for $292.99 (including shipping). The robot is not available on the company’s Aliexpress store just yet, but a similar model with four wheels called the RaspRover can be purchased for $212 and up. Waveshare also offers a UGV Rover base kit (without Pi or Pan-Tilt Module) for $244.99 on their store.
Banana Pi BPI-M5 Pro, also known as Armsom Sige5, is a low-profile single board computer (SBC) powered by the Rockchip RK3576 octa-core Cortex-A72/A53 SoC for the AIoT market that offers a mid-range offering between Rockchip RK3588 and RK3399 SoCs.
The board comes with 16GB LPDDR4X and 128GB eMMC flash by default, offers dual GbE, WiFi 6 and Bluetooth 5.2 connectivity, an M.2 2280 PCIe socket for expansion, HDMI and MIPI DSI display interfaces, two MIPI CSI camera interfaces, a few USB ports, an
Banana Pi BPI-M5 Pro, also known as Armsom Sige5, is a low-profile single board computer (SBC) powered by the Rockchip RK3576 octa-core Cortex-A72/A53 SoC for the AIoT market that offers a mid-range offering between Rockchip RK3588 and RK3399 SoCs.
The board comes with 16GB LPDDR4X and 128GB eMMC flash by default, offers dual GbE, WiFi 6 and Bluetooth 5.2 connectivity, an M.2 2280 PCIe socket for expansion, HDMI and MIPI DSI display interfaces, two MIPI CSI camera interfaces, a few USB ports, and a 40-pin GPIO header.
4x Cortex-A72 cores @ 2.2GHz, four Cortex-A53 cores @ 1.8GHz
Arm Cortex-M0 MCU at 400MHz
GPU – ARM Mali-G52 MC3 GPU with support for OpenGL ES 1.1, 2.0 and 3.2, OpenCL up to 2.0 and Vulkan 1.1
NPU – 6 TOPS (INT8) AI accelerator with support for INT4/INT8/INT16/BF16/TF32 mixed operations.
VPU
Video Decoder: H.264, H.265, VP9, AV1, and AVS2 up to 8K @ 30fps or 4K @ 120fps.
Video Encoder: H.264 and H.265 up to 4K @ 60fps, (M)JPEG encoder/decoder up to 4K @ 60fps.
System Memory – 8GB or 16GB 32-bit LPDDR4x
Storage
32GB or 128GB eMMC flash
MicroSD card slot
M.2 Key-M socket for M.2 2280 NVMe SSD (See Expansion section)
Footprint for UFS storage, but apparently unused
Video Output
HDMI 2.0 port up to 4Kp120
MIPI DSI connector up to 2Kp60
DisplayPort 1.4 via USB-C up to 4Kp120
Audio – Speaker header, digital audio output via HDMI
Camera I/F – 2x 2-lane MIPI CSI connectors
Networking
2x Gigabit Ethernet ports
WiFi 6 and Bluetooth 5.2 via Realtek RTL8852BS
USB – 1x USB 3.0 port, 1x USB 2.0 port, 1x USB Type-C port
Expansion
40-pin GPIO header
M.2 Key-M socket (PCIe 2.1 x1)
Misc
MaskROM button, Reset button, Power button
Fan connector
RTC battery connector
2x LED
Power Supply – 4.5V to 23V via USB Type-C PD port
Dimensions: 92 x 62mm (8-layer PCB)
Weight – 43 grams
Temperature Range – 0°C to 80°C
The Banana Pi BPi-MP5 Pro will support Android 14, Debian 11, and Buildroot through official Rockchip support, as well as third-party Armbian Ubuntu and Debian images. Work-in-progress documentation is available on both Banana Pi and ArmSom websites. I initially thought the latter was better with links to Android, Debian, Ubuntu, and OpenWrt, and some source code, but all those are for older SBCs from the companies…
Banana Pi previously released the BPI-M5 SBC with an Amlogic S905X3 SoC, so you’d assume the Pro version would have many features in common, but the two boards have completely different designs…
Block diagram for ArmSom Sige5 / Banana Pi BPI-M5 Pro
The Banana Pi BPI-M5 Pro is not for sale on SinoVoIP’s Aliexpress store just yet possibly because the OS images are not available yet, but you’ll find the ArmSom Sige5 Pro Max for pre-order for $148 with 16GB RAM and 128eMMC flash. It does not seem particularly good value as the Orange Pi 5 Pro with similar features (but only one GbE port) and a more powerful Rockchip RK3588S processor goes for $128 on Amazon in the same 16GB/128GB configuration. Maybe it is still worth it for use cases where the low-profile design may be important.
Taiwanese wireless equipment manufacturer, Alfa Networks, and US-based Wi-Fi HaLow semiconductor company, Newracom, have collaborated to develop the AHST7394S Wi-Fi HaLow module.
The solder-down module is based on Newracom’s NRC7394 SoC, a low-cost and high-efficiency HaLow SoC chipset providing up to 17dbm output power and capable of connecting to a maximum of 8K devices within a single network. The NRC7394 also supports a standalone mode, which enables the execution of a wide range of IoT app
Taiwanese wireless equipment manufacturer, Alfa Networks, and US-based Wi-Fi HaLow semiconductor company, Newracom, have collaborated to develop the AHST7394S Wi-Fi HaLow module.
The solder-down module is based on Newracom’s NRC7394 SoC, a low-cost and high-efficiency HaLow SoC chipset providing up to 17dbm output power and capable of connecting to a maximum of 8K devices within a single network. The NRC7394 also supports a standalone mode, which enables the execution of a wide range of IoT applications on the embedded ARM Cortex-M3 CPU. While HaLow is more energy-efficient than other Wi-Fi types, low-power operation modes such as legacy, TWT, and WMM-PS can reduce power consumption further.
The AHST7394S Wi-Fi HaLow module supports a data rate of up to 15Mbps, over 600 times faster than LoRaWAN while maintaining a good range (over 1km). The second-generation HaLow module is perfect for building long-range, ultra-low-power networks in sub-1 GHz, license-exempt frequency bands. It is a compact module, measuring 20 x 17mm, and supports Wi-Fi HaLow for North America, Europe, and Asia.
It is ideal for various IoT applications such as weather stations, industrial monitoring, medical patient monitoring, agriculture, and surveillance cameras.
The AHST7394S WiFi HaLow module comes in three different product variants, targeted at different regions: AHST7394S-915 (902 MHz – 928 MHz), AHST7394S-868 (863 MHz – 868 MHz), and AHST7394S-924 (920.5 – 927.5 MHz). The module’s product page is live on Alfa Networks’s website but there is no information about pricing and availability. However, you can request more information / ask for a quote via the Contact Us link.
Compulab IOT-DIN-IMX8PLUS is a DIN Rail IoT gateway powered by an NXP i.MX 8M Plus AI SoC that takes various expansion modules with digital inputs and outputs (DIO), RS232, RS485, or ADC.
The new model looks to be an evolution of the Compulab IOT-GATE-IMX8PLUS Arm Linux IoT gateway introduced in 2022 with many of the same features, except the IOT-DIN-IMX8PLUS is designed to be installed in a cabinet or other installation with a DIN Rail, and can easily be connected to additional I/O modules.
Com
Compulab IOT-DIN-IMX8PLUS is a DIN Rail IoT gateway powered by an NXP i.MX 8M Plus AI SoC that takes various expansion modules with digital inputs and outputs (DIO), RS232, RS485, or ADC.
The new model looks to be an evolution of the Compulab IOT-GATE-IMX8PLUS Arm Linux IoT gateway introduced in 2022 with many of the same features, except the IOT-DIN-IMX8PLUS is designed to be installed in a cabinet or other installation with a DIN Rail, and can easily be connected to additional I/O modules.
Compulab IOT-DIN-IMX8PLUS with two DIO modules, one RS485 module, and one ADC module
Operating – Commercial: 0° to 50°C; industrial: -30° to 70°C
Storage – -40° to 85°C
Relative Humidity – 10% to 90% (non-condensing)
MTTF – > 200,000 hours
Compliance
Regulatory – CE, FCC, UKCA
EMC – EN 55032/5, EN 61000-6-2, EN 61000-6-3
Safety – EN/UL/IEC 62368-1
Compulab provides Yocto 4.0 and Debian 12 images with support for Docker, MS Azure IoT, Node-RED, and OTA updates with Mender, as well as Modbus RTU, Modbus TCP, and MQTT libraries. The gateway is compatible with the BalenaOS platform for device management. I could not find documentation specific to the IOT-DIN-IMX8PLUS at the time of writing, but the wiki provides documentation for the Debian 12 and Yocto images for the similar IOT-GATE-IMX8PLUS gateway.
Compulab sells the IOT-DIN-IMX8PLUS for $145 in its minimal configuration for 1K orders, and you’ll find more details and pricing for options and IO modules on the product page. The company also offers a 5-year warranty and a 15-year availability commitment for the gateway.
The Yocto Project 5.0 codenamed “Scarthgap” has just been released with Linux 6.6, glibc 2.39, LLVM 18.1, and over 300 other recipe upgrades. As a result of the release, the developers have made it available for download (bz2 tarball).
The Yocto Project, or Yocto for shorts, is a popular framework used to create custom embedded Linux distributions, and we’ve played with it over the year showing how to create a minimal image for the Raspberry Pi, and last year, we used it again when reviewing two
The Yocto Project 5.0 codenamed “Scarthgap” has just been released with Linux 6.6, glibc 2.39, LLVM 18.1, and over 300 other recipe upgrades. As a result of the release, the developers have made it available for download (bz2 tarball).
The Yocto Project, or Yocto for shorts, is a popular framework used to create custom embedded Linux distributions, and we’ve played with it over the year showing how to create a minimal image for the Raspberry Pi, and last year, we used it again when reviewing two industrial development boards, namely the VOIPAC IMX8M and ADLINK i-Pi SMARC 1200. Yocto is quite a powerful framework/build system with plenty of options that make it highly customizable, but the learning curve is fairly steep.
Some other changes in Yocto Project 5.0 include:
New variables:
CVE_DB_INCR_UPDATE_AGE_THRES: Configure the maximum age of the internal CVE database for incremental update (instead of a full redownload).
RPMBUILD_EXTRA_PARAMS: support extra user-defined fields without crashing the RPM package creation.
OPKG_MAKE_INDEX_EXTRA_PARAMS: support extra parameters for opkg-make-index.
EFI_UKI_PATH, EFI_UKI_DIR: define the location of UKI image in the EFI System partition.
TARGET_DBGSRC_DIR: specifies the target path to debug source files
USERADD_DEPENDS: provides a way to declare dependencies on the users and/or groups created by other recipes, resolving a long-standing build ordering issue
Architecture-specific enhancements:
genericarm64: a new MACHINE to represent a 64-bit General Arm SystemReady platform.
Add Power8 tune to PowerPC architecture.
arch-armv9: remove CRC and SVE tunes, since FEAT_CRC32 is now mandatory and SVE/SVE2 are enabled by default in GCC’s -march=armv9-a.
arm/armv*: add all of the additional Arm tunes in GCC 13.2.0
Some new core recipes:
bmaptool: a tool for creating block maps for files and flashing images, being now under the Yocto Project umbrella.
core-image-initramfs-boot: a minimal initramfs image, containing just udev and init, designed to find the main root filesystem and pivot to it.
lzlib: a data compression library that provides LZMA compression and decompression functions.
lzop: a compression utility based on the LZO library, that was brought back after a (now reverted) removal.
python3-jsonschema-specifications: support files for JSON Schema Specifications (meta-schemas and vocabularies), added as a new dependency of python3-jsonschema.
python3-maturin: a project that allows building and publishing Rust crates as Python packages.
etc…
QEMU has been upgraded to version 8.2.1 and had several improvements
Rust has been upgraded to version 1.75, the Rust profiler options were enabled back, etc…
Several wic Image Creator enhancements
Various SDK-related improvements including:
nativesdk: let MACHINE_FEATURES be set by machine-sdk configuration files.
nativesdk: prevent MACHINE_FEATURES and DISTRO_FEATURES from being backfilled.
Support for riscv64 as an SDK host architecture.
Extend recipes to nativesdk: acpica, libpcap, python3-setuptools-rust.
Several enhancements to testing, utility scripts, BitBake build tool, etc…
AAEON UP Squared Pro 710H Edge is a mini PC with an onboard Hailo-8 edge 26 TOPS AI accelerator and an Intel Processor N97 or Core i3-N305 Alder Lake-N CPU designed for edge AI applications.
The fanless Edge AI mini PC comes up to 16GB LPDDR5 soldered-on RAM, up to 128GB eMMC flash, and offers dual 4K video output through HDMI and DisplayPort connector, dual 2.5GbE, optional WiFi and cellular connectivity, two USB 3.2 ports, an RS232/RS422/RS485 COM port, a 40-pin female GPIO header, and a wide
AAEON UP Squared Pro 710H Edge is a mini PC with an onboard Hailo-8 edge 26 TOPS AI accelerator and an Intel Processor N97 or Core i3-N305 Alder Lake-N CPU designed for edge AI applications.
The fanless Edge AI mini PC comes up to 16GB LPDDR5 soldered-on RAM, up to 128GB eMMC flash, and offers dual 4K video output through HDMI and DisplayPort connector, dual 2.5GbE, optional WiFi and cellular connectivity, two USB 3.2 ports, an RS232/RS422/RS485 COM port, a 40-pin female GPIO header, and a wide 12 to 36V DC input.
Intel Processor N97 quad-core processor up to 3.6 GHz with 6MB cache, 24EU Intel UHD Graphics Gen 12 @ 1.2 GHz; TDP: 12W
Intel Core i3-N305 octa-core processor up to 3.8 GHz with 6MB cache, 32EU Intel UHD Graphics Gen 12 @ 1.25 GHz; TDP: 15W
AI accelerator – Onboard Hailo-8 edge AI processor with up to 26 TOPS for AI inferencing, data analytics, and multimedia processing. CNXSoft: “onboard” implies the chip is soldered on the motherboard instead of on an M.2 module as usual
System Memory – Up to 16GB onboard LPDDR5
Storage – Up to 128GB onboard eMMC
Video Output – HDMI 2.0b and DP 1.2
Networking
2x 2.5GbE RJ45 ports via Intel I226-IT controllers
Certifications – CE/FCC Class A, RoHS Compliant, REACH
The company provides optional VESA, Mount, and DIN Rail mounting with the mini PC, and the UP Squared Pro 710H Edge supports Windows 10 IoT Enterprise, Ubuntu 22.04 LTS, and Yocto 4.0. AAEON previously launched the UP Xtreme 7100 robotics SBC and mini PC with the same choice of processors, but different features and form factor. The 40-pin GPIO header aims to facilitate the installation of cameras, sensors, and industrial communication solutions. Target applications include identity management, AMR (autonomous mobile robots), healthcare imaging solutions, digital signage, and retail kiosks.
AAEON is taking pre-orders for samples of the UP Squared Pro 710H Edge mini PC on its eShop for $568 for the Intel Processor N97 SKU with 8GB/64GB configuration, and $739 for the Intel Core i3-N305 model with 16GB/128GB configuration. The lead time is said to be three weeks. More details may be found on the product page and in the press release.
We’ve already looked at GEEKOM XT12 Pro specifications and hardware design in the first part of the review and followed up by reviewing the Intel Core i9-12900 mini PC with Windows 11 Pro in the second part. We’ve now had time to review the GEEKOM XT12 Pro mini PC with the just-released Ubuntu 24.04 operating system to evaluate the compatibility and performance in Linux.
In this third and final part of the review, we’ll test features in detail, evaluate performance with benchmarks, test storage
In this third and final part of the review, we’ll test features in detail, evaluate performance with benchmarks, test storage and network capabilities, check YouTube video playback at 4K and 8K resolution, perform a stress test to check the cooling solution and provide numbers for fan noise and power consumption of the GEEKOM XT12 Pro mini PC.
Ubuntu 24.04 installation and system information
We’ve resized the Windows 11 by half in order to install Ubuntu 24.04 in dual boot configuration. After that, we inserted a USB drive with the latest version of Ubuntu, and the installation of the Linux distribution went smoothly with no particular issues. The installation procedure has changed slightly, for instance, we have to select between “interactive installation” or “automated installation”, and it’s still mostly straightforward.
