AAEON’s UP WCL is an upcoming credit card-sized single board computer (SBC) powered by an Intel Wildcat Lake processor up to the Core 7 350 hexa-core CPU paired with up to 24GB LPDDR5 and 256GB UFS storage.

The board also features an HDMI 2.1 video output, a 2.5GbE RJ45 jack, an M.2 Key-E socket for optional WiFi and Bluetooth connectivity, three USB 3.2 ports, and a few wafers for I/O expansion.

AAEON UP WCL specifications:

• Wildcat Lake SoC (one or the other)
  • Intel Core 3 304
    • 5-core CPU – 1x P-cores @ 1.5/4.3 GHz (Turbo) + 4x LPE-cores @ 1.4/3.3 GHz (Turbo)
    • GPU – 1-core Intel Xe3 Graphics @ 2.3 GHz (9 TOPS)
    • NPU – 15 TOPS
  • Intel Core 5 320
    • 6-core CPU – 2x P-cores @ 1.5/4.6 GHz (Turbo) + 4x LPE-cores @ 1.4/3.4 GHz (Turbo)
    • GPU – 2-core Intel Xe3 Graphics @ 2.5 GHz (20 TOPS)
    • NPU – 16 TOPS
  • Intel Core 7 350
    • 6-core CPU – 2x P-cores @ 1.5/4.6 GHz (Turbo) + 4x LPE-cores @ 1.4/3.6 GHz (Turbo)
    • GPU – 2-core Intel Xe3 Graphics @ 2.6 GHz (21 TOPS)
    • NPU – 17 TOPS
  • Common features – 15W PBP, 6 MB Intel Smart Cache
• System Memory  – 8GB or 16GB LPDDR5 on default SKUs (up to 24GB LPDDR5)
• Storage – 64GB or 128GB UFS on default SKUs (up to 256GB UFS)
• Video Output – HDMI 2.1 port
• Networking
  • 2.5GbE RJ45 port
  • Optional WiFi and Bluetooth via M.2 Key-E socket
• USB
  • 3x USB 3.2 Gen 2 Type-A ports
  • 2x USB 2.0 via wafer (See Expansion section)
• Expansion
  • M.2 2230 E-Key socket
  • 10-pin wafer (3.3V) with 2x I2C, 2x PWM, 2x SPI
  • 10-pin wafer with 8x GPIO
  • 10-pin header with 2x USB 2.0, 1x UART
• Security – TPM 2.0 support (fTPM)
• Misc – RTC
• Power Supply
  • Input – 12V DC-in, 5A
  • Type – AT/ATX
  • Consumption (Typical) – 30~36W under load
• Dimensions 85 x 56mm
• Weight – 150 grams
• Temperature Range – -20°C to +60°C with active cooler
• Humidity – 0% ~ 90% relative humidity, non-condensing
• Certifications – CE/FCC Class A, RoHS Compliant, REACH

The six PCIe lanes of the Wildcat Lake SoC are used for 2.5Gbps Ethernet and the M.2 2230 E-Key socket, so it looks like there are still some spare ones that could not be used in the small board.

AAEON will provide support for Windows 11 LTSC and Ubuntu 24.04 LTS. The main benefits over previous generation credit card-sized UP boards like the UP TWL, are the high maximum memory capacity (24GB LPDDR5), faster UFS storage, HDMI 2.1 video output, and 40 TOPS of AI performance. Somehow, the 40-pin Raspberry Pi-compatible GPIO header has not made it to the new UP WCL.

The credit card-sized Wildcat Lake SBC will also power the UP WCL Edge embedded system/mini PC.  We had covered other Intel Core Series 3 Wildcat Lake industrial platforms such as the Advantech MIO-5356 3.5-inch SBC, as well as congatec conga-TC300 COM Express module, but the UP WCL is the first pro-hobbyist grade SBC to be announced.

Wildcat Lake availability

AAEON says the UP WCL and UP Nexus WCL (101.6 x 101.6mm, UP Squared Pro form factor) will be available in late Q3 2026, and in the meantime you can check out the UP developer boards page for preliminary information. I’ll cover the UP Nexus WCL in more detail once the board is launched.

This brings me to the actual availability of Wildcat Lake platforms. I’ve received emails about consumer-grade laptops or mini PC from companies such as CHUWI, Beelink, MINISFORUM, and others, but in many cases, they don’t have proper photos of the products or even a product page, let alone pricing information. Beelink has started contacting reviewers about test samples, so it’s a good sign that their mini PCs and NAS should soon become available.  One of the first Wildcat Lake platforms that can be purchased/pre-ordered today is the Dell XPS 13 (DX13260) laptop going for $699.99 or $899.99 with an Intel Core 5 320 SoC, 8GB or 16GB RAM, a 512GB SSD, and a 13.4-inch touchscreen display.

The post UP WCL – A credit card-sized Wildcat Lake SBC with up to 24GB LPDDR5, 256GB UFS appeared first on CNX Software - Embedded Systems News.

Broadcom has launched three new Wi-Fi 8 access point SoCs: the BCM6772 for mass-market Ethernet routers, extenders, and repeaters, the BCM6774 for high-volume Ethernet routers and extenders, and the BCM6776 for premium Ethernet tri-band routers and extenders. The latter is also part of a 5G FWA router platform based on Samsung B1320 5G Modem.

Broadcom’s first Wi-Fi 8 chips were introduced in May 2025 for clients and access points, and the company followed up with BCM6714 and BCM6719 dual-band radios earlier this year, and the BCM67142, BCM67192, and BCM68565 chips for low-cost Wi-Fi 8 10 Gbps fiber access points last month. So far, all the Broadcom Wi-Fi 8 chips enable multi-chip solutions with an application processor connected to one or more Wi-Fi 8 chipsets over a PCIe connection. The new BCM677x chips are fully-integrated SoCs with (Armv8>) application cores and 802.11bn connectivity in a single chip, except for 6 GHz Wi-Fi 8, as we’ll see further below.

Broadcom BCM6772 – 4-stream Wi-Fi 8 SoC

BCM6774 specifications:

• CPU – “Quad-core CPU complex”
• System Memory – 16-bit DDR4, DDR5
• Storage – Flash memory
• Wireless
  • Wi-Fi Standards
    • IEEE 802.11bn (Wi-Fi 8)
    • IEEE 802.11be (Wi-Fi 7)
    • IEEE 802.11a/b/g/n/ac/ax (Wi-Fi 6/6E/legacy)
  • 4x Spatial Streams – Dual Radio: 2×2 MIMO (2.4 GHz) + 2×2 MIMO (5 GHz)
  • Spectral Bands – 20/40 MHz (2.4 GHz), 20/40/80/160 MHz (5 GHz)
  • Modulation – Up to 4096-QAM (for UHR/Wi-Fi 8 operation)
  • Multi-Link Operation (MLO) supported on all bands
  • OFDMA and MU-MIMO supported in both uplink and downlink direction
  • Integrated 2.4 GHz Power Amplifiers (iPA)
• Wired networking
  • Multi-gigabit PHY
  • USXGMII (10GbE)
  • Dedicated network processing engine
• Expansion – USB controller/PHY (version/speed unspecified)
• Security
  • Hardware-accelerated WPA3 (including 192-bit mode)
  • Secure Boot with PQC (Post-Quantum Cryptography) support
• Package – 15×15 mm FCBGA
• Temperature Range – 0°C to 70°C (commercial temperature)

Broadcom BCM6774 – 6-stream Wi-Fi 8 SoC

BCM6774 specifications:

• CPU – “Quad-core CPU complex”
• System Memory – 16-bit DDR4, DDR5
• Storage – Flash memory
• Wireless
  • Wi-Fi Standards
    • IEEE 802.11bn (Wi-Fi 8)
    • IEEE 802.11be (Wi-Fi 7)
    • IEEE 802.11a/b/g/n/ac/ax (Wi-Fi 6/6E/legacy)
  • 6x Spatial Streams – Dual Radio: 2×2 MIMO (2.4 GHz) + 4×4 MIMO (5 GHz)
  • Spectral Bands – 20/40 MHz (2.4 GHz), 20/40/80/160 MHz (5 GHz)
  • Modulation – Up to 4096-QAM (for UHR/Wi-Fi 8 operation)
  • Multi-Link Operation (MLO) supported on all bands
  • OFDMA and MU-MIMO supported in both uplink and downlink direction
  • Integrated 2.4 GHz Power Amplifiers (iPA)
• Wired networking
  • Multi-gigabit PHY
  • USXGMII (10GbE)
  • Dedicated network processing engine
• Expansion – USB controller/PHY (version/speed unspecified)
• Security
  • Hardware-accelerated WPA3 (including 192-bit mode)
  • Secure Boot with PQC (Post-Quantum Cryptography) support
• Package – 15×15 mm FCBGA
• Temperature Range – 0°C to 70°C (commercial temperature)

The BCM6774 is virtually identical to the BCM6772 but adds two more 5 GHz spatial streams.

Broadcom BCM6776 – 6-stream Wi-Fi 8 SoC with PCIe Gen 3 for tri-band  support

BCM6776 specifications:

• CPU – “Quad-core CPU complex”
• System Memory – 16-bit DDR4, LPDDR4, DDR5 ,LPDDR5
• Storage – Flash memory
• Wireless
  • Wi-Fi Standards
    • IEEE 802.11bn (Wi-Fi 8)
    • IEEE 802.11be (Wi-Fi 7)
    • IEEE 802.11a/b/g/n/ac/ax (Wi-Fi 6/6E/legacy)
  • 6x Spatial Streams – Dual Radio: 2×2 MIMO (2.4 GHz) + 4×4 MIMO (5 GHz)
  • Spectral Bands – 20/40 MHz (2.4 GHz), 20/40/80/160 MHz (5 GHz)
  • Modulation – Up to 4096-QAM (for UHR/Wi-Fi 8 operation)
  • Multi-Link Operation (MLO) supported on all bands
  • OFDMA and MU-MIMO supported in both uplink and downlink direction
  • Integrated 2.4 GHz Power Amplifiers (iPA)
• Wired networking
  • Multi-gigabit PHY
  • USXGMII (10GbE)
  • Dedicated network processing engine
• Expansion
  • Dual PCIe Gen3 for tri-band Wi-Fi 8
  • USB controller/PHY (version/speed unspecified)
• Security
  • Hardware-accelerated WPA3 (including 192-bit mode)
  • Secure Boot with PQC (Post-Quantum Cryptography) support
• Package – 19×19 mm FCBGA
• Temperature Range – 0°C to 70°C (commercial temperature)

The BCM6776 further builds upon the BCM6774 and adds support for LPDDR4/LPDDR5 memory and a dual PCIe Gen3 interface to connect a BCM6718 for 6.0 GHz 4×4 MIMO WiFi 8 connectivity. No information about software was provided, but we can expect support for Linux-based operating systems like OpenWrt.

5G and Wi-Fi 8 FWA Platform

The Broadcom BCM6776 SoC is also part of the 5G and Wi-Fi 8 FWA (Fixed Wireless Access) router reference platform developed in collaboration with Samsung. Besides the tri-band WiFi 8 solution, it also features Samsung’s B1320 broadband-optimized 5G platform with the following highlights:

• CPU – 1.6 GHz quad-core ARM Cortex-A55 CPU
• Memory – LPDDR4x / LPDDR5x support
• Cellular
  • 3GPP Release 17
  • 4Rx/2Tx radio chain support
  • Power Class 1.5 support (TDD bands)
  • 5G NR-NTN and NB-NTN satellite support for n255 and n256 (L- and S-bands)
• GNSS
• Networking – 5 Gbps USXGMII
• Host interfaces – PCIe Gen 3, USB 2.0

HUMAX Networks is one of the first companies to develop a 5G CPE based on the solution.

Broadcom is currently sampling the BCM677x Wi-Fi 8 family to early access partners and customers, while the Samsung B1320 / Broadcom BCM6776 FWA platform is sampling to OEM partners. More details about the BCM677x family can be found on the Broadcom website. There’s no product page for the 5G FWA reference platform, and it was just introduced through a press release.

Thanks to TLS for the tip.

The post Broadcom unveils fully-integrated quad-core Wi-Fi 8 access point SoCs, 5G FWA router reference platform appeared first on CNX Software - Embedded Systems News.

HS89 T15 4K Android TV stick is powered by the new Rockchip RK3539 quad-core Cortex-A55 SoC with AV1 and H.265 video codecs and HDR10 support, and is equipped with up to 4GB of RAM and a 32GB eMMC flash.

The device features a male HDMI 2.1 port, WiFi 6 and Bluetooth 5.4 connectivity, a USB Type-A host port, and a USB-C port for power only, either directly from the TV or through a 5V/2A power adapter. The device also ships with a Bluetooth voice remote control. Let’s have a look at both the streaming device and the RK3539 chip itself since the datasheet is available.

HS89 T15 TV stick

Let’s start with the HS89 T15 TV stick’s specifications:

• SoC – Rockchip RK3539
  • CPU – Quad-core Cortex-A55 @ 1920 MHz
  • GPU – Mali-310
  • VPU – 4Kp60 AV1, H.265, H.264 video decoding
• System Memory/Storage options
  • 2GB RAM, 16GB eMMC flash
  • 4GB RAM, 32GB eMMC flash
• Video Output – HDMI 2.1 port up to 4Kp60 with HDR10 support
• Audio – Digital audio via HDMI
• Connectivity – WiFi 6 and Bluetooth 5.4
• USB – 1x USB Type-A port
• Misc – 3.5mm jack for IR extender
• Power Supply – 5V DC/2A via USB-C port
• Dimensions – 100 x 33 x 13 mm
• Weight – TBD

The TV stick ships with an HDMI cable, a Bluetooth and IR remote control with voice control support, and a user manual. It runs Android 14 with Google Play Store support, but there’s no mention of DRM certifications such as Widevine L1, so it may not be for everyone.

The HS89 T15 TV stick is sold on AliExpress for $33.72 in its 2GB/16GB configuration, while the 4GB/32GB model goes for $44.96. Note that it doesn’t ship with a 5V/2A power supply, and you may need one if your TV’s USB port doesn’t output enough power. The IR extender doesn’t appear to be included either. If you are not into the TV stick form factor, VONSTAR HS89 PRO 15 TV box sells for about $30 (2GB/16GB) or $40 (4GB/32GB) on AliExpress.