We can get a few more details by running a few command in the Linux terminal:
ey@xt12-pro-cnx:~$ cat /etc/lsb-release
DISTRIB_ID=Ubuntu
DISTRIB_RELEASE=24.04
DISTRIB_CODENAME=noble
DISTRIB_DESCRIPTION="Ubuntu 24.04 LTS"
aey@xt12-pro-cnx:~$ uname -a
Linux xt12-pro-cnx 6.8.0-31-generic #31-Ubuntu SMP PREEMPT_DYNAMIC Sat Apr 20 00:40:06 UTC 2024 x86_64 x86_64 x86_64 GNU/Linux
aey@xt12-pro-cnx:~$ free -mh
total used free shared buff/cache available
Mem: 31Gi 3.3Gi 25Gi 905Mi 3.9Gi 27Gi
Swap: 8.0Gi 0B 8.0Gi
Ubuntu 24.04 uses Linux 6.8 as expected and our system comes with 32GB RAM.
We can get additional details with the inxi program:
The utility shows the GEEKOM XT12 Pro mini PC features a 12th Gen Intel Core i9-12900H processor with 14-core (6-mt/8-st) up to 5,000/3,800 MHz and 20 threads, an Intel Alder Lake-P GT2 graphics [Iris Xe Graphics], and 1TB Lexar NM7A1 SSD, The CPU temperature is 52°C at idle. The MediaTek MT7922 wireless module with 802.11ax (WiFi 6) and USB Bluetooth is detected, but we’ll note that there’s no MAC address for the Bluetooth device and it is down. More on that later…
GEEKOM XT12 Pro benchmarks on Ubuntu 24.04
Let’s start with Thomas Kaiser’s sbc-bench.sh script:
aey@xt12-pro-cnx:~/Downloads/sbc-bench-master$ sudo ./sbc-bench.sh -r
Starting to examine hardware/software for review purposes...
sbc-bench v0.9.65
Installing needed tools: distro packages already installed. Done.
Checking cpufreq OPP. Done.
Executing tinymembench. Done.
Executing RAM latency tester. Done.
Executing OpenSSL benchmark. Done.
Executing 7-zip benchmark. Done.
Throttling test: heating up the device, 5 more minutes to wait. Done.
Checking cpufreq OPP again. Done (14 minutes elapsed).
Results validation:
* Measured clockspeed not lower than advertised max CPU clockspeed
* No swapping
* Background activity (%system) OK
* Powercap detected. Details: "sudo powercap-info -p intel-rapl" -> https://tinyurl.com/4jh9nevj
Full results uploaded to https://sprunge.us/WON9T3
# GEEKOM XT12 Pro / i9-12900H
Tested with sbc-bench v0.9.65 on Wed, 01 May 2024 09:32:27 +0700. Full info: [https://sprunge.us/WON9T3](http://sprunge.us/WON9T3)
### General information:
Information courtesy of cpufetch:
Name: 12th Gen Intel(R) Core(TM) i9-12900H
Microarchitecture: Alder Lake
Technology: 10nm
P-cores:
Max Frequency: 5.000 GHz
Cores: 6 cores (12 threads)
AVX: AVX,AVX2
FMA: FMA3
L1i Size: 32KB (192KB Total)
L1d Size: 48KB (288KB Total)
L2 Size: 1.25MB (7.5MB Total)
E-cores:
Max Frequency: 3.800 GHz
Cores: 8 cores
AVX: AVX,AVX2
FMA: FMA3
L1i Size: 64KB (512KB Total)
L1d Size: 32KB (256KB Total)
L2 Size: 2MB (4MB Total)
L3 Size: 24MB
The CPU features 2 clusters of different core types:
i9-12900H, Kernel: x86_64, Userland: amd64
CPU sysfs topology (clusters, cpufreq members, clockspeeds)
cpufreq min max
CPU cluster policy speed speed core type
0 0 0 400 4900 Golden Cove
1 0 1 400 4900 Golden Cove
2 0 2 400 4900 Golden Cove
3 0 3 400 4900 Golden Cove
4 0 4 400 5000 Golden Cove
5 0 5 400 5000 Golden Cove
6 0 6 400 5000 Golden Cove
7 0 7 400 5000 Golden Cove
8 0 8 400 4900 Golden Cove
9 0 9 400 4900 Golden Cove
10 0 10 400 4900 Golden Cove
11 0 11 400 4900 Golden Cove
12 0 12 400 3800 Gracemont
13 0 13 400 3800 Gracemont
14 0 14 400 3800 Gracemont
15 0 15 400 3800 Gracemont
16 0 16 400 3800 Gracemont
17 0 17 400 3800 Gracemont
18 0 18 400 3800 Gracemont
19 0 19 400 3800 Gracemont
31860 KB available RAM
### Policies (performance vs. idle consumption):
Status of performance related policies found below /sys:
/sys/module/pcie_aspm/parameters/policy: default [performance] powersave powersupersave
### Clockspeeds (idle vs. heated up):
Before at 60.0°C:
cpu0-cpu11 (Golden Cove): OPP: 4900, Measured: 4883
cpu12-cpu19 (Gracemont): OPP: 3800, Measured: 3786
After at 83.0°C:
cpu0-cpu11 (Golden Cove): OPP: 4900, Measured: 4883
cpu12-cpu19 (Gracemont): OPP: 3800, Measured: 3785
### Performance baseline
* cpu0 (Golden Cove): memcpy: 22375.8 MB/s, memchr: 37037.2 MB/s, memset: 27398.0 MB/s
* cpu12 (Gracemont): memcpy: 8867.0 MB/s, memchr: 19378.9 MB/s, memset: 14313.0 MB/s
* cpu0 (Golden Cove) 16M latency: 20.53 19.04 20.35 19.14 19.91 19.23 18.27 19.85
* cpu12 (Gracemont) 16M latency: 30.69 28.36 31.01 28.65 30.40 27.44 27.26 30.72
* cpu0 (Golden Cove) 128M latency: 81.88 78.39 82.26 79.50 81.82 75.93 75.87 69.25
* cpu12 (Gracemont) 128M latency: 109.8 104.6 110.4 104.0 110.2 105.5 106.7 118.4
* 7-zip MIPS (3 consecutive runs): 43783, 38345, 38431 (40190 avg), single-threaded: 5502
* `aes-256-cbc 1288693.71k 1534784.87k 1670116.78k 1697173.50k 1655133.53k 1661583.36k (Golden Cove)`
* `aes-256-cbc 898520.42k 1316679.08k 1361087.40k 1372689.41k 1374838.78k 1376299.69k (Gracemont)`
### PCIe and storage devices:
* MEDIATEK MT7922 802.11ax PCI Express Wireless Network Adapter: Speed 5GT/s, Width x1, driver in use: mt7921e, ASPM Disabled
* Intel Ethernet I225-V: Speed 5GT/s, Width x1, driver in use: igc, ASPM Disabled
* O2 SD/MMC Card Reader: Speed 2.5GT/s, Width x1, driver in use: sdhci-pci, ASPM Disabled
* 953.9GB "Lexar SSD NM7A1 1TB" SSD as /dev/nvme0: Speed 16GT/s, Width x4, 0% worn out, unhealthy drive temp: 67°C, ASPM Disabled
* Macronix MX25L25635E 32MB SPI NOR flash, drivers in use: spi-nor/intel-spi
"smartctl -x /dev/nvme0" could be used to get further information about the reported issues.
### Challenging filesystems:
The following partitions are NTFS: nvme0n1p3,nvme0n1p4 -> https://tinyurl.com/mv7wvzct
### Swap configuration:
* /swap.img on /dev/nvme0n1p5: 8.0G (0K used)
### Software versions:
* Ubuntu 24.04 LTS (noble)
* Compiler: /usr/bin/gcc (Ubuntu 13.2.0-23ubuntu4) 13.2.0 / x86_64-linux-gnu
* OpenSSL 3.0.13, built on 30 Jan 2024 (Library: OpenSSL 3.0.13 30 Jan 2024)
### Kernel info:
* `/proc/cmdline: BOOT_IMAGE=/boot/vmlinuz-6.8.0-31-generic root=UUID=f2d0aac4-01dc-4393-ac63-4d6501e5ebb2 ro quiet splash vt.handoff=7`
* Vulnerability Reg file data sampling: Mitigation; Clear Register File
* Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl
* Vulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization
* Vulnerability Spectre v2: Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
* Kernel 6.8.0-31-generic / CONFIG_HZ=1000
Waiting for the device to cool down.............................................. 53.0°C
The maximum temperature (100.0°C) is reached during 7-zip single and multi-core test, at which point the CPU frequency drops as low as 400 MHz. The CPU temperature hovers around 89.0°C during the multi-core cpuminer test. That’s due to modern CPU designs that enable a burst of performance for a short time. This is also confirmed by the 7-zip test where the first run (43783 MIPS) is faster than the subsequent ones (38345 and 38431 MIPS). The utility also complains that the Lexar NM7A1 SSD’s temperature is too high at 67°C.
We can check the power limit as recommended by the script using powercap-info:
PL1 is set to 45W and PL2 to 80W with the Intel Core i9-12900H TDP advertised as 45W. We had 35W and 80W in Windows 11.
We can further test the CPU with Geekbench 6.3.0.
The single-core score was 2,575 points and the multi-core one was 10,447 points. You’ll find the full results on Geekbench website.
We’ll start testing GPU performance in Linux with the Unigine Heaven Benchmark 4.0 program where the GEEKOM XT12 Pro managed to render the scene at an average of 51.3 FPS and got a score of 1,293 points at the standard 1920×1080 resolution.
We then test 4K and 8K YouTube video playback in Chrome.
No issue at 4K 30 FPS with no frames dropped at all while watching the video for over 7 minutes.
8K 30 FPS was good too with only 9 frames dropped out of 13786.
4K 60 FPS is a bit more challenging, but still perfectly watchable with 4k p60 with 329 dropped frames out of 25691.
8K 60 FPS is another story with the system struggling, and the video is not watchable with 7,993 frames dropped out of 22,149, or a 36% drop rate…
We evaluated web browsing performance in the latest version of Firefox using Speedometer 2.0.
The score was 298 runs per minute, one of the best so far, but looking at the details for 10 iterations, there were some variations with the score ranging from 274.5 to 311.3 runs per minute.
GEEKOM XT12 Pro Ubuntu benchmarks comparison against other mini PCs
Let’s compare Ubuntu 24.04 benchmark results for the Intel Core i9-12900H-powered GEEKOM XT12 Pro against other systems (running Ubuntu 22.04) such as GEEKOM Mini IT12 (Intel Core i7-12650H), GEEKOM Mini IT13 (Intel Core i9-13900H), GEEKOM A7 (AMD Ryzen 9 7840HS), and the Khadas Mind Premium (Intel Core i7-1360P).
But first, let’s list the key specifications of the five systems.
GEEKOM XT12 Pro
GEEKOM Mini IT12
GEEKOM Mini IT13
GEEKOM A7
Khadas Mind Premium
SoC
Intel Core i9-12900H
Intel Core i7-12650H
Intel Core i9-13900H
AMD Ryzen 9 7840HS
Intel Core i7-1360P
CPU
14-core/20-thread (6P+8E) processor up to 5.00 GHz (P-cores) and 3.80 GHz (E-Cores)
10-core/16-thread (6P+4E) processor up to 4.70 GHz (P-cores) and 3.5 GHz (E-cores)
14-core/20-thread (6P+8E) processor up to 5.4 GHz (P-Cores) and 4.1 GHz (E-Cores)
8-core/16-thread processor up to 4.0GHz
12-core/16-core (4P+8E) processor up to 5.0 GHz (P-Cores) and 3.7 GHz (E-Cores)
GPU
96EU Intel Iris Xe Graphics
64 EU Intel UHD Graphics
96 EU Intel Iris Xe Graphics
AMD Radeon 780M Graphics
96 EU Intel Iris Xe Graphics
System Memory
32GB DDR4-3200
32GB DDR4-3200
32GB DDR4-3200
32GB DDR5-5600
32GB LPDDR5-5200
Storage
1TB NVMe SSD
1TB NVMe SSD
2TB NVMe SSD
2TB NVMe SSD
1TB NVMe SSD
Default OS
Windows 11 Pro
Windows 11 Pro
Windows 11 Pro
Windows 11 Pro
Windows 11 Home
And now for the benchmark results.
GEEKOM XT12 Pro
GEEKOM Mini IT12
GEEKOM Mini IT13
GEEKOM A7
Khadas Mind Premium
sbc-bench.sh
- memcpy
22,375.8MB/s (P-core)
24,180.5 MB/s
24,014.4 MB/s (P-core)
20,406.0
25,389.5 MB/s (P-core)
- memset
27,398.0MB/s (P-core)
27,539.6 MB/s
26,647.9 MB/s (P-Core)
62,491.7
24,731.8MB/s (P-core)
- 7-zip (average)
40,190
35,730
56,540
71,110
44,430
- 7-zip (top result)
43,783
41,360
60,981
72,496
50,396
- OpenSSL AES-256 16K
1,661,583.36k (P-Core)
1,630,005.93k (P-Core)
1,844,401.49k (P-Core)
1428559.19k
1,771,334.31k (P-Core)
Geekbench 6 Single
2,575
2,575
2,745
2,535
2,093
Geekbench 6 Multi
10,447
9,874
11,974
12,914
8,891
Unigine Heaven score
1,293
1,002
1,333
2,032
1,349
Speedometer (Firefox)
298
302
273
249
242
The GEEKOM XT12 Pro delivers slightly better performance than the GEEKOM Mini IT12 in most tests, including 3D graphics, but not overly so despite featuring a 14-core Intel Core i9 processor. The performance is not that different from the Khadas Mind Premium either with a 13th Gen Core i7-1360p CPU, but the GEEKOM Mini IT13 (Intel Core i9-13900H) and especially the GEEKOM A7 (AMD Ryzen 9 7840HS) are a step ahead from it comes to performance. But somehow both the Mini IT12 and XT12 Pro outperform their peers in Speedometer 2.0 web browser test.
Storage and USB performance tests
We tested the 1TB NVMe SSD performance with iozone3:
That’s pretty good with about 4,315 MB/s sequential read speed and 4,323 MB/s sequential write speed. By comparison, CrystalDiskMark utility in Windows reported 5,096 MB/s and 4,438 MB/s respectively for similar tests.
We don’t have a spare M.2 2242 SATA SSD, so we did not test the M.2 SATA socket in the GEEKOM XT12 Pro.
We’ll now test the USB4, USB 3.2, and USB 2.0 ports on GEEKOM XT12 Pro using an ORICO M234C3-U4 M.2 NVMe SSD enclosure for USB 3.x/4 ports and a USB 3.0 hard drive for the USB 2.0 port, as well as the iozone3 and lsusb or boltclt to confirm the transfer speed.
Here’s an example with the left USB 3.2 Type-A on the front panel:
aey@xt12-pro-cnx:~$ lsusb -t | grep uas
|__ Port 002: Dev 002, If 0, Class=Mass Storage, Driver=uas, 10000M
aey@xt12-pro-cnx:/media/aey/EXT4-REVIEW$ sudo iozone -e -I -a -s 1000M -r 16384k -i 0 -i 1
random random bkwd record stride
kB reclen write rewrite read reread read write read rewrite read fwrite frewrite fread freread
1024000 16384 963719 970302 870059 873834
iozone test complete.
USB4 testing relies on boltctl. Here’s the output for the left USB4 port on the rear panel:
We’ll use iperf3 to test both 2.5GbE and WiFi 6 networking starting with 2.5GbE using an UP Xtreme i11 Edge mini PC running Ubuntu 22.04 on the other side.
We did not manage to find a solution, so if you need Bluetooth in Linux, devices with a MediaTek MT7922 should be avoided until a solution is found. In a general sense, Intel wireless modules usually work fine in Linux, but MediaTek ones may be hit or miss. For example, I have a laptop with an MT7902 wireless module and there aren’t any Linux drivers for it, so I’m using a 2.5GbE USB dongle or USB tethering to my phone when Ethernet is not available…
Checking thermal performance with a stress test
We ran a stress test on all 20 threads of the Intel Core i9-12900H processor and monitored the package temperature and CPU frequency with psensor and sbc-bench.sh to check for CPU throttling.
The Psensor chart shows the package temperature jumps quickly to 97°C in the first 30 seconds before stabilizing to around 89-91°C in the long run. The P-cores max out at around 3,100 MHz, before dropping to around 2,600 MHz, and E-cores started at 2,400 MHz before running at 2,200 MHz.
Fan noise
GEEKOM XT12 Pro’s fan is barely audible at idle or under light loads but becomes noisier under heavier loads. As usual, we measured the fan noise with a sound level meter placed at around 5 centimeters from the top of the enclosure:
Idle – 37.8 – 41.3 dBA
Stress test on all 20 threads – 49.5 – 54.7 dBA
For reference, the meter measures around 37 to 38 dBA in a quiet room.
GEEKOM XT12 Pro power consumption on Ubuntu 24.04
We measured the power consumption with a wall power meter:
Power off – 1.4 Watts
Idle – 6.1 – 7.2 Watts
Video playback – 56.3 – 63.3 Watts (Youtube 8K 60fps in Firefox)
CPU stress test (stress -c 20)
First couple of seconds – 68.0 – 70.9 Watts
Longer run – 51.1 – 51.9 Watts
During the measurements, the mini PC was connected to WiFi 6, a USB RF dongle was used for a wireless mouse and keyboard combo, and a 15.6-inch CrowVi Full HD display was connected via HDMI using its own power source.
Conclusion
The GEEKOM XT12 Pro works well and fast in Ubuntu 24.04 thanks to its powerful Intel Core i9-12900H 14-core/20-thread processor, 32GB of RAM, and fast M.2 NVMe SSD storage. Its six USB ports (USB 2.0, USB 3.2, and USB4 ports) provide plenty of expansion, and 2.5GbE and WiFi 6 networking works great. The main downside is that Bluetooth 5.2 does not work at all due to driver issues.
YouTube video playback works well up to 4Kp60 and 8Kp30, but 8K 60 FPS was too much to ask in Ubuntu with around 36% of frames being dropped during our test, although it worked just fine in Windows 11 Pro. It does provide some small boost of performance compared to the Mini IT12, but somehow the latter had no issue playing an 8K 60 FPS in Ubuntu 22.04 when we tried, many because of the slightly higher room temperature (TBC). As with all other compact mini PCs, the CPU does get hot underload (stabilizes at 90°C) and the fan is barely audible under light load, and the noise may only become an issue for some people under heavy loads.
We’d like to thank GEEKOM for sending the XT12 Pro mini PC with an Intel Core i9-12900H processor, 32GB RAM, and a 1 TB NVMe SSD for review. It can be purchased for $664 on Amazon or GEEKOM US after applying the coupon code cnxXT12Pro for a 5% discount valid until May 31, 2024. Readers based in the UK can also use that coupon on GEEKOM UK.
Pineboards, previously known as PineBerry, has launched four new Raspberry Pi HAT+ with a PCIe interface: the Hat uPCIty Lite, HatDrive! Piano, Hat mPCIe, and Hat Ai! Dual.
The Polish company decided to change the name from Pineberry to Pineboards since the “berry” name implied they were manufacturing single-board computers, while Pineboards, not to be confused with Pine64 boards, apparently does not :). Nevertheless, let’s have a look at the four new HAT+ boards.
Hat uPCIty Lite – PCIe x4 slot
Pineboards, previously known as PineBerry, has launched four new Raspberry Pi HAT+ with a PCIe interface: the Hat uPCIty Lite, HatDrive! Piano, Hat mPCIe, and Hat Ai! Dual.
The Polish company decided to change the name from Pineberry to Pineboards since the “berry” name implied they were manufacturing single-board computers, while Pineboards, not to be confused with Pine64 boards, apparently does not :). Nevertheless, let’s have a look at the four new HAT+ boards.
Hat uPCIty Lite – PCIe x4 slot and ATX power supply
PCIe X4 slot (output) to connect PCIe cards such as NVIDIA or AMD graphics cards; Note: only PCIe x1 Gen2 and Gen3 are supported
Misc – 12V and PWR LEDs
Power Supply
12V/8A via ATX power connector
12V/8A via power barrel jack
Safety – PCIe Express power delivery is isolated from the Raspberry Pi 5
Dimensions – 65 x 56.50 mm; HAT+ compatible, 4-layer PCB
That Hat ships with an FPC PCIe ribbon cable, spacer pins, and M2.5 screws, so you only need to add your Raspberry Pi 5 SBC and a PCIe card (graphics, networking., etc…) of your choice to get started.
HatDrive! Piano – NVMe storage meets audio
Specifications:
Supported SBCs – Raspberry Pi 5 and other compatible SBCs
Host interfaces
16-pin PCIe FPC connector
40-Pin Raspberry Pi GPIO header
Storage – M.2 Key-M 2230/2242 socket for NVMe SSD (PCIe x1 Gen2 or Gen3 only)
Audio
3.5mm headphone jack (which was removed from the Pi 5, but present in all previous versions) with headphone amplifier
RCA Line Out
112dB Audio Stereo DAC from Texas Instruments with 32-bit, 384kHz PCM Interface
Expansion – 40-Pin Raspberry Pi HAT
Misc
PWR and ACT (M.2 activity) LEDs
Dedicated I2C EEPROM memory with HAT ID, allowing for advanced identification and configuration.
Power Management
Continuous monitoring and diagnostics of the power supply bus, measuring voltage, current, and power parameters in real-time via an I2C interface.
Integrated voltage regulator delivering up to 3A for the 3.3V power rail, compliant with M.2 (NGFF) standard.
M.2 card power options – FPC PCIe ribbon cable or 40-pin GPIO connector.
Dimensions – 65 x 56.50 mm; HAT+ compatible, 4-layer PCB
The board ships with an FPC PCIe ribbon cable, spacer pins, M2.5 screws, a spacer pin for M.2 card mounting, and a 40-pin male-female HAT connector.
Hat mPCIe – Cellular and WiFi connectivity via PCIe or USB
Specifications:
Supported SBCs – Raspberry Pi 5 and other compatible SBCs
Host interfaces
16-pin PCIe FPC connector
USB 2.0 Type-C connector to use the mini PCIe module with another host
mPCIe socket with Nano SIM card for WiFi and cellular modules; supports PCIe x1 Gen2 or Gen3 only
Power Supply
Integrated voltage regulator capable of delivering up to 3A for the 3.3V power rail
Input via 16-pin PCIe FPC connector or USB-C port
Dimensions – 65 x 56.50 mm; HAT+ compatible, 4-layer PCB
The HAT+ board ships with an FPC PCIe ribbon cable, spacer pins, M2.5 screws, and standoffs to connect a mPCIe card.
Hat AI! Dual – Google Coral Dual Edge TPU for the Raspberry Pi 5
Specifications:
Supported SBCs – Raspberry Pi 5 and other compatible SBCs
Integrated ASM1182E PCIe packet switch to share the single PCIe lane between two TPUs. Note: Pineboards claims the processing performance is still up to twice the standard TPU module since most data transfers occur when your application is loading the TensorFlow models.
Misc – LED for monitoring the power input of the board
Power Management – Integrated voltage regulator capable of delivering up to 3A for the 3.3V power rail, adhering to the M.2 (NGFF) standard.
Dimensions – 65 x 56.50 mm; HAT+ compatible, 4-layer PCB
The board ships with a FPC PCIe ribbon cable, spacer pins to mount it to the Raspberry Pi SBC, M2.5 screws, and a spacer pin for mounting the M.2 card.
Pricing
All four HAT+ boards are available for the following prices:
SpacemiT, a chip design company from China with RISC-V as its core technology, recently unveiled the Muse Book laptop based on the K1 octa-core RISC-V chip. Unlike our daily laptops, it has many interesting unique features and is mainly sold to hardware engineers and DIY enthusiasts.
This Muse Book runs the Bianbu OS operating system based on the Debian distribution and optimized to run on the SpacemiT K1 octa-core RISC-V SoC. Let’s first take a look at its external interfaces. On the left side
SpacemiT, a chip design company from China with RISC-V as its core technology, recently unveiled the Muse Book laptop based on the K1 octa-core RISC-V chip. Unlike our daily laptops, it has many interesting unique features and is mainly sold to hardware engineers and DIY enthusiasts.
This Muse Book runs the Bianbu OS operating system based on the Debian distribution and optimized to run on the SpacemiT K1 octa-core RISC-V SoC. Let’s first take a look at its external interfaces. On the left side of the laptop, there are two USB Type-C interfaces, a USB 3.0 Type-A port, a 3.5mm headphone jack, a microSD card slot, and a reset pinhole.
The 8-pin header on the right side of the laptop is quite interesting, and SpacemiT hopes the Muse Book can become one of the most convenient hardware development platforms for RISC-V. In addition to the power pins, users will find multiplexed I2C, UART, PWM, GPIO, and other interfaces. There’s also another USB Type-A port on that side and an additional pinhole to enter USB download mode for flashing the firmware.
Muse Book IO specifications:
Storage – MicroSD card slot
USB
1x USB Type-C port (USB 3.2 Gen 1 data transmission, PD fast charging)
1x USB Type-C port (USB 3.2 Gen 1 data transmission, PD fast charging, DP video signal)
2x USB 3.0 Type-A ports
Audio – 3.5mm headphone jack
Misc
Reset pinhole
USB start pinhole to set the USB interface to download mode
The laptop features a 14.1-inch IPS display with a resolution of 1920×1080 and a 60 Hz refresh rate, a relatively conventional configuration. It also comes with a basic touchpad and keyboard whose Windows button has been replaced by a RISC-V button.
The notebook adopts a relatively lightweight design, with a weight of 1.3kg and a thickness of approximately 18mm. A teardown of the Muse Book reveals an all-metal fanless design, indicating that SpacemiT is very confident in the heat dissipation control of the K1 SoC.
The main board features the K1 SoC, up to 16GB of LPDDR4X memory, eMMC storage, and a WiFi 6 wireless module (Fn-Link 6252C-PUB). A 36.48Wh/7.6V battery was selected for the laptop with support for USB PD 3.1 fast charging and a battery life of over 6 hours in typical conditions.
Muse Book specifications:
SoC – SpacemiT K1 octa-core X60 RISC-V core compliant with RVA22 (See SoC specifications details below)
System memory – Up to 16GB 32bit LPDDR4-2400MT, 32bit LPDDR4X-2666MT
Storage
32GB eMMC flash (default), option up to 128GB
M.2 (PCIe 2.1 2x) socket for NVMe SSD up to 1TB
MicroSD card slot
Display
14.1-inch IPS display with 1920×1080 resolution, 60Hz refresh rate
72% NTSC (≈100%sRGB) Color gamut
250 nits brightness
Networking – RTL8852BE-based WiFi 6 module
Power Supply – USB PD 3.1 Type-C port (65W power adapter)
Dimensions – 322.6 x 209.2 x 17.8mm
Weight – 1.36kg
OS: Bianbu OS by Debian, Ubuntu, Linux
Browser: Chromium
The SpacemiT K1, the main controller of the laptop, is the world’s first RISC-V SoC that complies with the RISC-V Foundation’s RVA22 standard and the 256-bit RVV 1.0 standard. It provides 2.0 TOPS of AI computing power.
SpacemiT K1 SoC specifications:
CPU – Octa-Core X60 64-bit RISC-V core (faster than Cortex-A55 in multi-core configuration)
GPU – Unnamed with support for OpenCL 3.0, OpenGL ES3.2, Vulkan 1.2
VPU – H.265/H.264/VP9/VP8 4K encoding/encoding
AI accelerator – 2.0 Tops AI NPU
Memory: Up to 16GB 32bit LPDDR4-2400MT, 32bit LPDDR4X-2666MT, Up to 10.6GB/S bandwidth
Temperature Range – -40°C to 85°C (industrial grade)
OS
Bianbu OS based on Debian
Mainline Linux
RTOS
Unfortunately, the laptop I have on hand cannot be turned on, so I am unable to conduct further evaluation. However, from the information released by SpacemiT, we can understand that in addition to commonly used office software such as Office and Chromium browsers, Muse Book also comes with optimized OpenCV, OpenBLAS, Slam Eigen, libpng, libjpeg, XNNPACK, and other algorithm libraries allowing developers to access high-performance computing libraries. In addition, foundation models (like ChatGPT) can be used to develop and deploy without having too much knowledge of the chip itself.
AI on K1 RISC-V chip
SpacemiT K1 integrates a wide range of I/O interfaces, including multiple PCIe, USB, and GMAC interfaces, and can operate in the -40°C to 85°C industrial temperature range. In addition to the Muse Book laptop, SpacemiT has plans for other hardware products as part of the Muse series, including the MUSE Card system-on-module, the MUSE Pi development board, the MUSE box mini PC, and the MUSE Shelf server/workstation.
It is reported that this laptop will be launched for sale soon with an expected price of around $300. There is currently no news of other MUSE series products being launched. You can find more information on the official SpacemiT community website (in Chinese only). The Muse Book is not the first RISC-V laptop on the market (or announced), as we’ve already covered the ROMA laptop and Lichee Console 4A portable development terminal both based on T-Head TH1520 quad-core RISC-V SoC, as well as an laptop based on the upcoming StarFive JH8100 that may or may not even see the light of the day.
CAPUF Embedded CH32V003 Dev Kit is an all-in-one development board with a USB-C interface, onboard sensors (temperature/humidity), an OLED display, SPI NOR Flash, and ample I/O options. Additionally, it features a Qwicc connector, an RGB LED, and a 3-pin header to connect the WCH-link programmer providing further flexibility for your projects.
After the initial release of this “10 cents” CH32V003 RISC-V microcontroller, we have seen $1.5 development boards and an open-source GGC toolchain, as we
CAPUF Embedded CH32V003 Dev Kit is an all-in-one development board with a USB-C interface, onboard sensors (temperature/humidity), an OLED display, SPI NOR Flash, and ample I/O options. Additionally, it features a Qwicc connector, an RGB LED, and a 3-pin header to connect the WCH-link programmer providing further flexibility for your projects.
4 Pin Header is provided for connecting other boards via jumper wires
SPI – 8Mbit NOR Flash connected via SPI
Indicators – LED Indications for 5V/3.3V/UART RX, TX
Misc
Option for connecting external 24MHz Oscillator
Power LED, RGB LED
reset buttons
potentiometer and buzzer
Board Dimensions – Not Specified
Inside the box, you will find a CH32V003 Dev Kit Development Board, a USB-A to USB-C cable, and a WCHLinkE programmer. Additionally, the company will provide resources such as a PDF schematic and a Dev Kit 3D model, which are currently unavailable at the time of writing. The company also provides a getting started guide with real-world examples on its website and code samples can be found on their GitHub repository.
The product is not yet available for purchase. However, the product page indicates that it will be released soon. Back in December 2023, CAPUF Embedded told CNX Software the board would be available soon for approximately 35-45 USD with free shipping worldwide.
SBC Case Builder V3.0 case design utility has just been released with the ability to create over 1,000 standard cases – not including customization – for popular SBCs from Raspberry Pi, Hardkernel, Orange Pi, Radxa, and others, as well as standard motherboards following Mini-ITX, Pico-ITX, NUC, Nano-ITX, etc.., and SBC adapters following these standards, meaning you could install a Raspberry Pi 5 into a mini-ITX case if needed.
SBC Case Builder started as a command line utility for designing DIY
SBC Case Builder V3.0 case design utility has just been released with the ability to create over 1,000 standard cases – not including customization – for popular SBCs from Raspberry Pi, Hardkernel, Orange Pi, Radxa, and others, as well as standard motherboards following Mini-ITX, Pico-ITX, NUC, Nano-ITX, etc.., and SBC adapters following these standards, meaning you could install a Raspberry Pi 5 into a mini-ITX case if needed.