Rockchip RK3539 SoC for smart IPTV/OTT and high-end multimedia applications

Rockchip RK3539 specifications:

• CPU – Quad-core Arm Cortex-A55 processor with NEON, FPU, ARMv8 Crypto…
• Cache
  • 32KB L1 instruction cache
  • 32KB L1 data cache and 64KB L2 data cache
  • 512KB unified system L3 cache
• GPU
  • Arm Mali-G310 3D GPU with support for OpenCL 3.0, OpenGL ES1.1/2.0/3.2, Vulkan 1.2
  • 2D Graphics Engine
• VPU
  • Decoder
    • H.265, H.264, VP9 up to 4096 x 2160 @ 60 FPS
    • AV1 up to 3840 x 2160 @ 60 FPS
    • VP8, VC1, MPEG-4, MPEG-2, MPEG-1 up to 1920×1088 @ 60 FPS (1088 is not a typo)
    • H.263 up to 720p60
    • (M)JPEG up to 8176×8176 @ 76 million pixels per second
  • Encoder – N/A
• MCU core – RISC-V MCU in PMU domain with dedicated mailbox and watchdog.
• Memory – 32-bit DDR3(L)/LPDDR3/DDR4/LPDDR4(4X) up to 4GB
• Storage – eMMC 5.1, SD 3.0/MMC 4.51, FSPI (Flexible Serial Flash Interface), NANDC interface
• Video Output
  • HDMI 2.1 up to 4Kp60 10-bit, with HDCP 2.3, HDR10+ support
  • CVBS/PAL TV interface
• Audio
  • 24-bit audio DAC
  • 2x 2-channel SAI (Serial Audio Interface), 1x 8-channel SAI; up to 192 kHz sample rate
  • 8-channel PDM up to 192 kHz sample rate
  • S/PDIF output
• Networking
  • 10/100 Mbps Ethernet controller and MAC PHY
  • 10/100/1000 Mbps Ethernet MAC with RMGII/RMII interfaces
• USB
  • USB 2.0 DRD
  • USB 2.0 Host
• Other peripheral interfaces
  • GPIO
  • SDIO 3.0
  • 6x UART, 2x SPI, 6x I2C
  • 8x PWM
  • 4-channel DMAC
  • Analog
    • 13-bit SARADC up to 2 MS/s
    • Temperature sensor (TS-ADC)
• Timers
  • 2x 64-bit secure timers with interrupt-based operation
  • 6x 64-bit non-secure timers with interrupt-based operation
  • 1x high precision timer
  • Watchdog timer
• Security
  • AES, DES, 3DES, SM4 symmetrical algorithms
  • RSA, ECC, SM2 asymmetrical algorithms
  • SHA-1, SHA-256/224, SHA-512/384, SHA-512, MD5, SM3 hash algorithms
  • Key ladder
  • Security TRNG and non-security TRNG
  • Secure OTP
  • Secure debug
  • Secure OS…
• Package – FBGA434L (body: 13.3mm x 13.5mm ;ball size: 0.3mm ;ball pitch: 0.65/0.6mm)
• Temperature Range – Junction: up to 125°C; storage: -40 to +125°C

Unsurprisingly, the specifications are very similar to those of the Rockchip RK3538 SoC we covered a few months ago. The main differences are support for 4K video output and decoding, a Gigabit Ethernet MAC, and support for up to 4GB of RAM (vs 2GB) for the RK3539 processor. Despite the differences, the RK3538 and RK3539 have the same number of balls (434) and dimensions. You’ll find more details in the datasheet on the Rockchip.fr website.

Via AndroidTVBox.eu

The post Rockchip RK3539 quad-core Cortex-A55 SoC shows up in low-cost Android 14 4K TV stick with AV1 codec support appeared first on CNX Software - Embedded Systems News.

Last year, we noted three upcoming high-performance RISC-V SoCs to watch out for: Zhihe A210, SpacemiT K3, and UltraRISC UR-DP1000. The K3 has already been launched, and I’ll work on the K3-Pico-ITX SBC/mini PC review this coming weekend, while the UR-DP1000 is (still?) expected on the Milk-V Titan motherboard. However, we did not have that many details about the Zhihe A210 so far.

This has now changed since documentation has surfaced for the Zhihe A210 and a development kit (A210 SODIMM V2) based on a carrier board and a system-on-module itself powered by the octa-core RISC-V processor. Let’s have a look at both.

Zhihe A210 octa-core RISC-V SoC

Zhihe A210 specifications:

• CPU  – Octa-core RISC-V RV64GCV processor
  • 4x 64-bit RISC-V C920 cores @ up to 2.3 GHz with 64 KB I Cache and 64 KB D Cache for each core, 1 MB L2 cache; note: 1.9 GHz is also shown for this cluster, depending on where we look in the docs.
  • 4x 64-bit RISC-V C908 cores @ up to 1.9 GHz with 32 KB I Cache and 32 KB D Cache for each core, 512 KB L2 cache
  • RVA23 Profile compliant
• GPU
  • Unnamed 3D GPU, potentially IMG BXM-4-64
    • Up to 50.34 GFLOPS
    • 3D graphics acceleration engine: 3.14 GPixel @ 786MHz
    • Support for Vulkan 1.1/1.2, OpenCL 1.1/1.2/2.0, and OpenGL ES 1.1/2.0/3.0/3.1/3.2
  • 2D graphics accelerator
• VPU
  • Decoding
    • Up to 4Kp120 H.265/HEVC, H.264/AVC, AV1
    • Up to 4Kp60 VP9, AVS2
    • Up to 1080p60 VP6/7/8/AVS/AVS+/VC1/MPEG4
    • (M)JPEG decoding up to 1080p @ 576 FPS
  • Encoding
    • 4Kp60 H.265/HEC, H.264/AVC
    • (M)JPEG encoding up to 1080p @ 312 FPS
• ISP – Up to 12MP resolution
• AI accelerator
  • 12 TOPS (INT8) neural processing unit with supports for INT4, INT8, INT16, FP8, FP16 and BF16
  • DeepSeek-7B can reach 8 tokens/s; “4-die multi-chip cascading DeepSeek-7B can reach 25 tokens/s”
  • Supported frameworks: TensorFlow, Caffe, HuggingFace, ONNX, etc.
• Memory – Up to 16GB; 2x 32-bit LPDDR4/LPDDR4X up to 4266 MT/s
• Storage
  • eMMC 5.1
  • SPI NOR/NAND flash
  • SD 3.0
  • NVMe or SATA interface via PCIe 3.1/SATA 3.0 combo interface (see PCIe section below for details)
• Display Interfaces
  • HDMI 1.4/2.0 up to 4Kp60 with HDCP 1.4 support
  • 4-lane MIPI DSI up to 4Kp60
  • DisplayPort up to 4Kp60 via USB3.1/DP 1.4/eDP1.5 combo interface (see USB section for details)
• Camera Interface – 4-lane MIPI CSI2 D-PHY interface, with flexible lane configurations of 2x 4-lane, 4x 2-lane, 1x 4-lane + 2x 2-lane
• Audio
  • 8-lane I2S/PCM interface (Rx/Tx); up to 384kHz sample rate, 16-/24-/32-bit resolution
  • 3x 2-lane I2S/PCM interface (Tx/Tx); up to 384kHz sample rate, 16-/24-/32-bit resolution
  • PDM interface for up to 8 channels; up to 192kHz sample rate, 16-/24-bit resolution
  •  TDM/PCM interface up to 8 channels;  up to 192kHz sample rate, 16-/24-/32-bit resolution
• Networking – 2x GMAC ports with RGMII/RMII support
• USB
  • USB3.1/DP 1.4/eDP1.5 combo interface (host or device)
    • USB 3.1 output via a Type-C interface
    • USB 3.1 + 2-lane DP via a Type-C interface
    • USB 2.0 + 4-lane DP output via a Type-C interface
    • USB 3.1 output via a Type-A interface + 2-lane DP/eDP interface
    • USB 2.0 output via a Type-A/C interface +4-lane DP/eDP interface
  • 2x USB 2.0 host/device interfaces
• PCIe – PCIe 3.1/SATA 3.0 combo interface configurable as follows:
  • 1x PCIe x4 RC/EP
  • 1x PCIe x2 RC/EP+ 1x PCIe x1 RC
  • 1x PCIe x2 RC/EP + 2x SATA
  • 1x PCIe x1 RC/EP + 1x PCIe  x1 RC + 2x SATA
• Other Peripherals
  • Analog I/O – 4-channel ADC
  • 10x UART ports
  • 10x I2C ports
  • 2x QSPI ports and 2x SPI ports
  • 3x 6-channel PWM ports
  • 3x CAN interfaces, supporting CAN-FD
  • Integrated JTAG, GPIO, etc.
• Security
  • Dual-layer security: TEE+REE
  • Encryption algorithms
    • Symmetrical algorithms: AES, TDES, DES, SM4
    • Asymmetrical algorithms: RSA1024/2048/3072/4096, SM2
    • HASH algorithms: SHA-1/224/256/384/512, SM3, MD5
  • Hardware random number generator (HRNG)
  • Hardware-enforced isolation
  • Secure boot
• Power Management Unit – RISC-V E902 core @ 100 MHz with RTC, POR, Watchdog…
• Operating voltage
  • Voltage for core: 0.8V
  • Voltage for IO: 1.8V
  • Voltage for LPDDR4 IO: 1.1V
  • Voltage for LPDDR4x IO: 0.6V
• Package – 25×25 mm; 1373-ball FC-BGA

A210 SODIMM V2 Development Board

A210 SODIMM V2 board specifications:

• Core board
  • SoC – Zhihe A210 octa-core 64-bit RISC-V SoC as described above
  • System Memory – 4GB, 8GB, or 16GB LPDDR4x up to 4266Mbps
  • Storage – eMMC flash
  • Networking – 2x Gigabit Ethernet PHY
  • Host interface – 260-pin SO-DIMM edge connector exposing the A210’s I/Os
  • Power
    • 5V DC input voltage
    • 2x PMIC chips
  • Dimensions – 69.94 x 59.66mm
• Carrier board (V2)
  • Storage – M.2 Key-B SATA port (
  • Video Output
    • HDMI 2.0 port
    • MIPI DSI connector
    • DisplayPort via USB-C port
  • Camera
    • 2-lane MIPI-CSI Board-to-Board (B2B) connector
    • 4-lane MIPI-CSI Board-to-Board (B2B) connector
  • Audio – 3.5mm audio jack
  • Networking
    • 2x Gigabit Ethernet RJ45 ports
    • Wi-Fi and Bluetooth V5.0 /4.2 support.
  • USB
    • USB 3.1 Type-C port with DP Alt mode support
    • USB 2.0 Type-A port for flashing
    • USB 2.0 Type-A port
  • Expansion
    • Mini PCIe socket (PCIe/USB)
    • M.2 Key-B SATA port
    • 20-pin 2.54mm pitch QSPI and GPIO header
    • 20-pin 2.54mm pitch UART and GPIO header
  • Misc
    • Power, Reset, and Flash buttons
    • 4-pin Debug UART header
    • PCIe RC/EP mode switch
    • 5V fan header
  • Power Supply – 12V DC
  • Dimensions – TBD

The system-on-module and development board are supported by a Linux SDK with buildroot and Debian images. The documentation has many more details about the hardware and basic instructions to get started with the SDK using a Docker image, as well as multimedia (video decode/encode) tutorials, and steps to use the TORQ-Toolkit to perform model conversion and deploy the AI models to the chip.

The development kit is sold for $334.33 on AliExpress as the “Sipeed ZHIHE A210 RISC-V development platform” in 8GB/64GB configuration, but note that the seller (chipboard store) often marks up the price a fair bit. If it’s indeed from Sipeed, it could soon be sold for about $200 or less.

The post Zhihe A210 octa-core RISC-V SoC with 12 TOPS NPU powers SoM-based development board appeared first on CNX Software - Embedded Systems News.

Amlogic has introduced the A311Y3 Edge AI SoC, which seems to be an upgrade over earlier designs such as the A311D and A311D2 used in boards like the Khadas VIM3 and Khadas VIM4.

Although not listed on the official Amlogic website, documentation from vendors such as Shenzhen Tomato Technology indicates that the A311Y3 SoC is built around a 6nm process node and features a newer CPU architecture based on Arm Cortex-A78/A55 cores. The SoC integrates an NPU with up to 8 TOPS of compute performance and supports LPDDR5/LPDDR5X memory, which can be used in both processing and Edge AI, compared to earlier A311-series chips.

Amlogic A311Y3 specifications

• CPU
  • 2x Arm Cortex-A78 cores (64KB L1 I-cache, 64KB L1 D-cache, 256KB L2 cache)
  • 6x Arm Cortex-A55 cores (32KB L1 I-cache, 32KB L1 D-cache, 128KB L2 cache) with a shared 2MB L3 cache
  • Neon and Crypto extensions included
• Co-processors – RISC-V core for system control processing; Built-in RISC-V core for Always-on power management and Sensor Hub
• GPU
  • ARM Mali-G625 MC1 GPU with OpenGL ES 3.2, Vulkan 1.4, and OpenCL 3.0
  • 2.5D graphics processor (VICP + GE2D) bitblt engine supporting up to 16 image compositions and 4096 x 4096 output resolution
• VPU (Video Engine – AVE-10)
  • Decode – up to 4K @ 120 fps (AV1, H.265, VP9, AVS3, etc.)
  • Encode
    • H.265 / H.264 up to 4K @ 60fps
    • MJPEG up to 4K @ 30fps
    • Multi-stream decoding (e.g., 2×4K@60 or 4×1080p@60)
  • 9th Gen Advanced Amlogic TruLife Image Engine with inline AI-SR (AI Super Resolution), Dolby Vision (Optional), HDR10, HDR10+, HLG, and Vivid HDR
• NPU
  • 8 TOPS Amlogic Deep Learning Accelerator (ADLA) supporting INT4, INT8, INT16, FP8, FP16, and BF16 precision
  • Works with TensorFlow, TensorFlow Lite, ONNX, PyTorch, Darknet, MxNet, and Caffe
• Memory – Up to 16GB LPDDR5/LPDDR5X (up to 6400 Mbps) or LPDDR4/LPDDR4X (up to 4266 Mbps)
• Storage
  • eMMC 5.1/5.2 interface (400Mbps) with Hardware Command Queue (HW CQ) and in-line encryption
  • SLC/SPI/QSPI NAND Flash with 8-bit ECC
  • SDSC/SDHC/SDXC and SDIO interface (1-bit/4-bit, up to UHS-I SDR104)
• Video Output
  • 1x HDMI 2.1a up to 4K @ 60Hz with eARC, ALLM, VRR, QMS, QFT, SBTM, and HDCP 1.4/2.3
  • 1x DisplayPort (DP) 1.4 up to 4K @ 60Hz with MST and DP Alt Mode via USB Type-C
  • 1x eDP 1.3/1.4 TX up to 4 lanes, 4K @ 60Hz with PSR/PSR2 support
  • 2x 4-lane MIPI-DSI up to 2.5 Gbps per lane (up to 2560 x 1600 @ 60Hz or 1920 x 1080 @ 120Hz)
  • 1x 8-lane V-by-One TX up to 4K @ 60Hz
  • 1x LVDS TX up to 1080p @ 60Hz
  • Dual independent displays supported (up to 4K + 4K duplicate mode or 4K + 2.5K extended mode)
• Video/Camera Input
  • 1x HDMI 2.1a input up to 4K @ 60Hz with eARC, ALLM, VRR, QMS, QFT, SBTM, and HDCP 1.4/2.3
  • 2x 4-lane MIPI-CSI interfaces (2.5 Gbps per lane), supporting up to 4 virtual channels per port (RAW, YUV, or RGB formats)
  • Integrated ISP wtih up to 13MP @ 30fps or 16MP @ 20fps (WDR); features 3A, PDAF, 2D/3D NR, and CAC
  • Dewarp Unit supporting panoramic, spherical, and perspective dynamic corrections up to 16MP
  • Computer Vision Engine (CVE) for background modeling, motion target detection, face tracking, etc
• Audio
  • 1x SPDIF input, 2x SPDIF output ports
  • 3x TDM/I2S input and 3x TDM/I2S output ports (up to 16 channels, 16/24/32-bit, up to 192kHz)
  • 8-channel PDM input supporting up to 8 digital microphones (DMICs)
  • Sound Event Detection (SED) and Asynchronous Sample Rate Converter (ASRC)
  • Optional HiFi 5 Audio DSP with single/half-precision vector FP and neural network extensions
• Networking
  • 2.5Gbps Ethernet MAC via HSGMII interface
  • 1Gbps Ethernet MAC via RGMII interface
  • Wireless network support expansion via PCIe, SDIO, USB, UART, or PCM (for WiFi/Bluetooth modules)
• USB
  • 1x USB 3.0 DRD (Dual Role Device, up to 5Gbps, combo with DP 1.4 TX)
  • 1x USB 3.0 Host (combo with PCIe 2.0 and HSGMII)
  • 1x USB 2.0 DRD
  • 1x USB 2.0 Host
• Expansion
  • 1x PCIe 3.0 x2 (supports lane bifurcation to 1×2 lane or 2×1 lane; Root Complex and Endpoint modes)
  • 1x PCIe 2.0 x1 (Root Complex mode, combo with USB 3.0 Host and HSGMII)
• Other I/O
  • Digital TV 2x serial or 1x parallel + 1x serial Transport Stream (TS) inputs with demux; ISO7816 smart card controller
  • 9x I2C, 2x I3C (supporting SDR, HDR-DDR, and HDR-BT modes)
  • 6x UART (up to 4Mbps, 4-wire with hardware flow control)
  • 14x PWM
  • 6x SPI (master/slave mode)
  • 1x CAN FD (FlexCAN) protocol specification compliant with ISO 11898-1:2015
  • IR interface (1x RX, 1x TX)
  • 5-channel 10-bit SAR ADC
• Security
  • Arm TrustZone (TEE)
  • Secure boot, encrypted DRAM, encrypted OTP
  • AES, RSA (up to 8K), ECC
  • Post-quantum crypto (ML-DSA / Dilithium)
  • TRNG, SHA-1/2, HMAC
• Power
  • Multi-voltage I/O design (1.8V/3.3V)
  • Multiple software-controlled power domains and DVFS support
  • Always-On (AO) domain contains dedicated I/Os (4x UART, 4x I2C, 1x I3C, 3x SPI, 7x PWM, 1x CEC, IR, PDM) to interface with an external PMIC.
• Package – TBD