Version 3.0 comes with many improvements but the main focus was to reuse existing and new PC standard form factor cases by creating SBC adapters and custom I/O Shields since Edward noted many users lamented the lack of enclosure or a poor choice of cases when SBCs were announced on websites such as CNX Software and others.
Another change is the use of the SBC Model Framework version 2, released in February of 2024, which incorporates openings for the SBC models at a component level in the library. Edward explained it to us as follows:
What this means is that anyone using this library for design has access to autonomous openings for the SBC model, including select-able and dynamic Heatsink, GPIO and UART mask openings. This is best illustrated by viewing the animated graphic at the git repository README.md or the Hardkernel forum thread. The included SBC Model Framework Viewer allows one to review models in the library, their related masks and technical meta data.
ODROID-M1S and UPS Kit housed in a standard mini-STX enclosure through a custom IO shield generated in SBC Case Builder V3.0
99 devices are now supported including 70 SBCs, 3 carrier boards, 8 compute modules, 4 microcontroller boards, and 14 standard motherboards:
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
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 on the bottom allows for easy mounting and installation on a DIN rail.
M5Stack CoreMP135 specifications:
MPU – STMicroelectronics STM32MP135DAE7 microprocessor, featuring single-core Arm Cortex-A7 CPU @ 1GHz
Memory – 4Gbit (512MB) DDR3L SDRAM
Storage – microSD card slot
Display
2.0-inch IPS capacitive touchscreen
1x HD video output (LT8618SXB HDMI 1.4 transmitter, supports up to 24 bits of color depth). Note: M5Stack does not use “HDMI” wording because it requires an extra license.
Ethernet – 2x Gigabit Ethernet
USB – 2x USB 2.0 Type-A, 1x USB-C (OTG, Power)
Expansion
2x Grove (I2C and UART) interfaces
PWR485 expansion – 9~24V power input and RS485 interface
Power Supply
12V @ 2A via DC power socket, 5V@3A via USB-C
AXP2101 PMIC
Misc – 1W speaker (16-bit, I2S), DIN rail base plate
Operating Temperature – 0°C – 40°C
Dimensions – 54 x 54 x 34.5mm
Weight – 99.5 grams
The documentation page contains more information if you want to learn more about the product. M5Stack has released similar devices to the CoreMP135 in the past, often based on ESP32 microcontrollers, including the Core2 devkit, Core3, and AtomS3 Lite, and Espressif Systems has very recently acquired a majority stake in the company. The CoreMP135, however, differs in its ability to run a full Linux operating system.
The CoreMP135 is mostly targeted at industrial automation, IoT edge gateways, Smart Home, multimedia entertainment devices, and robot motion control. It is currently available for $75 on Aliexpress or M5Stack’s store.
SONOFF has been gradually refreshing its Zigbee product lineup since August 2023, and CNX has recently written about the Temperature/Humidity Sensor (SNZB-01P), Switch/Button (SNZB-02P) and Motion Sensor (SNZB-03P), and the SONOFF SNZB-06P human presence sensor. I’ve now had the opportunity to review both the SONOFF SNZB-06P and the new Door/Window Sensor (SNZB-04P). After we keep trying them out for a few weeks, we think reliability is much improved, especially with the Door/Window Sensor, the
SONOFF has been gradually refreshing its Zigbee product lineup since August 2023, and CNX has recently written about the Temperature/Humidity Sensor (SNZB-01P), Switch/Button (SNZB-02P) and Motion Sensor (SNZB-03P), and the SONOFF SNZB-06P human presence sensor. I’ve now had the opportunity to review both the SONOFF SNZB-06P and the new Door/Window Sensor (SNZB-04P). After we keep trying them out for a few weeks, we think reliability is much improved, especially with the Door/Window Sensor, the materials made as well as the appearance look better than the previous models. As for the Human Presence Sensor, it’s the first generation, which is quite impressive considering its price and features. Let’s take a closer look.
Door/Window Sensor (SNZB-04P) review
Unboxing
Obviously, the rounded design differs from the previous rectangular version. It integrates better with the surroundings than the previous model. Additionally, the lithium battery capacity is larger (CR2450 @550mA vs. CR2477 @1000mA), resulting in a longer lifespan. SONOFF claims up to 5 years, but we know it really depends on usage.
The large white base on the left side houses the battery and circuit board, while the right side contains a regular magnet. This is a sort of reed switch which uses a magnet to open and close the switch circuit. A problem with the previous Door/Window Sensor (SNZB-04) was that the magnet was relatively weak and couldn’t detect if installed too far apart or at different levels. Sometimes, the installation location might not allow for close proximity. It seems that the SNZB-04P model addresses this issue to allow for more flexible installation. The image below shows the maximum distance between the two sides that can still be detected, as well as the level of discrepancy compared to the previous model on the right which indicates the furthest distance the sensor could detect.
Therefore, anyone experiencing installation issues with the Door/Window Sensor in challenging locations might appreciate the solution. It eliminates the need to purchase stronger magnets separately to address excessive spacing. Overall, it makes installation more flexible, a definite advantage of this new sensor.
Another addition to this new model is the Tamper Detection feature. Essentially, there’s a tiny black button at the back of the sensor (above the SONOFF letters). When installed, this button is normally pressed. However, if someone removes the sensor, the button is no longer pressed, allowing the Smart Home app to detect the event and send notifications if set.
Specifications
Wireless connection – Zigbee 3.0
Installation distance – ≤20mm
Battery – CR2477(3V)
Dimensions
Transmitter – 50.5x32x21.9mm
Magnet – 27x12x12.4mm
Weight – 33 grams
Casing material – PC
Temperature Range – -10°C to 60°C
Humidity – 5-95% RH, non-condensing
Testing SNZB-04P with eWeLink mobile application
First, you need to register an account on eWeLink. Since this is a Zigbee device, you’ll also need a SONOFF Zigbee Hub/Bridge to act as a gateway to connect to the eWeLink Cloud via Wi-Fi. In our test, we used the SONOFF Zigbee Bridge Pro. Once we made SNZB-04P in paring mode, press the “+” button on the top right in eWeLink to add the SNZB-04P. The process is similar to adding other SONOFF devices.
As the Door/Window Sensor is a very basic device with a few options to tweak, we’ll focus only on the new features to streamline the review. We found that eWeLink has added some features for the SNZB-04P. It introduces new options for setting triggers in scenes. In the previous model, we could only set trigger conditions in scenes when the door/window is closed or opened. However, the new model offers more flexibility, such as “Remain closed for a duration” and “Remain opened for a duration.” in addition to “Closed” and “Opened”. This allows for more flexible actions, such as triggering an action if the door/window remains open for more than 60 seconds. Refer to the below image when creating a scene with new trigger conditions.
Testing SNZB-04P with Home Assistant
We’ve tested the SNZB-04P with Home Assistant 2024.4.0 in combination with the SONOFF Zigbee USB Dongle “P” version used both the ZHA integration and the Zigbee2MQTT integration. The test results indicate that Zigbee2MQTT provides more accurate and detailed information compared to ZHA. For example, Zigbee2MQTT includes sensor states for normal status, tamper detection, and low battery status, while ZHA lacks thos. It is also worth mentioning that certain features seen in eWeLink, such as Duration Setting (how long before reporting a status change after undetected), are not available in ZHA and Zigbee2MQTT. However, Home Assistant fanboys may not find it’s not a big issue because they can easily create something similar to Duration Setting themselves within Automations.
Finally, SONOFF has jumped into the market of mmWave (5.8GHz) sensors with the SNZB-06P model, which utilizes Zigbee 3.0 protocol at an attractive price point. The basic principle of operation for the SNZB-06P is to emit Frequency Modulated Continuous Wave (FMCW) radar signals and measure the time it takes for the reflected signal to come back. Then, it calculates the distance and movement of the detected object using software algorithms. SONOFF claims it can detect motion at a level of 1mm and a minimum speed of 3mm/s, which covers even human breathing and slight movements. This makes it ideal for detecting stationary objects within its range. For instance, it could be used to detect sleep, occupants in a bathroom, or individuals sitting quietly during meetings or work. As mentioned earlier, it constantly emits radar signals for detection, so devices using mmWave technology like the SNZB-06P need to be powered continuously. The SNZB-06P comes with a USB-C port for a 5V power supply, and it can also act as a Zigbee Router to expand the Zigbee network in the home.
The mechanical design looks similar to the SNZB-03P (Motion Sensor) we recently reviewed. However, the primary difference lies in their intended use. The Motion Sensor is more suitable for areas with frequent motion, while the Human Presence Sensor is better suited for detecting stationary objects. See the comparison between the two models in the image below.
Unboxing
Inside the box, you’ll find everything necessary for use, as usual including a mini manual, the device itself, a magnetic base, a USB-C cable, mounting screws, and double-sided tape for attaching the magnetic base. The device has a pairing button and a QR code for installation. The SNZB-06P should be used with the SONOFF Zigbee Bridge Pro to fully make use of its capabilities. If used with a regular SONOFF Zigbee Bridge, SNZB-06P will function only as a basic motion sensor.
We received the SONOFF SNZB-06P some time ago but encountered issues with unstable hardware, necessitating a replacement and delaying the review more than anticipated. However, this turned out to be a benefit to readers, as early adopters encountered installation and usage issues and provided feedback to SONOFF. In response, SONOFF quickly released a new firmware (1.0.6) and additional documentation to address these issues. This documentation, called the Pre-Use Instructions, discusses various factors affecting performance and solutions, such as installation location, the impact of wind, heat, pets, vacuum robots, walls, mirrored surfaces, and the distance between SNZB-06P units. These are valuable recommendations. See the below image for quick reference.
In our testing, we found that the optimal mounting height for accurate object detection is approximately 1.5m to 1.8m, as shown in the photo below. If there are vacuum robots, ensure that the detection level is higher than their height. We found this to be a common issue with mmWave sensors, as they can still detect moving fans, robots, or trees. Therefore, finding a suitable mounting location is crucial for optimal performance.
Sensitivity configurable from 2.5 meters to 4 meters
Wireless – Zigbee 3.0 (router type) up to 30 meters
Misc – Red LED, magnetic base, button
Power Supply – 5V DC/1A via USB Type-C port
Dimensions – 44.2 x 44.2 x 59mm
Weight – 31 grams
Temperature Range – -10°C to 60°C
Humidity – 0 to 90%RH, non-condensing
Materials – ABS+PC
Certifications – CE/FCC/ISED/RoHS
Safety standard – EN IEC 62368-1
Using the SONOFF SNZB-06P with eWeLink mobile application
As usual, we added the SNZB-06P from SONOFF Zigbee Bridge Pro by pressing the add device button. Before adding the device, we pressed and held the single button on the SNZB-06P until the light flashed, indicating it was ready for pairing. See the below image for reference.
After the SNZB-06P is added to eWeLink, it’s now ready to use. However, two important settings can be adjusted to maximize the performance of the SONOFF SNZB-06P depending on the environment:
Sensitivity: This setting determines the detection range. Setting it to “Low” means the sensor can detect up to 2.5 meters away, “Medium” up to 3.5 meters, and “High” up to 4 meters. Starting with the lowest sensitivity setting (Low) is recommended as it reduces the impact of disturbances.
Detection Duration: This refers to the time duration after detection where the sensor will update its status to “No Present” and be ready to detect again. This acts as a cooldown period to prevent frequent triggering of actions such as lighting, alarms, or notifications. The minimum duration is 15 seconds, which is quite fast in our opinion.
We tested the SNZB-06P’s capabilities by creating a Scene (see screenshots below for reference). We found that SNZB-06P comes with a built-in light sensor, which can be used to create more diverse conditions in a Scene. For example, we set a trigger that if a person is detected and there is no light, it sends a notification.
Furthermore, we tried to reduce the Detection Duration to the minimum of 15 seconds to see how it responds. It worked as expected. If any users find discrepancies in this functionality, make sure you follow the Pre-Use instructions as mentioned above, as it may indicate a hardware issue that needs to be addressed with the seller.
Review of SONOFF SNZB-06 with Home Assistant
We tested the integration with Home Assistant using both ZHA and Zigbee2MQTT which are popular Zigbee integrations. The good news is that there’s more detail available compared to the earlier firmware versions we tested months ago (version 1.0.6). This version added options like Detection Duration and Sensitivity settings on the main page, making it easy to configure without having to go down into Zigbee clusters like the first firmware version. We also found that the light sensor statuses “Dark” and “Bright” are separate entities providing even more flexibility compared to eWelink’s native app. FYI these illumination values only update when the Occupancy entity status is updated.
Functionality remains consistent across ZHA and Zigbee2MQTT, with fast responsiveness and updates to the “No Present” state following the configured settings. Additionally, the entities obtained are similar in both integration, even on par with using the eWelink software.
The SNZB-04P has seen significant improvements over previous models from SONOFF, such as the SNZB-04 (without the “P” suffix) and the DW2 WIFI. It features a larger battery, stronger magnet, better energy efficiency (compared to the WiFi version), and faster responsiveness. It can also work with other platforms such as Google Home, Alexa, and SmartThings via the Zigbee 3.0 protocol. We love the strong magnetic at most as it provides more flexibility for installation.
Meanwhile, the SNZB-06P is the first mmWave sensor released by SONOFF, and our review showed its performance is commendable considering its relatively affordable price compared to other products on the market. It can also function as a Zigbee Router to build a reliable Zigbee network in the house. As mentioned in the comparison table together with motion sensors, it’s suitable for comfortably detecting stationary humans, such as in meeting rooms, offices, or bathrooms. However, like most mmWave sensors on the market, it requires constant power input via USB due to its energy-intensive nature in emitting waves to detect objects, which may make the installation tricker since the user needs to find a wall socket nearby. Pick the right one to fit your needs.
We’d like to thank ITEAD for sending the SONOFF SNZB-06P human presence Zigbee sensor and SNZB-04P door/window Zigbee sensor for review. The SNZB-06P can be purchased for $14.90, while the SNZB-04P goes for $10.90. As usual, you can get a 10% discount on any order on the ITEAD store with the coupon code CNXSOFTSONOFF, and standard shipping is free for orders over $89.
The Raspberry Pi 5 Arm SBC is now powerful enough to challenge some Intel systems in terms of performance, while Intel has made the Intel Alder Lake-N family, notably the Intel Processor N100, inexpensive and efficient enough to challenge Arm systems when it comes to price, form factor, and power consumption.
So we’ll try to match the Raspberry Pi 5 to typical Intel processor N100 mini PCs with a comparison of features/specifications, performance (benchmarks), and pricing with different use case
The Raspberry Pi 5 Arm SBC is now powerful enough to challenge some Intel systems in terms of performance, while Intel has made the Intel Alder Lake-N family, notably the Intel Processor N100, inexpensive and efficient enough to challenge Arm systems when it comes to price, form factor, and power consumption.
So we’ll try to match the Raspberry Pi 5 to typical Intel processor N100 mini PCs with a comparison of features/specifications, performance (benchmarks), and pricing with different use cases. That’s something I’ve been wanting to look into for a while but I was busy with reviews and other obligations (Hello, Mr. Taxman!), and this weekend I had some spare time to carry on the comparison.
Raspberry Pi 5 vs Intel N100 mini PC specifications
I’ll start by comparing the specifications of a Raspberry Pi 5 against the ones for typical Intel Processor N100-based mini PCs also mentioning optional features that come at extra cost.
VideoCore VII GPU @ 800 MHz with support for OpenGL ES 3.1, Vulkan 1.2
24EU Intel HD graphics @ 750 MHz with support for DirectX 12.1, OpenGL 4.6, OpenCL 3.0
- Video Decode on GPU
4Kp60 HEVC decoder
4Kp60 AVC (H.264), HEVC (H.265), VP9, JPEG, AV1
- Video Encoder on GPU
N/A
4Kp30 AVC, HEVC (H.265), VP9, JPEG
System Memory
4GB or 8GB LPDDR4X-4267
8GB to 32GB DDR4/DDR5/LPDDR5
Storage
MicroSD card
M.2 NVMe or SATA via HAT board
2.5-inch/3.5-inch SATA drive via HAT board
MicroSD or SD card slot
Included M.2 NVMe or SATA drive
2.5-inch SATA drive slot on some systems
Video Outputs / Display Interfaces
2x micro HDMI ports up to 4Kp60
2x 4-lane MIPI DSI display connectors
Up to 3 displays supported via HDMI, DisplayPort, and/or USB-C up to 4Kp60
Audio
Digital audio via HDMI
Various HATs for additional audio ports
Digital audio via HDMI
3.5mm audio jack (on most systems)
Camera
2x 4-lane MIPI CSI camera connectors (multiplexed with MIPi DSI)
Usually only through USB cameras
Ethernet
Gigabit Ethernet
Gigabit Ethernet, 2.5GbE, or multiple Ethernet interfaces depending on system
Wireless
WiFi 5 and Bluetooth 5.0
WiFi 5/6/6E and Bluetooth 4.2 to 5.4 depending on M.2 wireless module selected
USB
2x USB 3.0 ports (5Gbps)
2x USB 2.0 ports
Depends on system, up to 5x USB ports, with 10 Gbps USB 3.2, USB 2.0, and USB-C with DisplayPort Alt mode ports supported
GPIO
40-pin Raspberry Pi GPIO header
Usually none, only available through USB to GPIO adapters
PCIe
Custom connector for PCIe Gen 2 x1 (official), or PCIe Gen3 x1 (unofficial, and may not work on all boards)
Standard M.2 socket with PCIe Gen3 x4
Power Supply
5V/5A
Varies, often 12V (30W)
Idle Power Consumption
3.0 to 3.6 Watts
7.5 to 10 Watts (Typ.)
Dimensions
85 x 56 mm
As small as 87.8 x 87.8 x 37 mm (Blackview MP80)
Supported OS
Official Raspberry Pi OS support
Certified Ubuntu 22.04/24.04 support
Several other OSes with various degree of supports
Usually ships with Windows 11 Pro
All x86 operating systems are supported (Linux, FreeBSD, ..) albeit full support depends on available drivers (e.g. for WiFi/Bluetooth)
Some remarks:
Intel N100 systems with DDR4/DDR5 usually rely on one (replaceable/upgradeable) SO-DIMM module while LPDDR5 is soldered on the main board. Raspberry Pi 5 always comes with soldered-on memory.
Not all USB-C ports found in Alder Lake-N mini PCs support Displayport Alt mode, it depends on the model.
Some Intel N100 SBCs such as the AAEON UP 7000 provide a GPIO header, but I wanted to focus on the features in typical mini PCs in this post. It’s also possible to add GPIO headers through a USB adapter
The Blackview MP80 mini PC is not powered by an Intel Processor N100, but by the similar Intel N95 or N97 Alder lake-N processor, and was used to show it’s possible to get a really small x86 mini PC. Most are larger than that, and the Raspberry Pi 5 should be a better option for space-constrained applications.
The Raspberry Pi 5 can be overclocked to get more performance, and some people managed to achieve 1,033 points (single-core) and 2146 points (multi-core) at 3.10 GHz, but it is still lower than on Intel Processor N100 mini PC, and may not work on all Raspberry Pi 5 boards.
The Intel N100-powered GEEKOM Mini Air12 is faster for most tasks, and in some cases up to almost three times as fast (Speedometer 2.0 in Firefox), except for memset (similar results) and OpenSSL AES-256 where the higher sustained single-core CPU frequency helps the Arm SBC.
Raspberry Pi 5 vs Intel N100 mini PC price comparison
This one will be tough as everybody has different requirements, local or import taxes, and so on. But I’ll first calculate the price of a minimum working system and the Raspberry Pi 5 equivalent of the MINIX Z100-0dB mini PC with 8GB RAM, a 256GB NVMe SSD, 2.5GbE, WiFi 6, and a fanless enclosure.
For the minimum working configuration, we’ll assume the user wants a Linux or Windows system that boots to the OS, and connects to the network and a display without any other specific requirements. The Raspberry Pi 5 4GB is good enough for this along with the active cooler, a 5V/5A power adapter (although 5V/3A might do too), and a microSD card. I also searched for the cheapest N100 mini PC I could find with storage and memory: the CHUWI Larkbox X (12GB RAM, 512GB SATA SSD) sold for $125.93 with free shipping on Aliexpress at the time of writing.
I tried to select low-cost items for the Raspberry Pi 5 and considered adding an enclosure unnecessary for the minimum configuration (it would add $5 to $20). Taxes and handling fees are not considered for either device, and the shipping fee is not included for the Raspberry Pi 5 kit which ends up being about $33 cheaper. The Larkbox X mini PC delivers higher performance and offers more memory, dual Gigabit Ethernet, and WiFi 6. The Raspberry Pi 5 remains the ideal candidate for use cases requiring GPIO, low power consumption, and a small size.
Now let’s switch to another user who will wonder “What year is this?!” when hearing or reading the words “gigabit Ethernet”, “WiFi 5”, “4GB RAM”, and/or “microSD card”. He won’t allow any noisy fan to pollute his room either, and he’d been fine with a fanless mini PC like the MINIX Z100-0dB with 8GB RAM that’s currently sold for $220.71 on Amazon excluding taxes with an 8% discount coupon selectable before order.
Let’s see what happens if we try to reproduce this setup with a Raspberry Pi 5 8GB. We’ll still need the 5V/5A power adapter and a micro HDMI cable, but we’ll replace the active cooler with a fanless metal case that can still take HAT expansion boards and the microSD card with an NVMe SSD with an M.2 PCIe Hat. We’ll need a WiFi 6 USB 3.0 dongle and a 2.5GbE USB 3.0 dongle, although HAT expansion boards could be daisy-chained to achieve the same result, but that would start to get messy and be more expensive.
The Raspberry Pi 5 system is still cheaper (by $20) before taking into account the shipping fees which may add up when purchasing from multiple vendors. The EDATEC fanless case is also hard to get as it’s not for sale on Aliexpress anymore, and finding another complete Raspberry Pi 5 case that takes a HAT+ expansion board is challenging. We’ve also created a monster with a HAT and all four USB ports would be used in a typical system with a USB keyboard, a USB mouse, and our two USB 3.0 dongles for WiFi 6 and 2.5GbE. In that specific use case, I’d consider the Raspberry Pi 5 to be undesirable, and people would be better served by a mini PC. I reckon I’ve pushed the requirements a bit far with WiFi 6 and 2.5GbE, as I’d expect many people would be fine the the built-in gigabit Ethernet and WiFi 5 connectivity, in which case the Pi 5 could still be considered.
Final words
As one would expect, there’s no simple answer to the question “Which is the best? A Raspberry Pi 5 SBC or an Intel N100 mini PC?” since it will depend on the user’s specific requirements. The Raspberry Pi SBC was first introduced as cheap hardware for the education market, and I would recommend the Raspberry Pi 4 over the Raspberry Pi 5 for this purpose since it’s cheaper and does the job. The Raspberry Pi 5 is more suitable for projects that require extra performance while keeping the small form factor, GPIO header, and camera connectors. Intel Processor N100 mini PCs offer a better performance/price ratio as a general-purpose computer running Windows 11 or a Linux distribution such as Ubuntu, although you may potentially save a few dollars by using a Raspberry Pi 5.
The Nuvoton NuMicro MA35D0 series is a cost-effective, dual-core Arm Cortex-A35 MPU designed specifically for edge IoT applications. The specifications of this MPU are very similar to the Nuvoton MA35H0, and it looks more like a low-power low-cost version of the NuMicro MA35D1. The MPU features integrated, stacked DDR SDRAM options (128MB or 256MB) to reduce PCB complexity and BOM costs. It also offers some security options with Arm TrustZone, secure boot, cryptographic accelerators, and a True
The Nuvoton NuMicro MA35D0 series is a cost-effective, dual-core Arm Cortex-A35 MPU designed specifically for edge IoT applications. The specifications of this MPU are very similar to the Nuvoton MA35H0, and it looks more like a low-power low-cost version of the NuMicro MA35D1. The MPU features integrated, stacked DDR SDRAM options (128MB or 256MB) to reduce PCB complexity and BOM costs. It also offers some security options with Arm TrustZone, secure boot, cryptographic accelerators, and a True Random Number Generator, making it ideal for industrial applications.
The MPU also offers various peripherals and connectivity options including 2x Fast Ethernet (IEEE 1588 v2) interfaces, high-speed USB, SD3.0/eMMC, three CAN FD, eleven UART, and more. Additionally, the MA35D0 MPU supports touchscreens and a TFT LCD controller with resolutions up to 1280×800.
Nuvoton NuMicro MA35D0 Series specifications:
CPU Sub-system
2x Arm Cortex-A35 cores running at up to 650 MHz
L1 Cache – 32 Kbytes instruction & 32 Kbytes data for each core
L2 Cache – 512 Kbytes shared
Memory Sub-system
On-chip SRAM – Up to 384 KB
SDRAM Controller – Supports DDR2, DDR3, DDR3L (up to 533MHz)
Storage
USB (device or host)
SD/eMMC
NAND Flash
SPI Flash (SPI-NOR/SPI-NAND)
Bootloader – Pre-loaded 128KB mask ROM with all boot options above
Display and Video Sub-system
TFT-LCD Interface – 24-bit, up to 1280×800 @ 60fps, various formats supported
2D Graphics Engine – Blits, rotation, filtering, alpha blending, etc.
Image Decoder – JPEG
Post Processing – Scaling, color conversion, rotation, and more
Networking – 2x 10/100Mbps Ethernet MAC
USB
1x USB 2.0 high-speed host/device (transceiver included)
1x high-speed host only
Other peripherals
Peripheral DMA (PDMA)
3x I2C (master/slave)
2x Secure Digital Host Controllers
3x CAN Bus controllers, compliant with CAN 2.0 and ISO 11898-1:2015
External Bus Interface (EBI) – Up to 3 memory banks
Real-Time Clock(RTC)
Built-in temperature sensor with ±5°C accuracy
Hardware Semaphore
Dimensions – 24 mm x 24 mm
Package – LQFP-EP 216-Pin Package
Temperature range – -40 ~ 85°C (Industrial-grade)
The company has introduced two MPUs with SKU MA35D03F764C and MA35D03F864C the only difference between the MPUs is that the F764C has 128 MB of DDR memory, while the F864C has 256 MB. Other than that, they have 154 GPIOs, 40-bit PDM NAND Flash, watchdog timers, video codecs, and a JPEG decoder.
Nuvoton offers software and hardware tools to shorten the development time. With support for graphics libraries such as SEGGER emWin, Qt, and LVGL, developers can effortlessly create graphical user interfaces (GUIs). The company also shows that it supports Yocto and Buildroot build systems for Linux, secure firmware, and many other resources that can be found on the products page resources section. The company also has its own GitHub repo where some example code and other documentation can be found.
NuMaker-IoT-MA35D0-A1 Dev Board
The company also offers NuMaker-IoT-MA35D0-A1 development board based on the MA35D03F864C MPU. This is great for those who want to get started with this board easily. The company offers Schematic, PCB, Gerber File & BOM for the dev board.
NuMaker-IoT-MA35D0-A1 Dev Board Specifications:
Main Chip – High-performance MA35D0 MPU with dual Cortex-A35 cores (up to 650MHz)
Memory – 256MB of DDR3L RAM in MCP (Multi-Chip Package) for streamlined design
Storage
512MB onboard Quad SPI NAND Flash
1GB on-board NAND Flash
SD2.0 compatible MicroSD slot
SD3.0 compatible Standard SD card slot
USB
1x USB Type-C for 5V power input
1x High-Speed USB Host/Device
1x High-speed USB Host.
Ethernet – 2x Gigabit Ethernet ports (RMII) with PHYs, transformers, and RJ45 connectors.
Audio – Audio codec with mic input and speaker output.
Debugging
USB Virtual COM – UART debug port accessible as a USB virtual COM port.
SWJ Connector – JTAG + SWD debug port.
Boot Options – DIP Switch to Select the boot source at power-on.
User Interface
Buttons – 3x user buttons
LEDs – 3x user LEDs
ADC – 8-channel ADC header
Misc
RS232/RS485 – 2x RS232/RS485 transceivers with DB9/header
CAN FD – 3x CAN FD transceivers with headers
RTC Battery – Header for battery backup of the Real-Time Clock.
Dimension – 143 mm x 106 mm
At the time of writing the company is not selling the MPUs directly but the NuMaker-IoT-MA35D0-A1 dev board is available for purchase for $119.00 on Nuvoton Direct. More information about the IC can be found on the product page or the press release.
QNAP TS-216G 2-bay NAS features a quad-core Arm Cortex-A55 processor with an NPU for AI-powered photo management, 4GB RAM, 2.5GbE and GbE networking ports, two hot-swappable 3.5-inch SATA bay, and a few USB ports.
It looks to be an update to the Rockchip RK3566-powered QNAP TS-133/TS-223 with more memory (4GB vs 2GB) and more advanced networking capability (2.5GbE+GbE vs GbE only), while still keeping the object and face recognition capabilities.
QNAP TS-216G specifications:
SoC – Unnamed but
QNAP TS-216G 2-bay NAS features a quad-core Arm Cortex-A55 processor with an NPU for AI-powered photo management, 4GB RAM, 2.5GbE and GbE networking ports, two hot-swappable 3.5-inch SATA bay, and a few USB ports.
It looks to be an update to the Rockchip RK3566-powered QNAP TS-133/TS-223 with more memory (4GB vs 2GB) and more advanced networking capability (2.5GbE+GbE vs GbE only), while still keeping the object and face recognition capabilities.
QNAP TS-216G specifications:
SoC – Unnamed but likely Rockchip RK3566
CPU – Quad-core Cortex-A55 clocked at up to 2.0 GHz
GPU – Mali-G52
Neural Processing Unit (NPU)
Hardware-accelerated Transcoding
Encryption Engine
System Memory – 4 GB RAM
Storage
4 GB eMMC flash (dual boot OS protection)
2x 3.5-inch SATA III bay also supporting 2.5-inch SATA SSDs; hot-swappable
Networking
2.5GbE RJ45 jack
Gigabit Ethernet jack
Wake-on-LAN (WoL) and Jumbo Frame support
Number of Concurrent Connections (CIFS) – Up to 200
The NAS runs QTS 5.1.3 Linux-based operating system with an App Center that features File Station for NAS file access, sharing, and management through a web browser, Hybrid Backup Sync to back up NAS files to the cloud or another NAS, and Qsync to enable synchronization across multiple users and devices. The NAS is compatible with various client operating systems including Windows 7 to 11, Windows Server 2008 R2 to 2022, several Linux distributions including Ubuntu 14.04 and greater, Apple Mac OS 10.10 or later, and more.
The built-in NPU can accelerate image recognition, and do so at lower power consumption, through the AI-powered QuMagie smart photo management to automatically sort thousands of images based on people, animals, things, or locations, and then allows the users to easily search for those without having to scroll through the photo collection manually.
Location search with QuMagie
QNAP TS-216G ships with a power adapter and power cord, Ethernet cables, flat head screws for SSD and HDD mounting, two drive tray keys, and a Quick installation guide (QIG). You’ll find it on Amazon US for $298.97. More details may be found on the product page.
The Bulgarian hardware manufacturer, Olimex, has designed a new ESP32 board with Power over Ethernet (PoE) functionality. The Olimex ESP32-POE2 board builds on the original ESP32-POE and features the same ESP32-WROOM-32 microcontroller module with Wi-Fi and Bluetooth connectivity. However, it supports up to 25W power delivery from the Ethernet port via PoE and can be used in more demanding projects.
The Olimex ESP32-POE2 board features a low-power design and consumes only 200µA in deep sleep mo
The Bulgarian hardware manufacturer, Olimex, has designed a new ESP32 board with Power over Ethernet (PoE) functionality. The Olimex ESP32-POE2 board builds on the original ESP32-POE and features the same ESP32-WROOM-32 microcontroller module with Wi-Fi and Bluetooth connectivity. However, it supports up to 25W power delivery from the Ethernet port via PoE and can be used in more demanding projects.