After going through the specifications, it’s clear that compared to earlier Amlogic SoCs, the A311Y3 uses a newer CPU configuration with up to 2× Cortex-A78 + 6× Cortex-A55, instead of the Cortex-A73/A53 setup used in the Amlogic A311D and Amlogic A311D2. It is also built on a 6nm process. The NPU performance increases to about 8 TOPS, compared to 5 TOPS (A311D) and 3.2 TOPS (A311D2). The platform supports up to 16GB LPDDR5/LPDDR5X memory. The I/Os are also enhanced with 2.5Gbps Ethernet, PCIe 3.0, dual CAN FD, and hardware support for post-quantum cryptography.

According to SZTomato, the A311Y3 supports Android 16 (64-bit) based on Linux Kernel 6.12, along with standard GNU/GCC development tools. On Linux, initial mainline kernel support has been submitted by an Amlogic engineer with patches adding device tree bindings for the Amlogic A9 family. These patches currently provide support for the CPU, GIC, interrupts, timers, and UART, with broader peripheral enablement expected as the series progresses through upstream review. While searching, I also found a basic device tree for the BY401 development board, which appears to be an internal Amlogic reference board for the A311Y3 SoC.

While searching for more details, I came across the dev board from 9tripod, which features the A311Y3 SoC octa-core CPU and supports up to 16GB of LPDDR5/LPDDR5X RAM, 4K video encoding/decoding, and multiple display outputs, including HDMI 2.1, DP, and MIPI-DSI. The board also offers various connectivity options, including PCIe 3.0, 2.5G Ethernet, USB 3.0, and various industrial interfaces like CAN FD, I2C, SPI, and UART, making it suitable for embedded and AI applications.

The specifications for the Amlogic A311Y3 SoC were derived from the SZTomato website, and they also published a separate article explaining why the new chip is suitable for Smart Retail and commercial AIoT devices, and comparing the A311Y3 to RK3572.

The post Amlogic A311Y3 octa-core Edge AI SoC features Cortex-A78/A55 cores, 8 TOPS NPU, LPDDR5 support appeared first on CNX Software - Embedded Systems News.

SCINTIX P4 is an ESP32-P4 RISC-V Compute Module with an ESP32-C6 for wireless connectivity that’s compatible with Raspberry Pi CM4/CM5 carrier boards, at least partially.

It should be the first MCU-based system-on-module in Raspberry Pi CM4/CM5 form factor, and RELOC says the SCINTIX P4 gives access to displays, cameras, Ethernet, USB, and all the peripherals the ESP32-P4 exposes when connected to a carrier board. It can also be programmed in standalone mode through its built-in USB Type-C port.

SCINTIX P4 (RM-CMP4) specifications:

• SoC – Espressif Systems ESP32-P4NRW32X
  •  CPU
    • Dual-core RISC-V @ 400 MHz with AI instruction extensions and single-precision FPU
    • Single-core RISC-V LP (low-power) MCU @ up to 40 MHz
  • GPU – 2D Pixel Processing Accelerator (PPA)
  • VPU – H.264 and JPEG codecs support
  • Memory – 768 KB HP L2MEM, 32 KB LP SRAM, 8 KB TCM, 32MB PSRAM
  • Storage – 128 KB HP ROM, 16 KB LP ROM
• Storage – 32MB SPI NOR flash
• Networking
  • Microchip KSZ8091RNACA Fast Ethernet transceiver
  • 2.4 GHz Wi-Fi 6, Bluetooth 5, and 802.15.4 (Zigbee / Thread / Matter) via ESP32-C6 module connected via  SDIO + UART to the ESP32-P4
• USB – 1x USB Type-C port for power and programming
• 2x 100-pin Hirose DF40C-100DO-0.4V(51) B2B connectors compatible with Raspberry Pi CM4/CM5 carrier boards
  • Display – 2-lane MIPI DSI
  • Camera – 2-lane MIPI CSI
  • Networking – 10/100Mps Ethernet
  • USB – USB 2.0 OTG
  • I/Os – 40x GPIO
• Power Supply  – 5V DC input via USB-C port or B2B connectors
• Dimensions – 55.0 × 40.0 × 5.0 mm; 8 mm thick including 1.27 mm programming connectors
• Weight – 8.2 grams
• Temperature Range – -20 to 70°C

While it looks neat, in practise, you’d lose HDMI and PCIe interfaces, USB is limited to 2.0 speed (480 Mbps), and Ethernet to 100 Mbps. It still exposes MIPI CSI/DSI camera and display interfaces and GPIOs. In our article about the ESP32P4C5 Core Board with two Raspberry Pi-compatible MIPI connectors, we noted that ESP32-P4 could only support the original Raspberry Pi Camera Module (OV5647), but not the more recent Sony IMX-based modules, and the situation is better for the official Raspberry Pi displays, although this requires an out-of-tree library, at least for now.

On the software side, it will obviously not run Raspberry Pi OS, and instead, RELOC mentions full compatibility with Espressif’s ESP-IDF SDK and demos,  and Arduino and PlatformIO support out of the box. As an upcoming crowdfunding campaign, nothing has been released yet, but RELOC promises schematics, Bill of Materials, pinout diagrams, 3D mechanical models, firmware examples covering the main features of the board (Wi-Fi / BLE, Ethernet, display, camera, etc.), a complete demo project showcasing the full board capabilities, and component libraries compatible with CAD tools such as KiCad. There’s no video demo of the solution either for now.

RELOC has set up a page on Crowd Supply for a future launch, but I also noticed the product page on the RELOC website says that the SCINTIX P4 is available and to contact them for pricing.

The post SCINTIX P4 ESP32-P4 Compute Module works with (some) Raspberry Pi CM4/CM5 carrier boards (Crowdfunding) appeared first on CNX Software - Embedded Systems News.

Last month, AMD introduced the Ryzen AI Halo Developer Platform based on the Ryzen AI Max+ 395 processor, delivering up to 126 TOPS, and pre-orders are now available for $3,999.99 at MicroCenter (US only, in-store pickup).

The computer ships with 128GB LPDDR5x memory and a 2TB NVMe SSD. It offers 10GbE and WiFi 7 networking, HDMI 2.1 and USB-C DisplayPort video output, and a few extra USB-C ports

Ryzen AI Halo developer kit specifications:

• SoC – AMD Ryzen AI Max+ 395
  • CPU – 16 cores, 32 threads, “Zen 5” Architecture clocked up to 5.1 GHz
  • GPU – AMD Radeon 8060S Integrated Graphics with 40 CUs, RDNA 3.5 architecture
  • NPU – AMD XDNA 2 NPU
  • TDP – 120W
  • Process – TSMC 4nm FinFET
• Memory – 128GB LPDDR5x up to 8,000 MT/s speed, 256GB/s bandwidth
• Storage – 2TB M.2 SSD
• Video Output
  • HDMI 2.1b
  • DisplayPort via USB-C port
• Networking
  • 10GbE RJ45 port
  • WiFi 7 and Bluetooth 5.4
• USB
  • 3x USB-C ports, including one with DP Alt mode support
  • 1x USB-C for power input
• Misc – Power button
• Power Supply – USB PD via USB-C port
• Dimensions – 150 x 150 x 45.4 mm
• Weight – 1.2 kg

The system can run Windows 11 or Linux to run AI workloads, and the company points to the AI Developer Program and “AMD AI Playbooks” to easily get started with step-by-step guides to run AI workloads on AMD hardware, from inference to fine-tuning. However, everything in the AI Playbooks is shown as coming soon, and we should expect the documentation to be ready once the development platform ships.

The 126 TOPS (INT8) Ryzen AI Halo Developer Platform appears to mainly compete against the 1,000 TOPS (FP4) NVIDIA DGX Spark introduced last year.  Both are designed for local AI workloads, and AMD shared a few (advantageous) benchmark results where they pitted the two platforms against each other for a range of AI models: GLM 7.4 Flash-30B-A3B, GPT-OSS-120B, Qwen 3.5-122B-A10B, and Qwen 3.6-35B-A3B.

The AMD AI computer delivers 4% to 14% extra performance in these specific benchmarks. The company also highlights support for both Windows and Linux operating systems, better token/sec/dollar metrics, and the presence of a 50 TOPS NPU, not that it matters that much.

For some reason, the company also compared the AMD Ryzen AI Halo to the Apple M4 Pro, and it’s 3.3 to 7.3 times faster on a range of workloads.

It’s unclear why AMD decided to launch its own Developer Platform, since several of its customers already provide Ryzen AI Max+ 395 systems with similar specs and a slightly lower price point on Amazon and AliExpress. For instance, the GMKtec EVO-X2 with 128GB LPDDR5, 2TB NVMe SSD, four display outputs, and 2.5GbE (no 10GbE) and WiFi 7 goes for $3,299.99.

A more powerful model based on the 133 TOPS AMD Ryzen AI Max+ PRO 495 processor, and paired with 192GB of memory, is also planned. More details can be found on the product page.

The post $4,000 AMD Ryzen AI Halo Developer Platform features 126 TOPS Ryzen AI Max+ 395 processor appeared first on CNX Software - Embedded Systems News.

ADLINK COM-HPC-mPTL is a COM-HPC R1.3 Mini computer-on-module powered by the Intel Core Ultra Processor Series 3 “Panther Lake” family, up to the 180 TOPS 16-core Core Ultra X7 358H processor.

The module supports up to 64GB LPDDR5x and optional NVMe SSD BGA storage, and features two 2.5GbE controllers, optional MIPI CSI camera connectors,  and a 40-pin  debug connector. All I/Os are exposed through a standard 400-pin high-density board-to-board connector, including four display interfaces (HDMI, DP, USB4, eDP), and up to 16x PCIe Gen4/Gen5 lanes.

COM-HPC-mPTL specifications:

• SoC – Intel Core Ultra Series 3 Panther Lake-H processors (one or the other)
  • Intel Core Ultra 5 325 8-core (4P+0E+4LPE) processor up to 2.1 GHz / 4.5 GHz (Turbo) with 12MB cache, 4-core Xe3 graphics (40 TOPS), 47 TOPS NPU; PBP: 25W
  • Intel Core Ultra 7 356H 16-core (4P+8E+4LPE) processor up to 1.9 GHz / 4.7 GHz (Turbo) with 18MB cache, 4-core Xe3 graphics (40 TOPS), 50 TOPS NPU; PBP: 25W
  • Intel Core Ultra X7 358H 16-core (4P+8E+4LPE) processor up to 1.9 GHz / 4.8 GHz (Turbo) with 18MB cache, 12-core Xe3 graphics (Arc B390, 122 TOPS), 50 TOPS NPU; PBP: 25W
• System Memory – Up to 64 GB (4x 16GB) LPDDR5x up to 7467 or 8533 MT/s depending on SKU
• Storage – Onboard NVMe SSD (optional co-lay with PCIe lane 4 & 5 w/ PCIex2 Gen5)
• Camera – 2x MIPI-CSI FFC connectors
• Networking – 2x Intel I226 2.5 Gbps Ethernet controllers with support for TSN by build option
• Host connectors – 400-pin high-density connector
  • Display Interfaces
    • 2x DDI ports (DDI 0/1) for DP 1.4a or HDMI 2.1 (Default DDI0)
    • 3x USB4 ports with support for DP alt mode
    • eDP 1.4b
    • Up to 4x independent displays
  • Audio – Realtek codec on the carrier board
  • Networking – 2x 2.5GbE
  • USB
    • 3x USB4 ports with DisplayPort Alt mode
    • 4x USB 3.2 Gen2x1 (10 Gbps)
    • 4x USB 2.0
  • Serial – 2x UART ports with console redirection
  • Expansion
    • Up to 16x PCIe lanes
      • 8x PCIe Gen5 lanes (8-15): x8, x4 configurations
      • 4x PCIe Gen4 lanes (0-3): x4, x2, x1 configurations
      • 2x PCIe Gen4 lanes (4-5): x2, x1 configurations
      • 2x PCIe Gen5 lanes (6-7): x2 (lane 6-7), x1(only lane 6) configurations
      • Note: Lanes 4-7 can be combined as PCIe Gen5 x4. (available w/o onboard NVMe SSD function).
    • SMBus (system), 2x I2C (user), 1x GP_SPI, 1x Boot_SPI and eSP
    • 12x GPIO (GPI with interrupt)
• Debug Headers – 40-pin FFC/FPC connector, compatible with the DB40-HPC debug module; used for BIOS POST code LEDs, EC access, SPI BIOS flashing, power test points, and debug LEDs
• Security – Infineon TPM 2.0 (SPI)
• Misc
  • Embedded BIOS AMI UEFI with CMOS backup in 32 MB SPI BIOS (dual BIOS optional)
  • SEMA Board Controller for manageability, voltage/current monitoring, power sequence debug support,  logistics & forensic information (event/history), general-purpose interfaces (I²C, UART, GPIO), watchdog timer, and fan control
• Power
  • Standard Input – AT mode: 12V±5%; Vin: 8-20V (single power rail)
  • Wide Input – AT: 8.5-20V
  • Power States – C0-C6, S0, S3, S5; ECO mode (Wake on USB S3/S4, WOL S3/S4/S5) (TBC)
• Dimensions – 95 x 70 mm (COM-HPC Mini Rev 1.3)
• Temperature Range
  • Standard: 0°C to 60°C (storage: -20°C to 80°C)
  • Industrial: -40°C to 85°C (storage: -40°C to 85°C)
• Humidity
  • 5-90% RH operating, non-condensing
  • 5-95% RH storage, non-condensing
• Shock and Vibration
  • IEC 60068-2-64 and IEC-60068-2-27
  • MIL-STD-202F, Method 213B, Table 213-I, Condition A and Method 214A, Table 214-I, Condition D

ADLINK provides standard support for Windows 11 IoT Enterprise LTSC and Ubuntu 24.04 LTS, as well as VxWorks upon request. The company also provides the COM-HPC Mini base ATX carrier board for evaluation with two Ethernet ports, DP and eDP connectors, PCIe x8 and x4 slots, audio jacks, a range of USB ports, and more.

The COM-HPC-mPTL is the first COM-HPC (Mini) computer-on-module we’ve covered here with an Intel Core Ultra Series 3 “Panther Lake” processor, but congatec is also working on the conga-HPC/mPTL with similar features. The COM-HPC Mini module can also be seen as a more compact variant of the ADLINK Express-PTL COM Express Type 6 module that takes DDR5 DO-DIMM memory, instead of soldered-on LPDDR5 memory.

ADLINK didn’t provide availability and pricing information for the COM-HPC-mPTL computer-on-module, and for reference, the product page is still shown as “preliminary”.

The post ADLINK COM-HPC-mPTL COM-HPC Mini computer-on-module features up to 180 TOPS Intel Core Ultra Series 3 CPU appeared first on CNX Software - Embedded Systems News.

While writing about the recent Linux 7.1 release, I came across the “OneThing Edge Cube” (OEC) based on the Rockchip RK3566 SoC. I assume it may have been some type of NAS, but instead it’s a box for the OpenThingCloud decentralized bandwidth sharing solution.

People purchase the OEC or OEC-Turbo device and install it at home to be part of a distributed P2P CDN (Content Delivery Network), caching data for other nodes operating nearby. As I understand it, there’s no user-facing functionality, and people install this type of hardware at home to get passive income.

OneThing Edge Cube (OEC) / Cube Pro (OEC-Pro) specifications:

• SoC – Rockchip RK3566
• System Memory – 2GB (OEC) or 4GB (OEC-turbo) LPDDR4X
• Storage
  • 8GB eMMC flash
  • SATA 3.0 connector
• Networking
  • Gigabit Ethernet port
  • 100+ Mbps upload bandwidth recommended
• USB – USB 3.0 port, USB 2.0 Type-C 2.0 port
• Power Supply – 12V DC Power supply

The DTS file implies it’s running Android since the model name is “Onething RK3566 BOX WXY4 V10 ANDROID Board”.  If you’re interested in mainline Linux support, the submitter (Jun Yan) has successfully tested Ethernet, USB 3.0, and SATA 3.0 ports. Besides the OEC models, the company also offers EOA (quad-core, 1GB DDR, 128GB eMMC, 50+ Mbps upload bandwidth recommended) and EOS (hexa-core, 4GB DDR, 8GB eMMC, 100-300Mbps upload bandwidth recommended) and variants. The XIYING OS NAS is another platform making use of the company’s P2P technology, but for storage. They also have a mobile app, PC software, and a program for other NAS systems.

You’ll find all that on the smart devices and cloud pages, except nothing can be purchased or downloaded until you switch to the Chinese version of the website, since the company currently operates in mainland China with 20+ million registered users and 1 to 2 million active nodes.

Earnings depend on the bandwidth a user shares, so users will earn more in a city with a high population density than in a rural area. It will also depend on upload bandwidth and storage capacity, and it’s not a get-rich-quick scheme, as 3 to 8 CNY per day is typical in cities, or about 44 cents to $1.2 a day.

A similar solution to OneThingCloud that operates outside of China is HoneyGain. It doesn’t require a specific box, and you can just install an app. Earnings appear to be even lower than OneThingCloud, with $2 to $20 reported per month. Some ISPs may also not appreciate it when their retail users are selling their upload bandwidth, so you may want to check your contract first, and it may not be worth the hassle considering the low earnings.

The post OneThingCloud decentralized bandwidth sharing/distributed CDN solution pays you for your “spare” bandwidth appeared first on CNX Software - Embedded Systems News.

Linus Torvalds has just released Linux 7.1 on LKML:

So it’s only Sunday morning back home, but it’s Sunday afternoon where I am right now, so I’m doing the 7.1 release at the regular time – just not in the regular timezone.

This obviously means that the merge window opens tomorrow, but I’ll be in yet another timezone by then, so timing will all be a bit irregular. Normally I try to front-load the merge window and do as much as possible the first few days – this time I’m not sure that will work out with my laptop and a couple of long flights without internet, but I’ve made sure that I have fetched the early pull requests (thank you – you know who you are), so I will be able to do some of it off-line.

Anyway, possible slight hiccups in the merge window aside, the news today is 7.1. Below is the shortlog for the last week – nothing particularly interesting or scary stands out, which is as it should be. It’s mostly various smaller driver updates (gpu, networking, sound, misc) with some networking and trace tooling fixes. And random minor changes elsewhere.

Please do keep testing despite the release, and apologies in advance if my merge window latency is going to be a bit random the next few days. I briefly  onsidered just extending the release for a week, but decided it wasn’t really worth it. I may come to regret that decision,

Linus

Released about two months ago, Linux 7.0 introduced AI coding assistants documentation, implemented a new generic API for file IO error reporting, the second phase of improved swapping performance with swap tables, and Rust got a rank upgrade of sorts, since it is no longer experimental. Now that Linux 7.1 has been released, it’s time to list some newsworthy changes and check out the embedded-focused Arm, RISC-V, and MIPS architectures in more detail.

Interesting changes in Linux 7.1

Some notable updates in Linux 7.1 include:

• New version of the NTFS filesystem (again) – I still remember NTFS being a pain on Arm in the mid 2000’s, as we had to use a fuse-based version (NTFS-3G) to get write support, but with limited performance. Since then things have improved with projects like ntfs3, but apparently not enough, as there’s now a “completely rewritten version” with full support for writes, a conversion to use iomap internally, and a promise of more active maintenance.
• Security enhancements – Linux 7.1 introduces the Landlock security module for Unix domain sockets, stricter default permission overrides for /proc/PID/mem,  security-module hooks for overlay filesystems, and the libcrypto library has gained support for a range of new algorithms.
• The high-resolution-timer core has been rewritten for better performance; in practical terms, the scheduler can now use high-resolution timers with no performance loss relative to a scheduler using coarse timers.
• Removal of the swap map after several kernel releases implemented various swap improvements. The only user-visible changes should be better performance and reduced memory usage by the swap subsystem. See details on LWN.
• Support has been removed for some old and unused 486 subarchitectures (M486, M486SX, and ELAN)

Arm changes in Linux 7.1

As usual, there were plenty of changes to the Arm architecture:

• Arm 9.6 LSUI feature adds instructions allowing the kernel to access user-space memory without disabling the “privileged access never” mode first. Linux 7.1 uses the new instructions to accelerate futex operations.
• Support for Memory Partitioning and Monitoring (MPAM) feature has been improved and exposed to user space. Check out the documentation for more information.
• Allwinner
  • MTD – SunXi driver support for new versions of the Allwinner NAND controller.
  • Power Management – Add support for the Allwinner A733 power domains
  • Clock driver – Support for the r-spi module clock in the A523 PRCM block,
  • Device tree changes for Linux 7.1
    • Allwinner A523 SoC family has LED controller enabled.
    • Avaota A1 board has SPI NAND enabled.
    • UART DMA channels added for A64 and H6. Standard resolution MMIO timer added for H616. This timer can be used as a broadcast timer for wakeup from idle states.
  • New devices – TaiqiCat (TQC) A01 set-top-box based on Allwinner H6
• Rockchip
  • Added support for Rockchip RV1103B 32-bit single-core vision processor (clock, pinctrl, etc.)
  • Clock –  Plug an OF node leak in Rockchip rk808 clk driver
  • PCIe controller driver
    • Add tracepoints for PCIe controller LTSSM transitions and link rate changes
    • Trace LTSSM events collected by the dw-rockchip debug FIFO
  • New devices
    • Khadas Edge 2L (RK3576)
    • Rockchip RK3576 EVB2 board
    • OneThing Edge Cube (OEC) based on RK3566
• Amlogic
  • Pinctrl – Fix a deadlock in the Amlogic driver, caused by playing around in sysfs
  • ARM device tree – drop iio-hwmon in favour of generic-adc-thermal
  • ARM64 device tree for Linux 7.1
    • Fix Ethernet PHY interrupt number for P230 reference board
    • Add missing cache information to cpu0 for Amlogic AXG
    • Fix Khadas VIM4 board model name
    • Fix GIC register ranges for Amlogic T7
    • Fix Khadas VIM4 memory layout for 8GB RAM
    • Drop CPU masks from GICv3 PPI interrupts for Amlogic S6
  • New device – N/A
• Samsung
  • Clock
    • Axis ARTPEC-9:  Add new PLL clocks and new drivers for eight clock controllers on the SoC
    • Exynos Auto v920: Add G3D (GPU) clock controller
      Exynos 850: Define missing clock for the APM mailbox
  • Fix booting of secondary CPU on Exynos5250 based Google Manta board – difference in TZ firmware.
  • SoC Drivers
    • Few cleanups in ACPM firmware drivers, used on Google GS101 and newer Samsung Exynos SoCs. Notable change is removing ‘const’ in ‘struct acpm_handle’ pointers, because even though the code does not modify pointed data, it immediately drops the const via cast. Also code is not logically readable when a reference getters/putters (e.g. acpm_handle_put()) take a pointer to const, because the meaning of “get” and “put” implies changing the memory, even if that changeable field is outside of pointed data.
  • DTS ARM changes – Few cleanups
  • Samsung DTS ARM64 changes for Linux 7.1
    • Add initial support for Axis ARTPEC-9 SoC and Alfred board using it. Just like ARTPEC-8, this is a derivative of Samsung Exynos SoC made for Axis, sharing most or all of core SoC blocks with Samsung designs.
    • Google GS101 Pixel phone: describe all PMIC regulators and Maxim fuel-gauge.
    • Exynos Auto v920: add G3D (GPU) clock controller (CMU).
  • Defconfig changes – N/A
  • New Devices
    • Exynos 5250 based Google Manta (Nexus 10)
    • Exynos 7870 based Samsung Galaxy J7 (2016) and Samsung Galaxy J5 (2017).
• Qualcomm
  • New SoCs
    • Qualcomm Glymur is a compute SoC using 18 Oryon-2 CPU cores (Snapdragon X2 Elite, SKU: X2E-96-100, X2E-88-100)
    • Qualcomm Mahua is a variant of Glymur with only 12 CPU cores (SKU: X2E-80-100), but largely identical.
    • Qualcomm Eliza is an embedded platform for mobile phone (SM7750) and IoT (QC7790S/M) workloads
    • Qualcomm IPQ5210 is a wireless networking SoC using Cortex-A53 cores
  • Clock
    • The parking of shared RCGs during registration parks the MDP source clock, disabling display until the msm display driver successfully initializes the hardware again. Mark this clock on Makena and Hamoa as “no_init_park” to retain a working recovery console, etc.
    • Global TCSR, RPMh, and display clock controller for the Qualcomm Eliza platform
    • TCSR, the multiple global, and the RPMh clock controller for the Qualcomm Nord platform
    • GPU clock controller for Qualcomm SM8750
    • Video and GPU clock controller for Qualcomm Glymur
    • Global clock controller for Qualcomm IPQ5210
  • Pinctrl
    • Qualcomm Eliza and Hawi families TLMM pin controllers
    • Qualcomm SDM670 and Milos family LPASS LPI pin controllers
    • Qualcomm IPQ5210 pin controller
  • remoteproc
    • Introduce support for more than 10 entries in the Qualcomm minidump implementation.
    • Add audio DSP remoteproc support for the Qualcomm Eliza platform. Add modem remoteproc support for the Qualcomm MDM9607, MSM8917, MSM8937, and MSM8940 platforms.
    • Add a list of Qualcomm QMI service ids to the QMI header file, in order to avoid sprinkling them across the various drivers using them. Migrate sysmon to use this constant.
  • PHY –  Qualcomm Eliza QMP UFS PHY
  • iommu – MMU-500 device-tree binding additions for Qualcomm Eliza & Hawi SoCs
  • PCIe controller driver – Advertise ‘Hot-Plug Capable’ and set ‘No Command Completed Support’ since Qcom Root Ports support hotplug events like DL_Up/Down and can accept writes to Slot Control without delays between writes.
  • WiFi (ath12k)
    • Monitor mode support on IPQ5332
    • Nasic hwmon temperature reporting
    • Support IPQ5424
  • Driver updates
    • Add ECS LIVA QC710, Glymur CRD, Mahua CRD, Purwa IoT EVK, and Asus Vivobook to the QSEECOM allow-list, to enable UEFI variable access
      through uefisecapp.
    • Register the Gunyah watchdog device if the SCM driver finds itself running under Gunyah. Clean up some locking using guards.
    • Handle possible cases where AOSS cooling state is given a non-boolean state.
    • Replace LLCC per-slice activation bitmap with reference counting. Also add SDM670 support.
    • Improve probe deferral handling in the OCMEM driver.
    • Add Milos, QCS615, Eliza, Glymur, and Mahua support to the pd-mapper.
    • Add support for SoCCP-based pmic-glink, as found in Glymur and Kaanapali.
    • Add common QMI service ids to the main qmi header file, to avoid spreading these constants in various drivers.
    • Add support for version 2 of SMP2P and implement the irqchip state reading support.
    • Add CQ7790, SA8650P, SM7450, SM7450P, and IPQ5210 SoC and the PM7550BA PMIC identifiers to the socinfo driver.
    • Add Eliza and Mahua support to the UBWC driver, introduce helpers for drivers to read out min_acc length and other programmable values, and disable bank swizzling for Glymur.
    • Simplify the logic related to allocation of NV download request in the WCNSS control driver.
  • Arm64 device tree updates
    • Snapdragon X Elite – Introduce UFS support and flatten the DWC3 node on Hamoa. Enable UFS, SDC, DisplayPort audio playback, and an EL2 overlay for the Hamoa IoT EVK. Enable DisplayPort audio on the Hamoa CRD and add HDMI support on the ASUS Zenbook A14. Reduce the duplication of thermal sensors across Purwa and Hamoa.
    • IPQ5332/IPQ9574 – Add the QPIC SPI NAND controller. Describe and enable the eMMC controller on IPQ9574.
    • Snapdragon 8 Elite Gen 5 (SM8850 and variants) – Add display, audio/compute remoteprocs, QUP devices, thermal sensors, display, and CoreSight on the Kaanapali platform. Enable audio, compute display, PMIC, Bluetooth, and WiFi on the MTP. Describe PMIC, audio, and compute remoteprocs on QRD.
    • Snapdragon Ride (SA8775P) – Add role-switching support for the tertiary USB controller on Lemans. Enable the tertiary USB controller and the GPIO expander on the Lemans EVK, and add an overlay for the IFP Mezzanine.
    • Snapdragon 7s Gen 3 (SM7635) – Add UFS, camera control interface, audio GPR, and FastRPC support on Milos. Enable UFS, camera EEPROMs, and hall effect sensor on the Fairphone FP6.
    • Dragonwing IQ8 – Add camera control interface and fix a variety of things on the Monaco platform, add missing FastRPC compute banks. Add eMMC support, describe the DisplayPort bridge and GPIO expander on the Monaco EVK. Add overlay for EVK camera and the IFP mezzanine. Describe DSI on the Monaco SoC and enable Bluetooth, WiFi, and DSI/DP bridge on the Ride board.
    • Add touchscreen to the Xiaomi Redmi 4A, 5A, and Go, and fix the board-id on the 4A.
    • Add the ambient light and proximity sensor on the Asus ZenFone 2 Laser/Selfie.
    • QCS6490/QCM6490 – On Kodiak-based boards, enable the Ethernet and USB Type-A ports on the Rb3 Gen2, correct the LT9611 routing on the RubikPi3, add Bluetooth on the IDP, and add front camera support on the Fairphone FP5. Introduce an overlay for the Rb3 Gen2 Industrial Mezzanine.
    • QRB2210/QRB4210 – Describe the WiFi/BT combo chip properly on the QRB2210 RB1 and QRB4210. Describe the DSI/DP brigde on the Arduino Uno Q.
    • SC8280XP – Introduce DSI display support
    • Add LLCC on SDM670 and another SPI controller on SDM630.
    • SDM845 – Properly describe the WiFi/BT chip on a variety of devices. Introduce the “alert slider” on the OnePlus 6 and OnePlus 6T devices.
    • SM6125 – Introduce the PRNG, describe the debug UART, and add the MDSS core reset. Enable the debug UART and fix various issues on the Xiaomi
      Redmi Note 8. Describe the touchscreen on the Xiaomi Mi A3.
    • Properly describe the WiFi/BT combo chip in SM8150 HDK.
    • SM8550 – Improve the EAS properties, in addition to various other fixes. Introduce a new overlay for the HDK display card.
    • Introduce various smaller fixes across SM8450 and SM8650.
    • SM9750 – Add display support and enable DSI and DisplayPort on the MTP. Also add tsens and thermal-zones.
    • QCS615 – Add ETR devices, flatten the USB controller node, and mark USB controllers as wakeup-capable devices, on Talos. Properly describe the IPA IMEM slice on a variety of platforms.
    • Drop redundant non-controllable regulator definitions from a variety of boards.
    • Drop redundant VSYNC pin state definition from various platforms.
  • Arm32 device tree updates
    • Qualcomm APQ8084 is incomplete and hasn’t seen functional contributions since 2016, so drop the platform (for now?). Also drop a number of unused IPQ-related dtsi files.
    • Clean up the RPM bus clocks in MSM8974 interconnect nodes.
  • Arm64 defconfig updates for Linux 7.1
    • Enable base drivers for booting boards based on Kaanapali, Glymur, Eliza, SM8750, and IPQ5210 SoCs.
    • Enable the Milos LPASS LPI driver. Enable the Qualcomm WCD937x headphone codec driver, used on QCM6490 boards.
    • Enable the QCOMTEE driver, to support the interface found in various Qualcomm SoCs.
    • Enable Ethernet and analog codecs used on the Arduino Ventuno Q.
    • Enable the Lontium LT8713sx driver, used for the Monaco EVK board.
    • Enable the PCI pwrctrl generic driver used in a few different Qualcomm-based boards with USB controllers on PCI buses.
  • New Devices
    • Redmi 4A, Redmi Go, Redmi Note 8T
    • Arduino Monza (Arduino VENTUNO Q) SBC
    • Purwa EVK
    • ECS Liva QCS710 mini PC
    • Additional variants of the DB820c Ayaneo Pocket S2
    • Thundercomm AI Mini PC G1
    • Samsung Galaxy Core Prime LTE Verizon Wireless
    • Wiko Pulp 4G
    • The Purwa-variant of ASUS Vivobook S15
    • The Eliza MTP, and the Glymur and Mahua CRDs.
• MediaTek
  • Pinctrl – Implement the GPIO .get_direction() callback in the Mediatek driver to rid dmesg warnings
  • PHY – Added Mediatek mt8167 dsi-phy
  • WiFi (mt76):
    • mt7996/mt7925 MLO fixes/improvements
    • mt7996 NPU support (HW eth/wifi traffic offload)
  • Bluetooth
    • Support MT6639 (MT7927)
    • Support MT7902 SDIO
  • DRM
    • mtk_dsi: enable hs clock during pre-enable
    • Remove all conflicting aperture devices during probe
    • Add support for mt8167 display blocks
  • MFD – Correct the hardware CIDs for the MT6328, MT6331, and MT6332 PMICs to allow proper driver binding
  • MediaTek PCIe Gen3 controller driver
    • Use dev_err_probe() to simplify error paths and make deferred probe messages visible in /sys/kernel/debug/devices_deferred
    • Power off device if setup fails
    • Integrate new pwrctrl API to enable power control for WiFi/BT adapters on mainboard or in PCIe or M.2 slots
  • ARM32 updates – Fix a dtbs_check validation error by changing the fallback compatible of the efuse node to the correct one in MT7623.
  • ARM64 device tree updates for Linux 7.1
    • The gpio-ranges pin was fixed in MT6795, MT7981B, and MT7986A SoCs, as the very last GPIO was unusable; Even though anyway unused, this fixes the hardware description.
    • Model string fixes for Banana Pi BPI-R4 Pro 4E/8X: now the correct model is shown.
    • The MT6359 PMIC gets disambiguation for two default regulator names, mainly fixing issues seen on U-Boot, but also making the regulators visually distinguishable in a summary…!
    • Aliases for eMMC/SD controllers added in MT8365 EVK board, MT8395 Radxa NIO-12L and Genio 1200 for consistency
    • Fixes to the MediaTek AUDSYS device tree binding
    • MT8195 Cherry Chromebooks get their WiFi on PCI Express and Bluetooth on USB described with the proper power supplies now tied to the correct devices (USB VBUS and PCIE3v3): this is now described almost perfectly, or at least links the right resources in the right places. This is also done as a preparation for when the M.2 E-Key connector binding will be upstreamed.
    • MT8195 Cherry Dojo gets its M.2 M-Key slot correctly described with the new pcie-m2-m-connector binding.
  • Defconfig updates – N/A
  • New devices – N/A
• Other new Arm hardware platforms and SoCs
  • Arm
    • Zena virtual platform in FVP using Cortex-A720AE cores, with additional versions planned to be merged in the future.
    • ARM corstone-1000-a320 is a reference platform for IOT, using low-end Cortex-A320 cores
  • Aspeed – Two more Aspeed BMC-based boards
  • Axis – ARTPEC-9 Cortex-A55 cores embedded SoC using the Samsung SoC platform
  • Microchip – LAN9691 updated 64-bit variant of the arm32 lan966x series of networking SoCs
  • NXP
    • S32N79 automotive Cortex-A78AE (8x) + Cortex-R52 (12x) SoC, a significant upgrade from the older S32V and S32G series
    • 22x additional industrial/embedded boards using 64-bit NXP i.MX8M or i.MX9 SoCs
  • Renesas – RZ/G3L (r9a08g046) is an industrial embedded chip using Cortex-A55 cores, similar to the already supported G3E and G3S variants
  • Texas Instruments – Four variants of the Toradex Verdin using TI AM62
• Raspberry Pi-specific changes
  • Add the V3D DT node to the 2712 (Raspberry Pi 5) SoC
  • Add the I2C controller, CSI (camera), ISP (image signal processor), fix the pinctrl node,  and update the UART10 interrupt for the RP1 sister chip to the 2712 (Raspberry Pi 5)
  • Move the firmware and GPU to the root level to fix DTC warnings
  • Enable EEE for Raspberry Pi RP1
  • gpiolib: handle gpio-hogs only once –  Fixes a behaviour change that breaks boot on Raspberry Pi 5 when using the firmware-supplied device tree

RISC-V updates in Linux 7.1

RISC-V support keeps progressing steadily:

• Add Kunit correctness testing and microbenchmarks for strlen(), strnlen(), and strrchr()
• Add RISC-V-specific strnlen(), strchr(), strrchr() implementations
•  Add hardware error exception handling
• Clean up and optimize our unaligned access probe code
•  Enable HAVE_IOREMAP_PROT to be able to use generic_access_phys()
•  Remove XIP kernel support (since it’s not used by anybody)
• Warn when addresses outside the vmemmap range are passed to vmemmap_populate()
• Update the ACPI FADT revision check to warn if it’s not at least ACPI v6.6, which is when key RISC-V-specific tables were added to the specification
• Increase COMMAND_LINE_SIZE to 2048 to match ARM64, x86, PowerPC, etc.
• Make kaslr_offset() a static inline function, since there’s no need for it to show up in the symbol table
• Add KASLR offset and SATP to the VMCOREINFO ELF notes to improve kdump support
• Add Makefile cleanup rule for vdso_cfi copied source files, and add a .gitignore for the build artifacts in that directory
• Remove some redundant ifdefs that check Kconfig macros
• Add missing SPDX license tag to the CFI selftest
• Simplify UTS_MACHINE assignment in the RISC-V Makefile
• Clarify some unclear comments and remove some superfluous comments
• Fix various English typos across the RISC-V codebase
• Alibaba T-Head
  • Hwmon – Adapt moortec,mr75203 bindings for T-Head TH1520
  • DRM – Support TH1520 HDMI plus DT bindings
  • Device tree updates for Linux 7.1
    • Update the T-Head TH1520 RISC-V SoC device tree to support the Verisilicon DC8200 display controller (called DPU in manual) and the Synopsys DesignWare HDMI TX controller. In addition, enable HDMI output for the LicheePi 4a board.
    • Enable the display pipeline for the TH1520-based BeagleV Ahead board by adding the HDMI connector node, connecting it to the HDMI controller, and activating the DPU and HDMI nodes.
    • Add coefficients to the TH1520 PVT node as the values in the TH1520 manual differ from the defaults in the driver.
• Microchip
  • Add support for Microchip PIC64GX embedded RISC-V chip using SIFIVE U54 CPU cores. “PolarFire SoC without the FPGA.”
  • SoC drivers
    • Add coverage for the pic64gx in the system controller and syscons.
    • Add an interrupt mux driver (akin to the one that Renesas recently added) that fixes a problem where the platform never properly modelled GPIO interrupts. There’s a GPIO driver change here that Bartosz has acked that adds the interrupt support to the GPIO driver itself.
  • Device tree
    • Add support for the picgx64 and its curiosity board.
    • Add the missing tsu_clk for ptp on the macb on PolarFire SoC and resolve a long-running problem with gpio interrupts being incorrectly described
      on the platform.
• SiFive – spi: sifive: fix controller deregistration. Make sure to deregister the controller before disabling underlying
  resources like interrupts during driver unbind.
• Sophgo
  • PCIe controller driver – Disable ASPM L0s and L1 on Sophgo 2042 PCIe Root Ports that advertise support for them (
  • DMA engine – Add support for CV1800B DMA
• SpacemiT
  • Device Tree changes for Linux 7.1
    • K3 SoC
      • Add I2C support
      • Add PMIC regulator tree
      • Add Ethernet support
      • Add pinctrl/GPIO/Clock
      • Enable full UART support
    • K1 SoC
      • Milk-V Jupiter – Enable PCIe/USB on, QSPI/SPI NOR flash, Enable EEPROM, and LEDs
      • Fix PMIC supply properties
      • Fix PCIe missing power regulator
  • Power Management – SpacemiT P1:
    • Drop the deprecated “vin-supply” property from the devicetree bindings
    • Add individual regulator supply properties to match actual hardware topology
• StarFive
  • USB – Add support for StarFive JHB100 SoC. JHB100 contains 2 dwc3 USB controllers and PHYs working at USB 2.0 speed. Note: it’s a BCM RISC-V chip.

MIPS changelog

As usual, MIPS was pretty quiet with just two lines for the summary:

• Support for Mobileye EyeQ6Lplus
• Cleanups and fixes

Here are a few of the commits:

• MIPS/mtd: Handle READY GPIO in generic NAND platform data
• MIPS/input: Move RB532 button to GPIO descriptors
• MIPS: validate DT bootargs before appending them
• MIPS: Alchemy: Remove unused forward declaration
• MAINTAINERS: Mobileye: Add EyeQ6Lplus files
• MIPS: config: add eyeq6lplus_defconfig
• MIPS: Add Mobileye EyeQ6Lplus evaluation board dts
• MIPS: Add Mobileye EyeQ6Lplus SoC dtsi
• clk: eyeq: Add Mobileye EyeQ6Lplus OLB
• clk: eyeq: Adjust PLL accuracy computation
• clk: eyeq: Skip post-divisor when computing PLL frequency
• pinctrl: eyeq5: Add Mobileye EyeQ6Lplus OLB
• pinctrl: eyeq5: Use match data
• reset: eyeq: Add Mobileye EyeQ6Lplus OLB
• MIPS: Add Mobileye EyeQ6Lplus support
• dt-bindings: soc: mobileye: Add EyeQ6Lplus OLB
• dt-bindings: mips: Add Mobileye EyeQ6Lplus SoC
• MIPS: dts: loongson64g-package: Switch to Loongson UART driver
• mips: pci-mt7620: rework initialization procedure
• mips: pci-mt7620: add more register init values

For more details, you can read the complete Linux 7.1 changelog generated with comments only using the command git log v7.0..v7.1 --stat. Alternatively,  Kernelnewbies should have a longer list of changes for the Linux 7.1 release.

The post Linux 7.1 Release – Main changes, Arm, RISC-V, and MIPS architectures appeared first on CNX Software - Embedded Systems News.