The Olimex ESP32-POE2 board features a low-power design and consumes only 200µA in deep sleep mode. The board can be powered via USB-C, Ethernet, or Li-Po battery. The Ethernet interface is built on Texas Instruments’ TPS2378PW chip with IEEE 802.3at classification and legacy PoE support. Powering with PoE requires at least 37V DC for seamless operation.
Two proprietary connectors (UEXT and EXT1) can be used to add sensors and other modules and expand the board’s operation for IoT projects. Users can select between 24V/0.75A or 12V/1.5A power output, with a 5V/1.5A output also available.
Olimex ESP32-POE2 specifications:
Wireless module – ESP32-WROOM-32 module with 802.11 b/g/n WiFi and Bluetooth LE
External Storage – micro SD card (1-bit mode)
Connectivity – 10/100M Ethernet with PoE power management based on TI TPS2378PW IEEE 802.3at Type 1-compliant chip
Expansion
UEXT connector for Olimex modules
EXT connector
Debugging/Programming – USB-C port + CH340 serial to TTL chip
Misc – User button, Reset button, battery level monitor pin, external power supply detection pin
Power Supply
USB-C port port
Via Ethernet cable thanks to PoE
2-pin header for LiPo battery
External Power Supply
24V/0.75A or 12V/1.5A to external circuit
5V/1.5A to external circuit
Operating Temperature – 0°C – 70°C
Dimensions – 59 x 90 mm
ESP32-POE2 is supported by all the popular development platforms: Espressif ESP-IDF, MicroPython, ESPHome, PlatformIO, and Arduino.
Like the original board, the ESP32-POE2 is open-source hardware and KiCAD hardware design files, software, and documentation are hosted in the project’s repository on GitHub. The board is priced at €21 (about $22) and can be bought at Olimex’s store, with bulk discounts available.
Synaptics has unveiled its new Astra platform with a range of SoC and a development kit for edge AI applications. These new processors and a supporting development kit are built to provide out-of-the-box AI capabilities for IoT devices, reducing reliance on cloud-based AI.
This new Synaptics Astra Platform is built around three main SoCs. The SL1680 is built for multi-modal IoT applications and features a quad-core Arm Cortex-A73 CPU, dedicated 7.9 TOPS NPU, and 4K video. The SL1640 is a cost an
Synaptics has unveiled its new Astra platform with a range of SoC and a development kit for edge AI applications. These new processors and a supporting development kit are built to provide out-of-the-box AI capabilities for IoT devices, reducing reliance on cloud-based AI.
This new Synaptics Astra Platform is built around three main SoCs. The SL1680 is built for multi-modal IoT applications and features a quad-core Arm Cortex-A73 CPU, dedicated 7.9 TOPS NPU, and 4K video. The SL1640 is a cost and power-optimized SoC with a quad-core Cortex-A55 CPU and 1.6+ TOPS NPU. Finally, the SL1620 is a graphics and AI accelerator with a quad-core Cortex-A55 CPU and dual-core Imagination BXE-2-32 GPU but does not feature an NPU.
The development kit features a module design where the new swappable compute modules allow flexible configurations. The devkit will support open Yocto Linux distribution and Synaptics AI toolkit for quick AI integration.
Synaptics Astra Platform SL1680 Specifications:
CPU
Quad-Core Arm Cortex-A73 processor clocked at 2.1 GHz delivering up to 40000 DMIPS
64KB I-cache and 32KB DCache
Dedicated Arm NEON technology/VFPU for each processor with 32 128-bit SIMD registers and crypto instructions
1 MB shared L2 Cache
Arm TrustZone technology
Memory/Storage Interfaces
Up to 4GB 64-bit LPDDR4/LPDDR4x-3733 and optional 32-bit DDR configuration
eMMC 5.1 Controller with x1, x4, or x8-bit interface
Neural Processing Unit
Dedicated hardware for localized NN/machine learning applications
Up to 7.9+ TOPS
Supports multiple DNN frameworks and is optimized for TensorFlow Lite inferencing via the SyNAP toolkit
Multi-Standard Video Decoding
Support for various video decoding formats including AV1, H.265, VP9, H.264, MPEG-2, and VP8
Flexible support for PIP and Multi-View configurations
Single-stream 2160p H.265/VP9 decode up to 90-100 fps
Support for up to 1080p120 single-stream decode
Multi-Standard Video Encoding/Transcoding
Support for up to two streams encoding at 1080p60 with H.264 or VP8 per stream
HDR-enabled with support for various graphics standards including OpenGL ES, DirectFB, OpenCL, and Vulkan
Up to 3840×2160 resolution with 3D capabilities
Video/Graphics Display Pipeline with QDEO
Two independent display output paths with support for HDR and gamma correction
Various image enhancement features including scaling, detail enhancement, and color remapping
Security
250MHz Secure CPU (Arm Cortex-M3)
Secure boot with RSA digital signature verification
True random number generator
DRM engine with support for AES, DES, 3DES, RSA, ECC
Memory and I/O space access control
DRAM scrambling support
Audio/Video Inputs & Outputs
Various input/output interfaces including MIPI CSI 2, HDMI, I2S, S/PDIF
Support for multiple camera inputs and HDMI input with HDR support
Standby Mode System Manager – Arm Cortex-M3 for managing standby mode with support for wake-up events such as Voice, LAN, CEC, and GPIO triggers
Peripherals
PCIe 2.0 root complex
Gigabit Ethernet MAC
USB 3.0 host interface, USB 2.0 OTG interface
SDIO host interface
UART interfaces, SPI interfaces, GPIOs, PWM, ADC, and IR receiver input
Package – 17 x 17mm FCBGA with 0.4mm ball pitch
Note! The specifications below are for the SL1680 SoC, the top-of-the-line processor in Synaptics’ SL-Series. While many features are shared across the series, key differences exist between the SL1680, SL1640, and SL1620. To highlight these differences, I’ve included a comparison table below.
Features
SL1620
SL1640
SL1680
CPU
Quad-core Arm Cortex-A55 @ up to 1.9GHz Each with 32KB I-cache / 32KB D-cache and 512KB shared L3 cache
Quad-Core Arm Cortex-A55 @ up to 2.1GHz Each with 64KB I-cache / 32KB D-cache and 1MB shared L2 Cache
Quad-Core Arm Cortex-A73 @ up to 2.1GHz Each with 64KB I-cache / 32KB D-cache
Memory Interface
32-bit DDR3/DDR4 up to 4GB with NAND/eMMC controller
32-bit DDR4-3200/LPDDR4x-3733 up to 4GB with eMMC 5.1 controller
64-bit LPDDR4x-3733 up to 4GB with eMMC 5.1 controller
Neural Processing Unit
-
Up to 1.6+ TOPS NPU
Up to 7.9+ TOPS NPU
Multi-Standard Video Decoding
-
Video decoding support with AV1, H.265, VP9, H.264, etc
Multi-standard video decoding support with AV1, H.265, VP9, H.264, etc.
Multi-Standard Video Encoding
-
-
Support for multiple encoding streams with (1080p60: H.264 or VP8 per stream)
Audio Decoding/Processing
Interface for DVFA101 support with TDM/I2S interfaces
Far-Field Voice, Keyword Detection with Audio decompression, post-processing
Far-Field Voice, Keyword Detection with Audio decompression, post-processing
Audio/Video Outputs
Dual display support
HDMI, MIPI DSI, I2S, S/PDIF, etc.
HDMI, MIPI DSI, I2S, S/PDIF, etc.
2D & 3D Graphics
Imagination BXE-2-32 GPU with OpenGL ES 3.2, OpenCL 3.0, Vulkan 1.3, etc.
Imagination PowerVR Series9XE GE9608 GPU with OpenGL ES 1.1/2.0/3.0/3.1/3.2, Vulkan 1.1
Imagination PowerVR Series9XE GE9920 GPU with OpenGL ES 1.1/2.0/3.0/3.1/3.2, Vulkan 1.1
Video/Graphics Display Pipeline with QDEO
-
Support for display paths, MP, GFX, etc.
Support for display paths, MP, GFX, etc.
Peripherals
USB, SDIO, TWSI, UART, SPI, GPIO, etc.
USB, PCIe, Ethernet, SDIO, TWSI, etc.
PCIe, USB, Ethernet, SDIO, TWSI, etc.
Standby Mode System Manager
Power gating, Wake-on-LAN, GPIOs
Power gating, Wake-on-Voice, GPIOs
Power gating, Wake-on-Voice, GPIOs
Power, Package and Layout
13mm x 13mm FCBGA, 0.4mm ball pitch
13mm x 13mm FCBGA, 0.4mm ball pitch
17mm x 17mm FCBGA, 0.4mm ball pitch
High-level block diagram of the SL1620, SL1640 and SL1680
In terms of software, the board supports a Yocto Linux distribution included in the Synaptics ESSDK available on Synaptics Astra’s GitHub. Synaptics also offers a guide to help you get started with their platform, streamlining the learning curve. Additionally, the company uses open-source libraries and frameworks like GStreamer and V4L2 to create robust video and display pipelines. This enables applications such as object detection, classification, and segmentation. The kit also facilitates the development of AI-enabled audio algorithms for far-field voice, wake words, echo cancellation, noise reduction, and even voice biometrics.
The Astra Machina Foundation Series evaluation kit supports the full range of SL1600 series SoCs (including the SL1680, SL1640, and SL1620). This kit features daughter cards (SYN43711, SYN43752,SYN43756(E)) for integrated Wi-Fi and Bluetooth connectivity, providing flexibility for seamless wireless integration.
Synaptics Astra Platform SL-Series AI-Native IoT SoC (SL1680, SL1640, and SL1620) are not available for purchase at the time of writing, although more details can be found on the product page and additional information may also be found on the press release.
Azulle Access Pro is a PC stick based on an Intel N100 quad-core Alder Lake” processor with up to 8GB RAM, 128GB eMMC flash, and a male HDMI port which is offered with Windows 11 Pro, Linux, or “Zoom” operating systems (more on that below).
The mini PC also provides a microSD card slot, gigabit Ethernet, WiFi 6 and Bluetooth 5.2 with an external antenna, USB 3.0 Type-A and Type-C ports, and a 3.5mm audio jack. We’ve written about pocket-sized mini PCs based on Alder Lake-N processors before, for
Azulle Access Pro is a PC stick based on an Intel N100 quad-core Alder Lake” processor with up to 8GB RAM, 128GB eMMC flash, and a male HDMI port which is offered with Windows 11 Pro, Linux, or “Zoom” operating systems (more on that below).
The mini PC also provides a microSD card slot, gigabit Ethernet, WiFi 6 and Bluetooth 5.2 with an external antenna, USB 3.0 Type-A and Type-C ports, and a 3.5mm audio jack. We’ve written about pocket-sized mini PCs based on Alder Lake-N processors before, for instance, the MeLe PCG02 Pro N100 or the slightly wider MeLe Quieter4C, but the Azulle model is different since it comes with a male HDMI connector which would allow users to connect it directly into a TV or monitor without a cable.
Azulle Access Pro specifications:
SoC – Intel Processor N100 quad-core Alder Lake-N processor @ up to 3.4 GHz (Turbo) with 6MB cache, 24EU Intel HD graphics @ 750 MHz; TDP: 6W
Memory/Storage configurations
4GB LPDDR4/LPDDR5 and 64GB eMMC flash
8GB LPDDR4/LPDDR5 and 128GB eMMC flash
MicroSD card slot on all variants
Video Output – Male HDMI connector up to 4Kp60
Audio – 3.5mm audio jack, and digital audio output via HDMI
Connectivity
Gigabit Ethernet RJ45 port
WiFi 6 and Bluetooth 5.2 via Intel AX201 wireless module
USB
1x USB 3.2 Gen 1 Type-A port
1x USB 3.2 Gen 1 Type-C port
Misc – Power button, Kensington lock slot
Power Supply – 12V/2A
Dimensions – 114 x 51 x 20 mm
Weight – 130 grams
The Azulle “Access Pro Alder Lake” ships with a power adapter, an “HD adapter”, and a Quick Guide. The “HD adapter” is an HDMI female-to-female adapter for connecting an HDMI cable, since while it’s indeed possible to connect the PC stick directly to the back or side of a TV, the weight may end up damaging the HDMI port on either mini PC or display over time especially if you connect an Ethernet cable.
Besides the form factor, the other differentiating factor of the Access Pro is the OS support: Windows 11 Pro, Linux, or Zoom. We asked for some clarifications from Azulle, and the company told CNX Software that Linux would be Ubuntu 22.04 for now, and the “Zoom” option is Window 10 IoT 2021 LTSC preconfigured to boot to Zoom which can be used for regular Zoom use, meetings, etc…
Arducam Pivistation 5 is an all-in-one Raspberry Pi 5 camera kit that aims to provide a turnkey hardware and software solution to quickly get started with computer vision applications and offered with a choice of camera sensors designed for various applications.
The system looks like a Raspberry Pi 5 SBC housed in the official case fitted with a camera. Three models are available, namely the “Hawkeye” featuring a high-resolution 64MP autofocus camera, the “Darksee” with an 8MP camera sensor with
Arducam Pivistation 5 is an all-in-one Raspberry Pi 5 camera kit that aims to provide a turnkey hardware and software solution to quickly get started with computer vision applications and offered with a choice of camera sensors designed for various applications.
The system looks like a Raspberry Pi 5 SBC housed in the official case fitted with a camera. Three models are available, namely the “Hawkeye” featuring a high-resolution 64MP autofocus camera, the “Darksee” with an 8MP camera sensor with ultra low-light sensitivity, and the “Klarity” with a 20MP camera with fixed focus and a large 1-inch sensor.
Dual mode Auto focus (PDAF + CDAF)
64MP resolution
Ultra-high sensitivity
Low-light performance
Low noise
Improved sensitivity in near-infrared (NIR) region
20MP resolution
Higher clarity and sharpness with 1-inch image sensor
Good balance between speed, resolution and sensitivity
All the cameras above rely on a rolling shutter, but the company is also working on the upcoming Arducam Pivistation 5 Swift model that includes a global shutter for robotics applications.
Example use case: traffic monitoring
Besides the pre-assembled hardware, the Arducam Pivistation 5 family aims to quicken the development process with pre-installed software typically used for computer vision applications including opencv-python, libopencv-dev (C++ version), libcamera, rpicam-apps, TensorFlow (Python), PyTorch, camera-streamer, and Arducam camera server.
You could probably reproduce such setup yourself relatively easily within a couple of days, but if you prefer getting a system that works out-of-the-box and focus on your application development instead of setting up systems, then the Pivistation 5 could be interesting.
Arducam has launched a crowdfunding campaign on Kickstarter with a lowly $5,000 funding target that’s been surpassed within a few hours. Pledges start at $139 for the Pivistation 5 Haweye, $179 for the Darksee, and $239 for the Klarity. They all include a Raspberry Pi 5, camera, and enclosure, and those are Early bird prices with a 30 to 40 percent discount over the expected retail price. Shipping adds about $35 and it’s not a fully global campaign with the company only shipping to North America, Europe, and a few countries in Asia and Oceania. Backers should expect their package to ship around August 2024 if everything goes according to plans.
Canonical has just released Ubuntu 24.04 LTS “Noble Numbat” distribution a little over two years after Ubuntu 22.04 LTS “Jammy Jellyfish” was released. The new version of the operating system comes with the recent Linux 6.8 kernel, GNOME 46, and a range of updates and new features we’ll discuss in this post.
As a long-term support release, Ubuntu 24.04 LTS gets a 12-year commitment for security maintenance and support, with five years of free security maintenance on the main Ubuntu repository, a
Canonical has just released Ubuntu 24.04 LTS “Noble Numbat” distribution a little over two years after Ubuntu 22.04 LTS “Jammy Jellyfish” was released. The new version of the operating system comes with the recent Linux 6.8 kernel, GNOME 46, and a range of updates and new features we’ll discuss in this post.