Last March, Gateworks introduced the Catalina SBC family, starting with the GW9200 board, based on NXP i.MX 95 SoC for industrial Edge AI applications, and featuring Flexible Socket Adapter (FSA) sockets to accommodate either mini PCIe or M.2 module adapters, depending on requirements.

We missed the original announcement, but the new SBC family was brought to our attention when Ezurio announced the acquisition of Gateworks last week. The GW9200 board comprises a system-on-module with the NXP i.MX 95 chip, 4GB LPDDR5 and 8GB eMMC flash by default, as well as a carrier board with 10GbE and GbE RJ45 ports, a MIPI DSI/CSI display/camera interface, two FSA sockets for M.2 or mini PCIe expansion, and a range of I/Os.

Ezurio/Gateworks Catalina GW9200 specifications:

• SoC – NXP i.MX 95
  • CPU
    • Up to 6x Arm Cortex-A55 application cores clocked at 1.8 GHz with 32KB I-cache and D-cache, 64KB L2 cache, and 512KB L3 cache
    • 1x Arm Cortex-M7 real-time core clocked at 800 MHz
    • 1x Arm Cortex-M33 safety core clocked at 333 MHz
  • GPU – Arm Mali-G310 V2 GPU for 2D/3D acceleration with support for OpenGL ES 3.2, OpenCL 3.0
  • VPU
    • 1080p60/4Kp30 H.265 and H.264 encode and decode
    • JPEG Encoder, JPEG Decoder
    • Decoding – 4Kp30/1080p60 HEVC/H.265, VP9, VP8, AVC/H.264 (Baseline / Main / High Profile)
    • Encoding – 4Kp30/1080p60 HEVC/H.265, AVC/H.264
  • AI Accelerator – NXP eIQ Neutron 2 TOPS neural processing unit (NPU) with 750 inf/sec
• System Memory – 4GB LPDDR5 (up to 16GB)
• Storage
  • 8GB eMMC flash (up to 64GB)
  • MicroSD card slot
• Display/Camera Interface – MIPI DSI or CSI connector
• Networking
  • 10GbE RJ45 jack
  • 2x Gigabit Ethernet RJ45 jacks
  • Optional Wi-Fi 6/7 and Bluetooth via Mini PCIe or M.2 adapter in FSA site
  • Optional 4G LTE or 5G module via FSA site plus two nano SIM card slots.
• GNSS – Optional u-blox GPS with GLONASS
• USB – USB 2.0 Type-C port
• Serial
  • 2x RS232 or 1x RS485
  • 2x CAN Bus
• Expansion
  • 2x Flexible Socket Adapter (FSA) sockets for Mini PCIe or M.2 adapters.
  • SPI connector
  • Digital I/O connector with I2C, ADC, UART
• Security
  • Optional NXP EdgeLock SE052F Security Chip
  • Optional TPM
• Debugging – JTAG connector
• Misc
  • User push-button
  • Power LED
  • 3-Axis accelerometer
  • Coin-cell battery for RTC
  • Fan connector
• Power Supply
  • 8 – 60V DC via power “header”
  • 802.3af/at PoE via optional PoE adapter)
• Dimensions – 100 x 90 mm
• Temperature Range – -40 to +85°C (Industrial)

The company provides support for Linux (Ubuntu, Buildroot, and more), and the GW11059 development kit with the board itself, a JTAG programmer, a power supply, and relevant cables. Check out the software documentation website for details.

The FSA (Flexible Socket Adapter) is an interesting concept, as customers can select whether to use M.2 or mini PCIe sockets during configuration or production to meet their specific requirements. I could not find any photos of the FSA connector, but we’re told that the “mating connector between the FSA and the SBC is very fine pitched” and “only rated for 30 mating cycles maximum”. Four FSA to M.2/mini PCIe are available:

• GW16FE0 – FSA to M.2 E Key (WiFi)
• GW16FB0 – FSA to M.2 B Key (Modems)
• GW16FM0 – FSA to M.2 M Key (NVME)
• GW16FP0 – FSA to Mini-PCIe

The GW6200 comes with two FSA sockets, but the upcoming, larger GW6400 (160 x 100 mm) SBC will feature four FSA sockets.

Gateworks/Ezurio says the GW6200 SBC is available now, and the larger GW6400 board is shown as being planned for Q2 2026 (about now).  There’s no public pricing, and interested parties are asked to “request a quote”. Additional information may be found on the product page.

The post Gateworks Catalina GW9200 NXP i.MX 95 SBC features Flexible Socket Adapter sockets for M.2 or mPCIe modules appeared first on CNX Software - Embedded Systems News.

The GEEKOM A8, an AMD Ryzen 7 874HS octa-core mini PC equipped with 16GB DDR5 and a 1TB NVMe SSD, is now offered at a 20% discount, dropping the price to $519/￡503 from $649/ ￡629 MSRP.

This promotional price is expected to be lower than the A8’s pricing during the June 2026 Prime Day event, making it the best time to buy. To enjoy the discount on the Windows 11 Pro mini PC, simply enter the coupon code CNXGKA820 on the GEEKOM US or GEEKOM UK stores before July 2.

GEEKOM A8 specifications:

• SoC – AMD Ryzen 7 8745HS
  • CPU – 8-core/16-thread processor up to 3.8GHz / 4.9 (Turbo)
  • L3 Cache – 16MB
  • GPU – AMD Radeon 780 Graphics
  • AI – NPU disabled. AI performance from CPU + GPU not specified
  • TDP: 45W
• System Memory – 16GB DDR5 SODIMM, upgradeable up to 64GB
• Storage
  • 1TB NVMe SSD via M.2 SSD socket (PCIe Gen4 x4), upgradeable up to 4TB
  • Full-size SD card reader
• Video Output
  • 2x HDMI 2.0 ports
  • USB4 and USB 3.2 Type-C ports with DisplayPort Alt mode
  • Up to 4x independent displays
• Audio – 3.5mm stereo headset jack; digital audio output via HDMI and USB-C
• Networking
  • 2.5GbE RJ45 port via RealTek RTL8125BG-CG controller
  • WiFi 6E and Bluetooth 5.3 via AW-EB600NF module
• USB
  • 3x USB 3.2 Gen 2 Type-A ports, including one with USB PD support (for charging peripheral)
  • USB4 Gen3 Type-C port with DisplayPort Alt mode, USB PD support
  • USB 3.2 Gen 2 Type-C port with DisplayPort Alt mode, USB PD support
  • USB 2.0 Type-A port
• Security – fTPM 2.0
• Misc
  • Power button
  • Kensington lock slot
  • RTC battery
• Power Supply – 19V/6.32A (120W) power supply adapter with geo-specific AC cord (IEC C5)
• Dimensions – 112 x 112 x 37 mm
• Certifications – CE, FCC, CB, CCC, RoHS

While the mini PC comes pre-loaded with Microsoft Windows 11 64-bit, GEEKOM claims support for Ubuntu and other Linux distributions. We did review the earlier GEEKOM A8 released in 2024 on both Windows 11 Pro and Ubuntu 24.04, and it worked well in both, but the updated model offers a different AMD Ryzen processor and WiFi module.

The GEEKOM A8 mini PC ships with an HDMI cable, a power adapter with a country-specific power cord, a VESA mount, and a user guide. As with all its mini PCs, GEEKOM offers shipping from a local warehouse for easy returns, a 3-year warranty, and a 30-day refund & return commitment.

Besides the GEEKOM A8 mini PC promotion, the company will be running its Summer Sale, with discounts starting from 15% off across the entire lineup, up to 50% off. It is their biggest promotion of the year so far, offering pricing that is even lower than select Prime Day deals. It will run until June 30.

The post Lower Than Prime Day Pricing: 20% off on GEEKOM A8 AMD Ryzen 7 8745HS mini PC (Sponsored) appeared first on CNX Software - Embedded Systems News.

At the end of last year, I received a Creality Sermoon S1 high-end 3D scanner for review. After checking the specifications and going through an unboxing in the first part of the review, I used the 3D scanner with an Intel Core i5-13500H laptop with 16GB of RAM running Creality Scan 4 software on Windows 11.

The laptop specs were below the minimum hardware requirements (NVIDIA GPU, 32GB RAM), and while I managed to scan a face and bust using infrared mode, it was a struggle with 4 to 5 FPS scanning, and I wasn’t able to use Blue light scanning at all. Luckily, shortly after the review, Khadas informed me they planned to send a Khadas Mind Graphics 2 dock featuring an NVIDIA GeForce RTX 560 Ti 16GB graphics card and a Mind 2 mini PC with 32GB of RAM and an Intel Core Ultra 7 155H 16-core Meteor Lake processor. What I didn’t know was that it would take around 5 months to receive it!

Nevertheless, I’ve already reviewed the Khadas Mind Graphics 2 with 3D graphics and AI workloads, so it’s finally time to test it with the Creality Sermoon S1 3D scanner.  Since I’ve already gone through Crealite Scan 4 software installation, firmware update, and calibration in the previous review, I’ll focus on re-testing infrared mode to check performance and trying Blue mode scanning for detailed objects.

Infrared mode scanning with Sermoon S1 + Khadas Mind 2 Graphics

When installing Creality Scan 4 software, it will estimate the performance of the host, and I went from a “Poor PC performance” on my laptop with up to 9 FPS in blue laser mode, and 16 FPS in infrared mode, to “Excellent PC performance” with the Khadas Mind 2 mini PC and Mind 2 Graphics dock handling up to 90 FPS in blue laser mode and 30 FPS in infrared mode.

I also had to update the scanner firmware and decided to redo the calibration since it had been over 6 months. The calibration was much easier with the higher frame rate, and it took me only 2 to 3 minutes to complete.

Let’s try a scan in infrared mode using a largish Santa Claus plush toy.

It was night and day compared to the previous review. On the faster NVIDIA-based machine, I completed the 3D scan in just over 2 minutes, compared to 15 minutes on the Intel laptop. The frame rate is 21 FPS on the screenshot, but most of the time it was around 29 FPS.

There were some extra parts and holes in the scan, but after going through Fusion and Mesh processing with parameters like “Remove Isolated Parts”, “Fill Small Holes”, and enabling the “Water tight” option…

… the result looks fairly good. Note the Fusion part still took several minutes even as I reduced the number of triangles to 1.2 million.

Here’s a video showing the full scan.

That’s all good. As a reminder, infrared mode is best used for fast scanning and tracking of geometry (e.g., human faces and bodies, sculptures), texture (e.g., a porcelain vase with drawings), and markers for objects with insufficient geometry and texture.

Blue laser mode with Sermoon S1 3D scanner

Blue laser mode is best suited for high-accuracy, high-detail scans and typically requires markers. I could not use it at all with my laptop last time around; let’s give it a try now that I have a more powerful machine. Since it can be used on objects as small as 5 x 5 x 5mm and with high details, I decided to go with a 20 Bath commemorative coin.

The coin is just 32mm in diameter, and the text can be hard to read with the naked eye. It’s probably quite a difficult scan. On this type of shiny object, I’m also supposed to use scanning spray, but I will do without.

First, we need to prepare the desk by adding several 3mm tracking markers provided with the scanner. For small objects, markers are placed on the desk, but for larger objects, you can also place them on the object itself.

Let’s go to the Creality Scan software, create a new project, and start a Scan with Blue laser mode, with markers, Pointcloud scan mode, and parallel lines mode. Exclude flat base should also be selected.

For reference, three line modes are available:

• Crossed Laser Lines – This mode uses a grid of lines to rapidly scan larger objects or broad areas. It is designed for efficiency from a greater distance and typically requires a lower marker density.
• Parallel Laser Lines – This mode projects a set of parallel lines (7 lines) for close-up, high-precision scanning. It is best for capturing fine details on smaller objects or specific areas, and requires a higher marker density to maintain tracking accuracy.
• Single Laser Line – This mode uses a 0.1mm thick single line to access deep holes, narrow gaps, and recessed areas that crossed or parallel lines cannot reach, ensuring metrology-grade accuracy in intricate details.

I went with the Parallel laser lines here. The scan is pretty quick to complete, but the level of detail is not sufficient. We need to switch to “Local Detail” scan mode and set the resolution is 0.05mm.

You need to select the area with fine details. For this coin, it’s basically the whole area. It will be red when selected.

We can now start to see some details like the text. Now I repeated the same procedure with another scan in the same project for the other side of the coin. I’d recommend cleaning the design, notably removing the blue dots around the coin for each scan. I did that after the fusion step, but the blobs kept reappearing even after I started the scans from scratch. It was extremely frustrating. I eventually did the final cleaning after the Meshing step. More on that later.

Let’s run the Fusion Batch Process on both scans, removing markers with ultra-detail and the finest resolution we can select. Noise removal was kept at 30%.

This is what one side of the coin looks like after Fusion. There are plenty of artifacts from the location of the markers, which is why I recommend cleaning before. The good news is that we can read text like “20 Baht” (๒๐ บาท). We still need to align the two faces of the coin with the “Alignment” tool.

The first time, I tried automatic alignment with markers, but it just merged the two faces instead of creating a coin. So I switched to manual feature alignment add point 1 to 4 to strategic locations for each face, forcing the program to rotate the face in the Floating window as shown above.

We can now go through the Meshing. I slightly reduced the number of triangles (maybe I shouldn’t have), selected “Remove Isolated Parts”, and “Fill Small Holes”. The screenshot below shows the result after the meshing process.

I still had to clean it up with the tools from the bottom toolbar. Here are the final results.

Here’s the video showing the “coin”. Obviously, the result is not ideal, since a lot of the text is not readable, and I didn’t scan the side of the coin quite enough.

Using scanning spray may have helped, or this coin just has too many small details to be properly scanned. I noticed another person scanned a simpler 20mm golden coin with scan-spray and got somewhat better results.

I decided to scan something easier in Blue laser mode but still with some level of detail: a comb. I added a few extra 6mm markers since the object is larger.

The process was similar to scanning the coins. I used Blue laser scanning with parallel lines with the default 0.5mm resolution.

It took less than one minute to get the shape of the comb since scanning was done smoothly at 90 FPS.

However, at this resolution, the Japanese text is not clear at all, so I enabled the “Local Detail” scan mode and selected the area.

I scanned the area again, and after a few seconds, the Japanese text became clear.

There was a lot of dirt around the scan, so this type around I cleaned it up before doing anything else. Select the parts you want to remove with the tools in the toolbar (red when selected) and select the delete icon.

It’s a bit hard to clean around the high-resolution area since the resolution shown is fairly low, so I went through the Fusion step first.

I did some more clean up around the Japanese characters after the Fusion was complete.

Since both sides of the combo are identical, I hoped for a duplicate scan function, but I didn’t find any. So I turned the comb around and repeated the steps above for a scan of the other side of the comb. Once the Fused scan was relatively clean, I used the same manual feature alignment as for the coin.

Once the Fusion was successful, I switched to the Mesh step. We can still see some yellow parts here and there. Those are open parts of the design.

The final step was to automatically close these by selecting the Hole Filling section in the Mesh Processing tab. I used the default values, and after clicking on Preview, the impacted areas were shown in red.

I got the final result, which you can check in the video below, and I find it looks pretty neat.

I also exported the file to STL to import it into Blender, and there was no issue. The scan would just benefit from a little more cleaning before printing it with a 3D printer.

Conclusion

The Creality Sermoon S1 is my first 3D scanner, so I can’t really compare it to others. It still looks like a great piece of equipment for detailed and large objects thanks to the blue laser and infrared modes. After a learning curve, I was able to perform some relatively decent scans with either method.

The scanner doesn’t work with Linux, so you’d need a Windows PC or a Mac computer, or alternatively, an Android smartphone or iPhone if you purchase a WiFi connection kit. It takes a little while to get used to the Creality Scan 4 software and the scanning process in general, so you may need to be patient if you’re new to 3D scanning.

The USB cable is great for achieving maximum frame rate, but for larger objects it will be a bit short. I had to adjust the position of the mini PC and Santa Claus plush toy when scanning when the cable was too short. WiFi is likely preferred for large objects (e.g., a car), although bandwidth may become an issue depending on parameters.

3D scanning is not a cheap hobby if you don’t do this as a commercial endeavor, and you’ll probably need to budget at least $4,000 if you don’t already own a powerful computer. The scanner itself costs $2,299 on Amazon or the Creality store, and as we’ve seen from the second part of the review, you can’t simply use any computer for it to be usable, and a $1,500+ machine is required. The Khadas Mind 2 and Mind Graphics 2 dock I used in this review go for $1,099 and $1,349 on Amazon, so the total cost of the hardware used for the review would be $4,747 US. If you need a portable and powerful 3D scanning workstation, the combination of the Creality Sermoon S1, Khadas Mind 2 mini PC, and Mind Graphics 2 dock offers a neat solution.

The post Creality Sermoon S1 review – Part 3: 3D scanning with Khadas Mind 2 and NVIDIA GeForce RTX 5060 Ti 16GB dock appeared first on CNX Software - Embedded Systems News.

The ESP32-E22 tri-band Wi-Fi 6E and Bluetooth 5.4 module has received a Wi-Fi CERTIFIED certificate from the Wi-Fi Alliance, and Espressif has also released WiFi and Bluetooth Linux drivers for the chip.

The ESP32-E22 was first unveiled at CES 2026 with a dual-core RISC-V processor clocked at up to 500 MHz, 1MB RAM, tri-band WiFi 6E tested up to 2.1 Gbps with iperf, and dual-mode Bluetooth 5.4/6.0. While it also features 41 GPIO pins, it’s not mainly designed for IoT projects, but instead targets host-based wireless systems needing WiFi 6E / Bluetooth 5.4 connectivity through PCIe 2.0 or SDIO interfaces. In other words, the ESP32-E22 will likely show on M.2 modules, competing against similar products from MediaTek and Intel.

Espressif has not said which module was tested for the Wi-Fi CERTIFIED certificate, but I suspect it should be the upcoming ESP32-E22-M2-1 M.2 module listed on the company’s website.

From the certificate, it was tested under Linux, rather than Windows, with a range of certifications for network management (WMM, Wi-Fi Agile Multiband), Security (WPA-Personal/Enterprise, WPA3-Personal/Enterprise, Wi-Fi Enhanced Open), and Wi-Fi MAC/PHY, including 2.4, 5, and 6 GHz bands, and 802.11b/g/n/ac/ax.

The chip was tested in Station mode with 2×2 MIMO, up to 40 MHz channel width at 2.4 GHz, 160 MHz at 5 GHz and 6GHz. I don’t see any access point tests from the certificate; however, AP mode will also be supported.

This brings us to the second part of today’s news: the open-source Linux driver for the ESP32-E22 chip/module. Currently, only station mode over PCIe and Bluetooth over USB are supported, but more is coming:

• Wi-Fi (STA) – PCIe (Supported), SDIO (Not yet)
• Wi-Fi (AP) – PCIe and SDIO (not yet)
• Bluetooth BR/EDR and BLE over USB – Supported
• Bluetooth over UART – Not yet
• Firmware download over PCIe or USB –  Supported
• Secure Download – Not yet
• Wi-Fi/BT coexistence – Not yet
• Suspend/Resume (sleep) – Not yet

The company says it works with Linux 5.14 and greater, but I suppose it will have to be upstreamed eventually and backported to popular Linux distributions, considering the target applications, as you don’t want users to build the code from source each time a new kernel is released, or enable DKMS manually. The GitHub repository also features a binary-only unified firmware for both the ESP32-E22 Linux driver and Windows driver.

The product page on the Espressif Systems website has limited information, and neither the ESP32-E22 chip nor the ESP32-E22-M2-1 M.2 module appear to be for sale for now.

The post Espressif ESP32-E22 WiFi 6E module gets Wi-Fi CERTIFIED certificate, open-source WiFi and Bluetooth Linux drivers appeared first on CNX Software - Embedded Systems News.

M5Stack Capsule v1.1 is a Stamp-S3A-based IoT controller with a microSD card slot, several sensors (6-axis IMU, microphone),  an IR transmitter, a built-in 250 mAh battery, a few buttons, a buzzer, an RTC, and expansion capabilities through GPIO headers and a Grove connector.

It’s an upgrade to the earlier Capsule based on the Stamp-S3 module. The new version still features an ESP32-S3 WiFi and Bluetooth microcontroller, 8MB flash, a USB-C port, and a few GPIOs, but benefits from the Stamp-S3A improvements, including an optimized antenna design and lower power consumption. We never had a look at the Capsule before, so let’s do it now.

M5Stack Capsule v1.1 specifications:

• M5Stack Stamp-S3A module
  • WiSoC – Espressif Systems ESP32-S3FN8
    • CPU
      • Dual-core 32-bit Xtensa LX7 microcontroller with AI vector instructions up to 240MHz
      • RISC-V ULP co-processor
    • Memory – 512KB SRAM
    • Storage – 8MB flash
    • Wireless – 2.4GHz WiFi 4 (802.11b/g/n), Bluetooth 5.0 LE + Mesh
  • Connectivity
    • 2.4 GHz WiFi 4, 20 MHz and 40 MHz bandwidth, IEEE 802.11 b/g/n protocol, up to 150 Mbps
    • Bluetooth 5, Bluetooth Mesh, with support for 125 Kbps, 500 Kbps, 1 Mbps, 2 Mbps bitrate, long-range support
    • 2.4GHz 3D antenna
  • USB – 1x USB Type-C port for power and programming
• Storage – microSD card slot
• Sensor
  • 6-Axis BMI270 attitude/IMU sensor
  • SPM1423 MEMS Microphone
• Expansion
  • 2x 9-pin headers routing Stamp-S3A module’s signals
  • 4-pin Grove connector
• Misc
  • Reset and Wake buttons
  • Power LED
  • Infrared transmitter with the following range at various angles
    • Straight line: 330 cm
    • 90°: 48 cm
    • 45°: 134 cm
  • BM8563 RTC for timed wake-up
  • Built-in buzzer for status feedback
  • 4x magnets and 2x M3 threads for mounting
• Power Supply
  • 5V charging via USB-C port on Stamp-S3A module
  • 250 mAh battery
  • Power consumption (while battery powered)
    • Sleep – 4.2V @ 35uA
    • Operating – 4.2V @ 144mA
• Dimensions – 40.0 x 24.0 x 16.2mm
• Weight – 18.7 grams

The company lists support for UIFlow2 visual programming IDE, the Arduino IDE, the ESP-IDF framework, and PlatformIO. However, the documentation focuses on Arduino and UIFlow2 instructions, besides providing additional technical details about the hardware such as schematics and pinout diagrams and tables.

M5Stack says the Capsule v1.1 targets attitude control and motion tracking, audio recording and voice wake-up, low-power timed data collection, infrared remote control, smart terminals, and generic embedded system development.

M5Stack sells the Capsule v1.1 IoT controller for $19.95 on the company’s store, and for reference, the original Capsule is listed for $21.95 on AliExpress, and the new v1.1 should soon appear there too. The video provides an overview of the original M5Stack Capture (aka M5Capsule), and showcases a few demos.

The post M5Stack Capsule Kit v1.1- A Battery-powered ESP32-S3 IoT controller with IMU sensor, MEMS microphone, and IR transmitter appeared first on CNX Software - Embedded Systems News.

Motor driver IC specialist Fortior Technology has recently introduced the FU75xx dual-core motor control MCU family, pairing a 32-bit RISC-V core and the company’s proprietary 2nd-generation Motor Engine (ME2) core.

The RISC-V core is used for parameter configuration and routine processing, while the ME core integrates FOC and CORDIC modules that enable fast calculation of FOC (as quick as 5µs)  or square-wave control for sensored/sensorless BLDC/PMSM motors. The chips have an impressive list of peripherals (see specs below) and target high-speed computing and real-time control for robotics and motion systems, such as industrial servo drives, robotic joints, smart home appliances, and new energy vehicle systems.

FU75xx MCU specifications:

• Dual-core CPU
  • RISC-V core @ 48 MHz
  • ME2 motor engine core @ 48 MHz with FOC module and CORDIC module
• Memory – 12kB SRAM, 4kB PRAM for program execution
• Storage – 64kB Flash with ECC and CRC, self-program and code protection
• I/Os (numbers depend on SKU)
  • 27x to 32x interrupt sources with 8 configurable priority levels
  • 22x to 53x GPIO
• Communication interfaces
  • Up to 2x I2 C
  • Up to 2x SPI
  • 2x UARTs, supporting single-wire mode
  • 1x LIN
  • 1x CAN FD
  • BiSS (Bidirectional Serial Synchronous) interface for full-duplex communications between sensors and actuators  (FU7562Q/FU7561Q/FU7512L/FU7512P)
  • 6-channel DMA for I2C/SPI/UART/LIN/DataMonitor
• Analog peripherals
  • 12-bit ADC, operating with 1μs conversion time and internal VREF or external VREF as reference voltage
    • 8- to 13-channel ADC1
    • 10- to 14-channel ADC2
    • Up to 3-channel ADC3
    • Internal VREF – 3V, 4V, 4.5V or VDD5 can be selected as the internal reference.
    • Internal VHALF with VREF/2, 1/4 VREF, 1/8 VREF, and 25/64 VREF as the internal reference
  • 5x to 7x operational amplifiers
  • 4x to 8x analog comparators
  • DAC
    • FU7571Q/FU7561Q/FU7541Q: Dual-channel 9-bit, dual-channel 6-bit
    • FU7562Q/FU7512L/FU7512P: Dual-channel 9-bit, single-channel 8-bit, and dual-channel 6-bit
    • FU7562L: Dual-channel 9-bit, single-channel 8-bit
  • Temperature sensor
• Drive Type
  • FU7571Q: 2x 6N Pre-driver outputs
  • FU7562Q/FU7561Q: 6N Pre-driver output + PWM output
  • FU7562L: 6N Pre-driver output
  • FU7541Q: 2x 3P3N Pre-driver outputs
  • FU7512L/FU7512P: 2x PWM outputs
• PFC (Power Factor Correction): 1x in FU7562Q/FU7562L, 2x in FU7512L
  • Automatic hardware
  • ADC automatic sampling
  • Over-current protection and wave-by-wave current limiting protection
  • Interleaved two-phase boost PFC
• Debugging – Two-wire FICE (Fortior Interactive Connectivity Establishment) protocol-based in-circuit emulation
• Timers
  • Timer2/Timer5/Timer6 (FU7562Q/FU7561Q/FU7512L/FU7512P): PWM output, measurement of duty cycle and period of input PWM wave, measurement of the time of set PWM wave numbers, QEP decoding, tailwind/headwind detection, rotation direction and speed detection of step motor
  • Timer5/Timer6 (FU7571Q/FU7541Q): PWM output, measurement of duty cycle and period of input PWM wave, measurement of the time of set PWM wave numbers, and tailwind/headwind detection
  • Timer2/Timer6 (FU7562L): Basic timer
  • Timer2 (FU7571Q/FU7541Q): Basic timer
  • Timer3/Timer4/Timer7/Timer8: PWM output, and measurement of duty cycle and period of input PWM wave. Timer4 supports FG generation and Timer3/Timer8 supports up to 96MHz input.
  • Systick Timer
  • RTC
  • IWDT, WWDT
• System clock
  • Built-in 24MHz high-speed RC oscillator
  • Built-in 32.8kHz low-speed RC oscillator
  • 12MHz external crystal clock
• Power Supply
  • FU7571Q/FU7561Q – 5V ~ 20V
  • FU7562Q/FU7562L – 10V ~ 20V and 3V ~ 5.5V
  • FU7541Q – 5V ~ 28V
  • FU7512L/FU7512P – 3V ~ 5.5V
  • Low-voltage detector (LVD)
• Packages
  • FU7571Q  – QFN64
  • FU7562Q – QFN64-56
  • FU7562L – LQFP52
  • FU7561Q – QFN56
  • FU7541Q – QFN48
  • FU7512L (LQFP64), FU7512P (LQFP48)

Nine SKUs are currently available with different drive types, number of I/Os, packages, and target applications.

The company appears to provide a development board (pictured below), but I could not find public details.

Having said that, we found out about the new FU75xx family through Banana Pi’s upcoming BPI-BJ2403N Servo motor controller based on the FU7512L SKU, and they’ve shared the datasheet for the FU75XX microcontrollers and a GitHub repo with source code (C language) and binaries that relies on Fortior Tech’s FTM32ForIDE. We’ll cover the Banana Pi in detail once it’s up for sale.

The shocker is when I check the price on JLCPCB: about 2 cents for the FU7512L, and other FU75xx parts when buying a few hundred pieces. In the screenshot below, it costs $8.97 to purchase 444 FU7512L chips, or $0.0202 per unit. Related products are shown to be $0.0392 per unit, but that’s for a single unit, and the minimum order is 444 pieces…

That’s only for people who purchase PCBs from JLCPC, but that’s still impressive. I double-checked the Raspberry Pi RP2040 price to make sure I didn’t misunderstand pricing, and it goes for 71 cents as expected, but I still feel the 2 – 4 cents price for the Fortior chips is likely a placeholder rather than the actual price. Additional information may also be found in the press release on the Fortior Tech website.

The post Fortior FU75xx dual-core motor control MCU family combines 32-bit RISC-V core with 2nd-gen Motor Engine (ME2) core appeared first on CNX Software - Embedded Systems News.

Seeed Studio Wio-S3 is a compact (21.6 x 16.5 x 3.3 mm) wireless module that combines an ESP32-S3R8 dual-core WiFi 4 and Bluetooth LE MCU and a Semtech SX1262 LoRa transceiver.

The module includes 16MB Flash and 8MB PSRAM, and supports LoRa (EU868/US915) with up to +20.9 dBm transmit power and -137 dBm sensitivity. It also supports Wi-Fi 4 and BLE 5.0, and comes with two IPEX connectors for external antennas. With interfaces such as UART, I2C, SPI, ADC, and USB, and support for -40°C to 85°C operation, it is designed for remote monitoring, industrial automation, smart agriculture, and IoT data logging.