As a long-term support release, Ubuntu 24.04 LTS gets a 12-year commitment for security maintenance and support, with five years of free security maintenance on the main Ubuntu repository, and Ubuntu Pro extending that commitment to 10 years on both the main and universe repositories (also free for individuals and small companies with up to 5 devices). This can be extended a further 2-year, or 12 years in total, for Ubuntu Pro subscribers who purchase the Legacy Support add-on.
Canonical explains the Linux 6.8 kernel brings improved syscall performance, nested KVM support on ppc64el, and access to the new bcachefs filesystem, and the company has also merged low-latency kernel features into the default kernel, hence reducing kernel task scheduling delays.
GNOME version 46 is said to deliver performance and usability improvements, including file manager search and performance (that’s been an issue for me with directories with many files loading slowly), expandable notifications, and consolidated settings options for easier access.
Developers get access to the latest version of the most popular languages with Python 3.12, Ruby 3.2, PHP 8.3, and Go 1.22, as well as .NET 8, OpenJDK 21 for Java programmer, and Rust 1.75 with a simpler Rust toolchain snap framework. Ubuntu 24.04 LTS also enables frame pointers by default on all 64-bit architectures to make it easier to optimize software thanks to a more complete CPU profiling and off-CPU profiling.
The installer has been revamped and relies on the same Subiquity back end as Ubuntu Server and a new Flutter-built front-end designed to improve access to accessibility options and increase clarity on the user experience.
The installer notably brings back support for experimental ZFS installs with or without hardware-backed full disk encryption, makes dual-boot installation easier, especially when BitLocker is enabled, and adds support for autoinstall in the graphical installer. The latter is used for custom installations and users can simply provide a local or remote autoinstall.yaml file to install a customized Ubuntu 24.04 image on multiple machines.
Other changes include a new Flutter-based App Center with a more modern design and improved user experience, and people will have to learn yet another method to configure networking, since Netplan 1.0 become the default tool on the desktop, as it has been on Ubuntu Server and Cloud since 2016. However, Canonical stresses that Netplan does not replace NetworkManager and will not impact workflows that prefer the previous configuration methods.
That means we’ll have to wave goodbye to Ubuntu 22.04 for all our mini PC reviews, and we’ll only use Ubuntu 24.04 forward, although I’m sure some Arm and RISC-V SBCs will still use older versions for a while. You’ll find the ISOs for Ubuntu 24.04 Desktop, Server, and other flavors on the download page. Raspberry Pi hardware was not left behind, and Ubuntu 24.04 Desktop and Server images can already be downloaded for the Raspberry Pi 5, Raspberry Pi Zero 2 W, Raspberry Pi CM4, Raspberry Pi 400, and Raspberry Pi 4 on a dedicated page.
Firefly AIBOX-1684X is a compact AI Box based on SOPHON BM1684X octa-core Arm Cortex-53 processor with a 32 TOPS AI accelerator suitable for large language models (LLM) such as Llama 2, Stable Diffusion image generation solution, and traditional CNN and RNN neural network architectures.
Firefly had already released several designs based on the SOPHON BM1684X AI processor with the full-featured Firefly EC-A1684XJD4 FD Edge AI computer and the AIO-1684XQ motherboard, but the AIBOX-1684X AI Box off
Firefly AIBOX-1684X is a compact AI Box based on SOPHON BM1684X octa-core Arm Cortex-53 processor with a 32 TOPS AI accelerator suitable for large language models (LLM) such as Llama 2, Stable Diffusion image generation solution, and traditional CNN and RNN neural network architectures.
Firefly had already released several designs based on the SOPHON BM1684X AI processor with the full-featured Firefly EC-A1684XJD4 FD Edge AI computer and the AIO-1684XQ motherboard, but the AIBOX-1684X AI Box offers the same level of performance, just without as many interfaces, in a compact enclosure measuring just 90.6 x 84.4 x 48.5 mm.
AIBOX-1684X AI box specifications:
SoC – SOPHGO SOPHON BM1684X
CPU – Octa-core Arm Cortex-A53 processor @ up to 2.3 GHz
TPU – Up to 32TOPS (INT8), 16 TFLOPS (FP16/BF16), 2 TFLOPS (FP32)
VPU
Up to 32-channel H.265/H.264 1080p25 video decoding
Up to 32-channel 1080p25 HD video processing (decoding + AI analysis)
Up to 12-channel H.265/H.264 1080p25fps video encoding
System Memory – 8GB, 12GB, or 16GB LPDDR4/LPDDR4X
Storage
32GB, 64GB, or 128GB eMMC flash
MicroSD card slot
Video and audio output – HDMI port up to 1080p30
Networking – 2x Gigabit Ethernet RJ45 ports
USB – 2x USB 3.0 ports, 1x USB-C serial debug port
Misc – Power button, LED
Power Supply – 12V/4A via 5.5/2.5mm power barrel jack
Dimensions – 90.6 x 84.4 x 48.5 mm
Weight – 420 grams
Temperature Range – Operating: -20°C to 60°C; storage: -20°C to 70°C
Humidity – 10%～90% (non-condensing)
Firefly has yet to provide documentation for the new AIBOX-1684X AI box, but the Wiki for one of the earlier BM1684X systems shows it runs Ubuntu 20.04 with Linux 5.4 LTS kernel (image released in July 2023) and docker is also apparently preinstalled in the image.
The company also explains the AIBOX-1684X is suitable for the local deployment of “ultra-large-scale parameter models under the Transformer architecture” such as large language models (LLM) such as LLaMa2, ChatGLM, and Qwen, as well as large vision models like ViT, Grounding DINO, and SAM. Stable Diffusion V1.5 image generation model is also supported and so are traditional network architectures such as CNN, RNN, and LSTM (Long short-term memory). Users can rely on popular frameworks such as TensorFlow, PyTorch, MXNet, PaddlePaddle, ONNX, and Darknet for pre-trained or custom models. More details about the software can be found in the SOPHON SDK section of the wiki.
Llama3 on Radxa Fogwise AirBox
Firefly is not the only company to launch an AI Box based on a SOPHON processor since the Radxa Fogwise Airbox has just been unveiled based on the SOPHON SG2300X processor that looks almost identical to the BM1684X, except it delivers (only) 24 TOPS. It seems odd to have similar CPUs doing more or less the same thing, so I asked, and I’ve been told SG2300X supports “open-source” generative AI, while the BM1684X does not. I’ll be able to find out more and test it by myself soon since I expect to receive a sample of the Fogwise Airbox by the end of the month or early May.
In the meantime, you can try the just-released Llama3 8B model on the Radxa Fogwise Airbox as I did when asking “What do you know about CNX Software?”. Some details are incorrect, but the text was rendered relatively fast.
The Firefly AIBOX-1684X is available now, but only in China (Taobao) for 2,499 CNY or about $344 US, but Firefly told CNX Software they’d also sell it on their online shop “very soon”, so anybody can get it. [Update April 28, 2024: The AIBOX-1684X with 16GB RAM and 64GB flash is now sold for $369] For reference, the Radxa Fogwise Airbox – which I’ll cover in more detail once I get a sample – can be pre-ordered for $321 and up on Arace Tech. Additional information about the Firefly AIBOX-1684X can also be found on the product page.
We’ve already checked out the specifications, and gone through an unboxing and a teardown of the GEEKOM XT12 Pro mini PC in the first part of the review, so we’ll now report our experience with the 12th Gen Intel Core i9-12900H mini PC with Windows 11 Pro operating system.
In this second part of GEEKOM XT12 Pro, we’ll test the features of the Alder Lake mini PC in Windows 11 Pro, run benchmarks, test networking and storage, evaluate cooling performance, and finally provide some numbers for fan n
We’ve already checked out the specifications, and gone through an unboxing and a teardown of the GEEKOM XT12 Pro mini PC in the first part of the review, so we’ll now report our experience with the 12th Gen Intel Core i9-12900H mini PC with Windows 11 Pro operating system.
In this second part of GEEKOM XT12 Pro, we’ll test the features of the Alder Lake mini PC in Windows 11 Pro, run benchmarks, test networking and storage, evaluate cooling performance, and finally provide some numbers for fan noise and power consumption.
Software overview and features testing
The System->About menu confirms that we have an XT12 Pro mini PC with a 12th Gen Intel Core i9-12900H processor clocked at 2.50 GHz (base frequency) and 32GB RAM that runs Windows 11 Pro operating system version 23H2. The system was tested with Windows 11 OS build 22631.3447.
HWiNFO64 program provides more details about the Intel Core I9-12900H 14-core (6x P-cores, 8x E-cores)/20-thread processor, the motherboard, and Intel Irix Xe graphics.
Additional details about the 96EU Intel Iris Xe Graphics (Alder Lake-P GT2) can also be found in TechPowerUp GPU-Z.
The PL1 and PL2 power limits are respectively set to 35W and 80W while the Intel Core i9-12900H has a 45W TDP/PBP (Processor Base Power). GEEKOM went with the 35W (Minimum Assured Power) value which may lower performance, but the CPU should run at cooler temperatures.
HWiNFO64 reports two 16 GB Lexar DDR4-3200MHz SO-DIMM modules are used in the mini PC using Shenzhen Longsys Electronics chips.
Windows Task Manager confirms the system comes with 32GB of RAM running at 3,200 MHz.
We can check the network adapters in Device Manager where we’ll find an an Intel i225-V 2.5GbE Ethernet controller and a MediaTek MT7922 wireless module for WiFi 6E and Bluetooth 5.2.
Further Information about the 2.5GbE interface in the GEEKOM XT12 Pro can be found in HWiNFO64…
… and also shows the “MediaTek MT7922 Wi-Fi 6E 160 MHz Wireless LAN Card” with up to 2,999 Mbps link speed and set to 1,200 Mbps at the time of the screenshot.
Let’s go based to the Device Manager windows to check the Bluetooth version.
LMP 12 looks up to Bluetooth 5.3, while Bluetooth 5.2 is advertised in the specifications. We also quickly tested Bluetooth connectivity by transferring a file from an Android smartphone to the computer.
We’ll now test the USB4, USB 3.2, and USB 2.0 ports on GEEKO< XT12 Pro using an ORICO M234C3-U4 M.2 NVMe SSD enclosure for USB 3.x/4 ports and a USB 3.0 hard drive for the USB 2.0 port, as well as the HWiNFO64 program to verify the version and speed and CrystalDiskMark to confirm the transfer speed.
Here are examples for each type of port starting with the left USB 3.2 Type-A port on the front panel…
… followed by the left USB4 Type-C port on the rear panel…
… and finally the USB 2.0 port.
Here’s a summary of the results of all 6 ports (from left to right)
Front panel
USB-A #1 – USB 3.2 – USB 3.1 SuperSpeedPlus (10 Gbps) – Read speed: 1,040 MB/s; write speed: 955 MB/s
USB-A #2 – USB 3.2 – USB 3.1 SuperSpeedPlus (10 Gbps) – Read speed: 1,046 MB/s; Write speed: 955 MB/s
USB-A #1 – USB 3.2 – USB 3.1 SuperSpeedPlus (10 Gbps) – Read speed: 1,045 MB/s; write speed: 966 MB/s
USB-A #2 – USB 2.0 – USB 2.0 Hight-Speed (480 Mbps) – Read speed: 43.40MB/s; write speed: 32.03 MB/s
All ports work as expected considering the read speed of the Apacer SSD we use tops at 2,200 MB/s.
The GEEKOM XT12 Pro mini PC supports up to 4 independent displays, so we connected four monitors through the two HDMI 2.0 ports and two USB 4 Gen3 ports.
We don’t own any (working) 4K or 8K capable display, but we could still confirm quad-display setups can work just fine using a 14-inch CrowView laptop monitor (USB-C), a 15.6-inch CrowVi USB-C display, an HDMI television, and the Kamvas Pro 16 (2.5K) drawing tablet with 2.5K resolution connected through HDMI.
GEEKOM XT12 Pro benchmarks on Windows 11 Pro
Before running Windows 11 benchmarks on the GEEKOM XT12 Pro mini PC, we set the Power mode to “Best performance” in the settings.
Note that the ambient temperature was 30 to 32°C, or slightly hotter than in previous reviews (26 to 30°C), and there may be an impact on the results.
Let’s start with PCMark 10.
That would be 6,063 points in PCMark 10.
The next benchmark was 3DMark’s Fire Strike where the system achieved 4,982 points.
The GEEKOM XT12 Pro got 5,085 points in PassMark PerformanceTest 11 with an impressive Disk Mark result of over 31,003 points. We used CrystalDiskMark to confirm the excellent performance of the 1TB Lexar NM7A1 SSD with 5,096 MB/s sequential read speed and 4,438 MB/s sequential write speed. Not bad at all, although not quite as performant as the WD PC SN740 SSD used in the Khadas Mind Premium.
The Cinebench R23 benchmark can help us estimate the single-core and multi-core performance of the Core i9-12900H CPU.
The GEEKOM mini PC achieved 7,466 points in the multi-core benchmark, and 1,769 points in the single-core test, with an MP Ratio of 4.22x for a 14-thread processor… So some thermal or power-throttling certainly occurs here.
We started benchmarking the Intel Iris Xe integrated GPU with Unigine Heaven Benchmark 4.0. The GEEKOM XT12 Pro managed to render the scene at 55.9 fps on average with a score of 1,409 points at the usual 1920×1080 resolution.
YouTube 4K and 8K video streaming was tested in Google Chrome web browser.
4Kp30 worked close to flawlessly with only 8 frames dropped out of 15,178.
Some system starts to struggle with 4K 60fps video playback, but the GEEKOM XT12 Pro had no such issue with no frames dropped at all after watching the video for over 8 minutes.
8K 30fps works fine on most systems, and no problem here either.
8K 60FPS often becomes a challenge in mini PCs, but the GEEKOM XT12 Pro passed that test as well with only 5 frames dropped out of 25,975 while playing the video for a little over 7 minutes.
The mini PC is one of the best when it comes to smooth YouTube video playback up to 8Kp60 videos with very few dropped frames, even despite the relatively high ambient temperature.
GEEKOM XT12 Pro Windows 11 benchmarks comparison against other mini PCs
Let’s compare Windows 11 benchmark results for the Intel Core i9-12900H-powered GEEKOM XT12 Pro against other systems such as GEEKOM Mini IT12 (Intel Core i7-12650H), GEEKOM Mini IT13 (Intel Core i9-13900H), GEEKOM A7 (AMD Ryzen 9 7840HS), and the Khadas Mind Premium (Intel Core i7-1360P).
But first, let’s list the key specifications of the five systems.
GEEKOM XT12 Pro
GEEKOM Mini IT12
GEEKOM Mini IT13
GEEKOM A7
Khadas Mind Premium
SoC
Intel Core i9-12900H
Intel Core i7-12650H
Intel Core i9-13900H
AMD Ryzen 9 7840HS
Intel Core i7-1360P
CPU
14-core/20-thread (6P+8E) processor up to 5.00 GHz (P-cores) and 3.80 GHz (E-Cores)
10-core/16-thread (6P+4E) processor up to 4.70 GHz (P-cores) and 3.5 GHz (E-cores)
14-core/20-thread (6P+8E) processor up to 5.4 GHz (P-Cores) and 4.1 GHz (E-Cores)
8-core/16-thread processor up to 4.0GHz
12-core/16-core (4P+8E) processor up to 5.0 GHz (P-Cores) and 3.7 GHz (E-Cores)
GPU
96EU Intel Iris Xe Graphics
64 EU Intel UHD Graphics
96 EU Intel Iris Xe Graphics
AMD Radeon 780M Graphics
96 EU Intel Iris Xe Graphics
System Memory
32GB DDR4-3200
32GB DDR4-3200
32GB DDR4-3200
32GB DDR5-5600
32GB LPDDR5-5200
Storage
1TB NVMe SSD
1TB NVMe SSD
2TB NVMe SSD
2TB NVMe SSD
1TB NVMe SSD
Default OS
Windows 11 Pro
Windows 11 Pro
Windows 11 Pro
Windows 11 Pro
Windows 11 Home
And now for the benchmark results.
GEEKOM XT12 Pro
GEEKOM Mini IT12
GEEKOM Mini IT13
GEEKOM A7
Khadas Mind Premium
PCMark 10
6063
5627
6681
7516
5904
- Essentials
11160
10714
11938
11528
11038
- Productivity
7376
7052
8341
10370
7589
- Digital content creation
7344
6401
8126
9639
6667
3DMark (Fire Strike)
4982
3997
5387
8534
5427
PerformanceTest 11.0
5085
3521
5580.4
8058.2
5378
- CPU Mark
23338
18532
25363.1
30719.8
21786
- 2D Graphics Mark
777
569
547.6
931.9
631
- 3D Graphics Mark
3444
2161
3728.2
7226.1
3622
- Memory Mark
3259
2939
3925.9
3391.4
3642
- Disk Mark
31003
22721
38135.5
38590
42395
Cinebench R23
- Single Core
1769
1781
1943
1831
1878
- Multi Core
7466
5273
11855
15231
9384
The performance of the GEEKOM XT12 Pro is quite similar to the performance of the Khadas Mind Premium, which might not be too surprising considering both are 14-core/20-thread processors, albeit the former comes with 12th Gen Core-i9 processor and the latter a 13th Core-i7 processor. The GEEKOM XT12 Pro has a lower multi-core score despite higher PL1/PL2 power limits (35W/80W vs 28W/64W), so cooling may be the reason or the higher room temperature as discussed above. GEEKOM A7 remains the fastest mini PC we’ve tested so far. As one would expect the XT12 Pro offers are nice performance boost over the mid-range Mini IT12 in all benchmarks.
Network performance testing (2.5GbE and WiFi 6)
We’ll use the iperf3 utility to test both Ethernet and WiFi using the UP Xtreme i11 Edge mini PC on the other side.
534 Mbps downloads and 300 Mbps uploads are quite underwhelming in our test environment since – for instance – the Khadas Mind Premium achieved 712 Mbps and 590 Mbps respectively in the same test. Having said that, the GEEKOM A7 achieves similar results using the same MediaTek MT7922 module instead of the Intel AX211 module found in the Khadas device. I also expect some potential issues in Ubuntu, especially with regard to Bluetooth, based on our experience with the A7. WiFi is perfectly stable here, and most people won’t care about the performance, for example, I have a 300 Mbps broadband connection here, so it would not make much difference to have a faster wireless connection, but people wanting a bit more throughput might consider a system with an Intel wireless module instead.
The situation is actually better than shown in the test above. There have been some discussions about iperf3 vs ntttcp utilities in recent days due to Microsoft telling people to avoid using iperf3 in Windows. While our testing did not reveal any differences between the two, others did find out that ntttcp would indeed perform better with WiFi 6/6E, especially at 5GHz and 6GHz. Eventually, we discovered that while the iperf 3.1.3 (which shows up first in web searches) might be slower in some cases, the more recent iperf 3.16 utility delivers the same performance as ntttcp in Windows. So let’s try it.
That’s more like it. Barring any driver issues, I’d expect the performance in Linux to be very similar, and from now on we will be testing WiFi 6 with iperf 3.16 or greater in Windows. The only downside will be that comparisons with earlier Windows mini PC reviews won’t be valid.
Thermal testing
We ran the 3Dmark’s Fire Strike benchmark while monitoring the system with HWiNFO64 to check the maximum CPU temperature under a CPU+GPU load, and the package reached up to 104°C with both CPU thermal and power occuring.
As usual, it’s quite challenging to cool a high-end mobile CPU in such a small form factor. Some of GEEKOM’s upcoming mini PCs implement a new “IceFlow 1.5 Technology” which the company describes as an
actively-cooled system that consists of a highly efficient CPU cooling fan, a large sink, dual heat pipes, and best-in-class thermal grease. The cooling fan and heat pipes are placed strategically to quickly transfer heat away from key internal components, letting you enjoy unthrottled performance and comfortable temperature in hardware-intensive computing duties”.
That’s something we’ll have to test once we get a sample…, and the first one should be the GEEKOM A7E, an upgraded version of the GEEKOM A7.
Fan noise
The mini PC comes with a fan that’s barely audible most of the time but that becomes a little noisy under heavy loads. We used a sound level meter 5 cm from the top of the device to measure the fan noise:
Idle – 37.8 to 41.3 dBA
3Dmark’s Fire Strike – 49.5 – 54.7 dBA
The room’s background noise is 38 to 39 dBa.
GEEKOM XT12 Pro power consumption
The power consumption of the GEEKOM XT12 mini PC running Windows 11 Pro was measured using a wall power meter as follows:
Power off – 1.2 to 1.4 Watts
Idle – 6.7 to 8.0 Watts
Web browsing – 9.4 to 27.6 Watts
Video playback – 14.7 – 24.3 Watts (YouTube 8K 60fps in Chrome)
3DMark’s Fire Strike – 47.2 – 52.2 Watts
Cinebench R23 Multi-core
First few seconds – 78 – 80 Watts
Long run – 51.5 – 53.0 Watts
The mini PC was connected to WiFi 6, an RF dongle for a mouse and a keyboard, and the Kamvas Pro 16(2.5K) drawing tablet to the mini PC through an HDMI cable and a separate USB-C power adapter.
Conclusion
GEEKOM XT12 Pro mini PC works great in Windows 11 Pro thanks to its 14-core Intel Core i9-12900H processor clocked up to 5.0 GHz with 32GB RAM, a fast 1TB M.2 NVMe SSD, 2.5GbE networking, decent WiFi 6 performance, and several USB 2.0, USB 3.2, and US4 port for expansion.
Benchmarks scores are all faster than the GEEKOM Mini IT12 (Core i7-12650H) on Windows 11, 4K and 8K YouTube video playback is close to flawless in Google Chrome, and overall, everything works as expected. The mini PC does throttle (thermal and power) in 3Dmar’sk Fire Strike, at least in the hot tropical climate we have in Thailand, and since it’s not a fanless computer, there’s some fan noise under heavy loads, but it’s not too disturbing/annoying. In the third and final part of the review, we will test the performance of the GEEKOM XT12 Pro mini PC with the just-released Ubuntu 24.04 to find out how the system performs in Linux.
We’d like to thank GEEKOM for sending the XT12 Pro mini PC with an Intel Core i9-12900H processor, 32GB RAM, and a 1 TB NVMe SSD for review. It can be purchased for $664 on Amazon or GEEKOM US after applying the coupon code cnxXT12Pro for a 5% discount valid until May 31, 2024. Readers based in the UK can also use that coupon on GEEKOM UK.
Raspberry Pi has announced three new versions of its Compute Module 4S (CM4S) offering 2GB, 4GB, and 8GB RAM options which is an upgrade from the 1GB variant introduced in 2022. These new modules feature the same Broadcom BCM2711 quad-core Cortex-A72 SoC used in Raspberry Pi 4 and Raspberry Pi CM4 and have a starting price of only $25. The main downside is that the modules can only be purchased in bulk (200-unit boxes) through Raspberry Pi Approved Resellers.
Raspberry Pi Compute Module 4S (CM4
Raspberry Pi has announced three new versions of its Compute Module 4S (CM4S) offering 2GB, 4GB, and 8GB RAM options which is an upgrade from the 1GB variant introduced in 2022. These new modules feature the same Broadcom BCM2711 quad-core Cortex-A72 SoC used in Raspberry Pi 4 and Raspberry Pi CM4 and have a starting price of only $25. The main downside is that the modules can only be purchased in bulk (200-unit boxes) through Raspberry Pi Approved Resellers.
Raspberry Pi Compute Module 4S (CM4S) specifications:
CPU – Quad-core 64-bit Cortex-A72 processor @ 1.5 GHz
GPU – VideoCore VI GPU
3D Graphics: OpenGL ES 3.0 and Vulkan 1.1
Video
H.265 (HEVC) (up to 4Kp60 decode)
H.264 (up to 1080p60 decode, 1080p30 encode)
Memory – 1GB, 2GB, 4GB, and 8GB LPDDR4-3200 SDRAM with ECC
Storage – 8GB, 16GB, or 32GB eMMC flash; Option for 0GB eMMC Flash (CM4S Lite only)
Ports and Interfaces
Display interfaces
HDMI 2.0 port (up to 4Kp60 supported)
2-lane MIPI DSI display interface
4-lane MIPI DSI display interface
Composite TV out (PAL or NTSC)
Camera interfaces
2-lane MIPI CSI camera interface
4-lane MIPI CSI camera interface
USB – 1x USB 2.0 port (high speed)
Other I/Os
46x GPIO signals
1x SDIO 2.0 (CM4SLite only)
Supply Voltage – Requires VBAT (2.5V to 5V) and +3.3V supplies. +1.8V is no longer used but can be supplied for backward compatibility.
Dimensions – 67.6mm × 31.0mm (compatible with JEDEC MO-224 mechanical specification for 200-pin DDR2), but not electrically compatible with DDR2 SODIMM modules.
Production lifetime – Raspberry Pi Compute Module 4S will remain in production until at least January 2034.
Raspberry Pi Compute Module 4S dimensions
The CM4S is different from other Compute Modules in a few ways. First, it uses the same form factor as the CM3+ but has a faster processor, the BCM2711 chip. Second, it lacks some of the features of the CM4, such as Wi-Fi, Bluetooth, USB 3.0, and PCI Express. CM4S exists because there are a lot of industrial users who have CM3-based designs that they can’t get parts for. The CM4S is a drop-in upgrade for those users because it uses the same form factor.
The new eMMC-equipped Raspberry Pi CM4S or the eMMC-free CM4S Lite can be purchased from the Raspberry Pi Approved Reseller listed on the company’s website. For more information, you can also check out their press release page.
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
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).
The BG95-S5 module supports satellite communication in both L-band (B255) and S-band (B256/23), making it versatile for various applications. The S-band is ideal for GPS navigation, weather radar, and air traffic control, while the L-band enables broader satellite communication, particularly for IoT applications.
At the time of writing, the module is not compatible with Skylo’s NTN network, but the company mentions that certification procedures are underway. If you need Skylo NTN compatibility now, consider modules like the Blues Starnote, Qualcomm 212S, or 9205S. For cellular IoT projects, Blues also recently announced the cost-effective Blues Notecard XP, which is a cost-down version of Notecard Cellular.
The company has not released any pricing information for the module, but more information can be found on the press release and the product page.
Austrian embedded equipment manufacturer, Ronetix, has released two i.MX93 system-on-modules (SoMs): the RNX-iMX93-OSM that follows the Open Standard Module (OSM) form factor, and the RNX-iMX93-SMARC compliant with the SMARC 2.1 (smart mobility architecture) standard.
RNX-iMX93-SMARC system-on-module
Ronetix i.MX93 SoMs are powered by the NXP i.MX93 processor featuring a 64-bit dual-core Arm Cortex-A55 application processor running at 1.7GHz and a Cortex-M33 core running at 250MHz for low-latenc
Austrian embedded equipment manufacturer, Ronetix, has released two i.MX93 system-on-modules (SoMs): the RNX-iMX93-OSM that follows the Open Standard Module (OSM) form factor, and the RNX-iMX93-SMARC compliant with the SMARC 2.1 (smart mobility architecture) standard.
RNX-iMX93-SMARC system-on-module
Ronetix i.MX93 SoMs are powered by the NXP i.MX93 processor featuring a 64-bit dual-core Arm Cortex-A55 application processor running at 1.7GHz and a Cortex-M33 core running at 250MHz for low-latency and real-time tasks. The SoC supports up to 2GB LPDDR4 RAM and 512GB eMMC storage, and the integrated Arm Ethos-U65 microNPU offers up to 0.5 TOPS of computing power for machine vision applications and intelligent energy management (IEM). Applicable for industrial automation and IoT, the SoMs also feature two CAN and Ethernet interfaces to aid high-speed data transfer.
The RNX-iMX93-SMARC system-on-module comes in 82 x 50mm SMARC form factor while the RNX-iMX93-OSM system-on-module comes in the solderable, 45 x 45mm, OSM-L form factor.
USB – 4x USB 2.0 (with USB hub), 2x USB 2.0 (without USB hub)
Up to 5x UART ports (OSM SoM only)
MMC/SD/SDIO
2x SPI, 2x I2C, 2x CAN
Up to 4x general purpose PWM signals (OSM SoM only)
GPIOs
Supply Voltage – 5V
Form Factor
OSM Size-L – 45x45mm
SMARC – 82x50mm
Temperature Range – 0°C to 70°C, -40°C to 80°C (optional)
Relative Humidity – 10% to 90%
Mean Time to Failure (MTTF) – >200000 hours
Software – Yocto, Linux Kernel, U-BOOT
RNX-iMX93-SMARC block diagram
The system-on-modules are provided with the Linux Yocto distribution, an open-source project for embedded development. Carrier boards, RNX-OSM-CARRIER and RNX-SMARC-CARRIER, are also available for evaluation and integration purposes.
RNX-iNX93-CARRIER board
The Ronetix RNX-iMX93-OSM module costs €120 (about $128) and the RNX-iMX93-SMARC module is priced at $135. The carrier boards are sold for $280 each. You can purchase the system-on-modules online from the shop and get more technical information on the wiki.
IBASE Technology ISR500 is a fanless digital signage player and Edge AI computer powered by either a MediaTek Genio 510 hexa-core Cortex-A78/A55 processor or Genio 720 octa-core Cortex-A78/A55 SoC with a real-time AI performance up to 3.2 TOPS.
The system ships with 4GB RAM and 64GB eMMC flash, features two HDMI video outputs, a gigabit Ethernet port, a few USB ports, a 3.5mm audio jack, a COM port, and an M.2 socket for wireless expansion.
IBASE ISR500 specifications:
SoC (one or the other)
IBASE Technology ISR500 is a fanless digital signage player and Edge AI computer powered by either a MediaTek Genio 510 hexa-core Cortex-A78/A55 processor or Genio 720 octa-core Cortex-A78/A55 SoC with a real-time AI performance up to 3.2 TOPS.
The system ships with 4GB RAM and 64GB eMMC flash, features two HDMI video outputs, a gigabit Ethernet port, a few USB ports, a 3.5mm audio jack, a COM port, and an M.2 socket for wireless expansion.
Hexa-core CPU – Dual-core Cortex-A78 @ 2.0GHz with 256KB L2 cache, quad-core Cortex-A55 @ 2.0GHz with 128KB L2 cache, shared 2MB L3 cache
GPU – Arm Mali-G57 MC2 with support for Vulkan 1.1, OpenGL ES 3.2, and OpenCL 2.2
AI accelerator – Tensilica VP6 MediaTek APU 3.0 up to 3.2 TOPS
VPU
Video Encode – 4Kp30 H.265/H.264
Video Decode – 4Kp60 AV1, VP9, H.265/H.264
System Memory – 4GB LPDDR4
Storage – 64GB eMMC flash (default)
Video Output
1x HDMI up to 4Kp60
1x HDMI up to 1080p60
Audio
3.5mm stereo audio jack
6-pin header for Audio Line-in and Line-out
Networking
Gigabit Ethernet RJ45 port
Optional WiFi and Bluetooth via M.2 module
USB – 1x USB 2.0 Type-A (with OTG support) port + 1x USB 3.1 Type-A port
Serial – RS232 via DB9 connector
Expansion
1x M.2 E-Key (2230) w/ SDIO, UART (for Wireless)
12-pin header (GPIO switch and LED)
Misc – Watchdog timer
Power Supply – 12V DC-in (40W or 60W adapter)
Dimensions – TBD (SGCC steel chassis)
Weight – 500 grams
Temperature Range – Operating: -10°C to 50°C; storage: -20°C to 80°C
Relative Humidity – 10%~90% (non-condensing)
Vibration – IBASE standard test
Certifications – CE/ FCC Class-B
IBASE offers DIN rail and wall mounting options and provides Android 13 and Yocto 5.15-based Linux images for the MediaTek Genio 510/700-powered Edge AI computer. Other OS are available upon request.
IBASE did not provide availability and pricing information for the ISR500 signage plater and Edge AI PC, but it should be up for sale already or very soon since the system was showcased at Embedded World 2024. A few extra details may also be found on the product page.
Přihlásit se