Seeed Studio Wio-S3 module specifications:

• SoC – Espressif ESP32-S3R8
  • CPU – Dual-core Tensilica LX7 microcontroller up to 240 MHz with vector instructions for AI acceleration
  • Memory – 8MB PSRAM
  • Wireless – 2.4 GHz 802.11b/g/n WiFi 4 and Bluetooth 5.0 LE + Mesh connectivity
• Storage – 16MB external SPI flash (option for 32MB)
• Wireless
  • LoRa via Semtech SX1262 with support for global frequency plans including EU868, US915, AS923, AU915, KR920, and IN865
  • Up to 20.9 dBm LoRa output power and -137 dBm sensitivity (@SF12, BW125kHz)
  • 2.4 GHz 802.11b/g/n WiFi 4 and Bluetooth 5.0 LE + Mesh connectivity via ESP32-S3
  • Integrated 32MHz TCXO for RF frequency stability
  • 2x IPEX antenna connectors
• Expansion – 38-pin SMT package exposing up to 25x GPIOs, 3x UART, 2x I2C, SPI, USB (D+/D-), ADC, PWM, I2S, and CAN (TWAI)
• Power
  • Supply – 3.0V to 3.6V (3.3V typical)
  • Consumption – Rx: ~5.5 mA; Tx: ~113–127 mA; Sleep: ~9.3 µA
• Dimensions – 21.6 x 16.5 x 3.3 mm
• Temperature Range – -40°C to 85°C
• Certifications – FCC, CE, Telec Certified

The module comes in two variants: the 100020327 SKU features an onboard IPEX connector for plug-and-play antenna attachment, while the 100079384 SKU features RF pads instead, to allow developers to route custom antenna paths or connect external RF front-ends.

On the software side, the Wio-S3 can be programmed with the ESP-IDF framework or Arduino. The company also mentions that it ships with pre-flashed command firmware for quick testing, and you can also build your own applications using Wi-Fi, BLE, and LoRaWAN. It also integrates with platforms like The Things Network and ChirpStack to easily get started with IoT projects.

At the time of writing, I don’t see a development board for the module, but the company provides a reference design instead of a full dev kit. It shows how to build a combined LoRa and Wi-Fi/BLE node. It includes a diagram with dual USB-C power inputs, a 3.3V regulator setup, and basic connections for programming like UART, reset, and boot pins. More technical details, including KiCad design files and STEP 3D models, libraries, and relevant SDKs, are available on the wiki. They also provide a separate module command list page that lists all the System Commands.

Previously, we have written about various development boards, industrial gateways, and handheld communicators that combine an ESP32 with a LoRa module, such as the OBJEX Link S3LW, the EBYTE ECM50-A, and the Heltec WiFi LoRa 32 (V4), but these are typically full-fledged development boards or designed for specific applications. The Wio-S3, on the other hand, is a low-cost 38-pin surface-mount module designed for direct integration into custom smart agriculture, industrial IoT, and edge AI hardware.

The Wio-S3 wireless module is available for $7.99 on the Seeed Studio store, and may soon show up on the company’s AliExpress store.

The post $7.99 Seeed Studio Wio-S3 WiFi, Bluetooth LE, and LoRa IoT module combines ESP32-S3 and SX1262 RF transceiver appeared first on CNX Software - Embedded Systems News.

SONOFF “Basic DIN Rail”, codenamed BASIC-1GSP, is a Matter Over WiFi smart switch designed to be installed in electrical distribution boxes with a DIN rail, and it also supports real-time energy monitoring to track voltage, current, power, and energy consumption.

The BASIC-1GSP can handle high-power circuits up to 32A (7,680W @ 240V AC), which makes it suitable for electric floor heating, EV charging stations, water heaters, and other heavy-duty devices. The device also focuses on safety with customizable overload protection & a double-pole safety switch.

SONOFF BASIC-1GSP specifications:

• SoC – Beken BK7238 wireless microcontroller as found in SONOFF Basic Gen5 and Tx Gen2
  • Core – 32-bit Arm Cortex-M MCU at up to 160 MHz
  • Memory – 288 KB RAM
  • Storage – 2 MB flash (SiP)
  • Wireless – 2.4 GHz 802.11b/g/n WiFi 4, Bluetooth LE 5.2, Matter
• Screw terminals for AC input and load
  • Wire gauge range – 11 AWG to 6 AWG (4 mm² to 10 mm²) SOL/STR copper conductors only
  • Max. load – 7,680 Watts
  • Rating – 100-240V~ 50/60Hz 32A Max
• Safety
  • Double-pole switching for full physical isolation
  • Overload protection; shutdowns within 3 seconds
• Misc
  • Button for power on/off and pairing
  • Red LED for relay status (Out)
  • Blue LED for network status
  • Real-time monitoring of voltage, current, power & energy usage
• Dimensions – 84 x 66.5 x 18 mm
• Weight – 92.1 grams
• Casing material – PC
• Temperature Range – -15°C to 50°C
• Humidity – 10% ~ 90% RH, non-condensing
• Certifications – CE, RoHS, FCC, CP65

The SONOFF BASIC-1GSP is supported by the eWeLink mobile app, and since it works over Matter, you can scan the Matter QR code on the Quick Guide or the device itself to add the device to Apple Home, Google Home, Amazon Alexa, Samsung SmartThings, Home Assistant, or other Matter-compatible apps.

However, note that some features like OTA firmware upgrades are only supported in the eWelink app, and power/energy monitoring is currently limited to eWelink, Home Assistant, and Samsung SmartThings platforms. The SONOFF Basic DIN Rail smart switch can also be used in LAN mode without access to the Internet, and supports features such as voice control, schedules, inching, and basic remote control (on/off). The user manual provides instructions to get started.

The device is not the first DIN rail-compatible switch from SONOFF, as the earlier 4CH Pro R3 switch offered a larger 4-gang switch, and the company also sells the SONOFF DR DIN Rail Tray as an accessory bracket for mounting various SONOFF modules. The BASIC-1GSP is still unique as a compact DIN Rail smart switch with Matter over WiFi connectivity.

SONOFF sells the BASIC-1GSP for $19.90 on its online store. The coupon code CNXSOFT can lower that price by 10 percent to $17.91.

The post SONOFF BASIC-1GSP Matter over WiFi smart switch supports DIN rail mounting, real-time energy monitoring appeared first on CNX Software - Embedded Systems News.

Ivan Svarkovsky’s S3-MSX-PC open-source project implements a bare-metal MSX2+ emulator running on an ESP32-S3 microcontroller and outputting 64-color VGA via a simple R-2R resistor ladder. It’s a fork of the Retro-Go emulator for ODROID-GO and other ESP32 devices, but with various optimizations.

It was tested on an off-the-shelf ESP32-S3 board with one core handling the game logic and the other video and audio output. VGA is implemented through a clever resistor network that converts digital data into an analog signal that any old monitor understands, while audio relies on Sigma-Delta modulation with a multi-stage PDM filter. The USB host port on the board allows for the connection of a keyboard.

S3-MSX-PC firmware highlights:

• Emulation Core – fMSX 6.0 — Full MSX1 / MSX2 / MSX2+ support
• Video Output
  • VGA 640×480@60Hz, 16-bit parallel RGB via LCD_CAM
  • Color Depth – 64 colors (2 bits per channel: R, G, B)
• Audio – PDM Stereo with hardware underrun protection
• USB Host
  • Plug-and-play keyboards
  • Input Latency – ~2–4 ms (USB interrupt-driven, software debounce bypassed)
• ESP32-S3 optimization
  • Z80 Core Optimization – The standard switch/case opcode dispatcher kills the Xtensa branch predictor, so Ivan implemented Computed Gotos (Threaded Code) instead to pin the PC and ICount variables directly to physical 32-bit registers.
  • MSX palettes are pre-shifted during init to map directly to the GPIO pins of the R-2R ladder. The ESP32’s LCD_CAM peripheral is used to DMA-blast pixels to the monitor, utilizing the hardware lcd_byte_order=1 flag for zero-CPU-cost byte swapping.
  • Fighting PSRAM Latency – Ivan integrated cycle-accurate studio audio chips (PSG, SCC, OPLL), but removed 400KB of PSRAM lookup tables. They were replaced with 1-cycle integer bit-shifts, and the core 11KB tables were locked into DRAM_ATTR.
  • Zero-Heap State Compression – Standard zlib crashes the ESP32-S3 when saving 4.4MB machine states, so Ivan wrote a custom streaming LZ77 compressor (“Delta-Stride LZ”) that understands the MSX VRAM geometry (128/256-byte strides). It performs a vertical rep-match XOR-delta pass and streams directly to the SD card, using exactly 0 bytes of heap RAM.

The project uses the ESP-IDF v5.4.4 framework. The S3-MSX-PC firmware was tested on a DIYables ESP32-S3 development board with an ESP32-S3-WROOM-1-N16R8 module ($4 on AliExpress, $12 on Amazon), but it should work on most other ESP32-S3 boards with similar flash and PSRAM capacity. You’ll also need to add a few resistors and capacitors for VGA video output and the PDM audio filter, plus the USB and VGA connectors wired as shown in the schematic below.

The result is a bit messy, although it could probably be cleaned up using a breakboard. I initially thought something like the Olimex ESP32-SBC-FabGL board might be a better option since there’s already a VGA connector. However, it’s based on ESP32 with LX6 cores, and Ivan told us that “getting standard emulator code (based on fMSX) to run smoothly on the ESP32-S3 required heavily modifying the architecture to bypass PSRAM latency and Xtensa LX7 pipeline stalls”. So any LX7-specific code might not work.

You’ll find the source code, binary release, instructions, and a “technical deep dive” with code snippets and compilation metrics on GitHub. Oh, and you can also watch a few short videos there proving the firmware indeed works…

The post Bare-metal MSX2+ Emulator for ESP32-S3 offers custom LCD_CAM VGA implementation & Z80 optimizations appeared first on CNX Software - Embedded Systems News.

Designed by Indonesian maker Ahmad Amarullah (amarullz), the piBrick PocketCM5 is an open-source hardware handheld Linux computer kit built around the Raspberry Pi CM5; it’s basically a smartphone-sized Linux machine with a physical keyboard and touchscreen. It is for developers, makers, and system administrators, for tasks such as general experimentation, embedded development, and remote access.

We have seen other handheld terminals and pocket computers based on Raspberry Pi SBCs and Compute Modules over the years, such as the PocketTerm35,  DevTerm, Carbon’s CyberT, Pi Slate, and many others. However, the piBrick PocketCM5 is based on the latest Raspberry Pi CM5 and features a 3.92-inch AMOLED touchscreen display, a physical BlackBerry QWERTY keyboard, and various expansion options.

piBrick PocketCM5 specifications:

• Compatibility – Raspberry Pi Compute Module 5 (CM5 and CM5 Lite)
• Auxiliary MCU – Raspberry Pi RP2040 for keyboard/trackpad input, rotary encoders and buttons, USB HID (keyboard/mouse) emulation, and accelerometer data processing
• Storage
  • MicroSD card slot
  • M.2 NVMe socket for SSDs (2230 or 2242 sizes)
• Display
  • 3.92-inch AMOLED touchscreen (1080×1240, 90Hz, 500 nits) with 5-point touch and Asahi glass
  • MIPI DSI interface
• Video Output
  • Full-size HDMI output
  • Micro HDMI output
• Camera I/F – Front camera support via MIPI CSI interface (compatible with Raspberry Pi Zero Camera modules)
• Audio
  • Integrated USB sound card with amplifier
  • Built-in stereo speakers
  • 3.5mm headphone jack
  • Microphone (integrated via the BBQ20 keyboard)
• Networking – Dual-band Wi-Fi 5 (802.11ac) and Bluetooth 5.0/BLE support via Pi CM5’s wireless module
• USB
  • USB 3.0 Type-C port with charging support
  • USB Type-C 2.0 port for peripherals and charging
  • USB 3.0 Type-A port
  • USB 2.0 Type-A port
  • Internal USB 2.0 expansion header
• User Input
  • BlackBerry (BBQ20) QWERTY keyboard with integrated trackpad (serves as a mouse)
  • Side rotary encoders with push switches
  • 5x Volume, Display brightness, and user buttons on the side
  • Internal USB button
• Expansion – I2C connector, GPIO extension breakout
• Misc
  • Built-in accelerometer
  • RTC with battery
• Power
  • USB Type-C port for charging and power
  • 5000 mAh LiPo battery (recommended dimensions: 80x50x10mm) connected through a JST-PH 2.0mm connector
• Dimensions – 145 x 80 x 19.6 mm (Device size)
• Weight – ~500 grams (approximate shipping weight)

The piBrick PocketCM5 runs standard Raspberry Pi OS and other Linux distributions on the CM5, with full desktop support. The RP2040 converts keyboard, trackpad, and control inputs into standard USB HID signals, so no custom drivers are required, and its open-source firmware allows custom key mapping and input adjustments.

The hardware is fully open-source, designed with EasyEDA Pro. The estimated cost is around $172 for the main components. Design files, schematics, STL enclosure files, and firmware are released under the GPL-3.0 license, with documentation and assembly resources available via GitHub and the OSHWLab website.

Until now, you had to build it yourself, but Ahmad is now selling the standard kit for $240 on Tindie with a fully assembled piBrick mainboard (PCBA), an AMOLED display with a flex PCB, a speaker board, a physical keyboard, and an SLA-3D-printed enclosure set with mounting hardware. A “Full Camera Kit” is also available, which bundles a Raspberry Pi Camera Zero module.

It is important to note that this Raspberry Pi CM5 handheld Linux computer is sold as a hardware-and-enclosure kit, not a complete plug-and-play consumer device. You must have your own Raspberry Pi CM5, an active or passive heatsink, a LiPo battery, NVMe/MicroSD storage, and the operating system (Raspberry Pi OS or compatible Linux distributions).

The post piBrick PocketCM5 – An open-source handheld Linux computer kit for Raspberry Pi CM5 appeared first on CNX Software - Embedded Systems News.

PZSDR P047 RF-ADC and RF-DAC software-defined radio (SDR) board is based on AMD Zynq UltraScale+ ZU47DR RFSoC with a quad-core Cortex-A53 processor, a dual-core Cortex-R5F real-time CPU, FPGA fabric, eight RF ADCs, and eight RF DACs.

The board features 6GB of DDR4 memory (4GB for PS, 2GB for PL), 8 GB eMMC flash, and 512 Mbit QSPI flash, plus microSD card and M.2 NVMe sockets for storage, a mini DP video output, a Gigabit Ethernet RJ45 port, two 100 Gbps QSFP28 cages, a USB 3.0 port, plus JTAG and serial debug interfaces, and a 32-pin IO header.  It makes use of all RF channels of the RFSoC with 8 receiver and 8 transmitter connectors, plus various clock inputs. The company says the PZSDR P047 is ideal for advanced MIMO wireless applications such as cellular and shortwave communication.

PZSDR P047 RF-ADC and RF-DAC specifications:

• SoC FPGA – AMD Xilinx Zynq UltraScale+ ZU47DR (XCZU47DR-2FFVE1156I)
  • Application processor – Quad-core Arm Cortex-A53 with NEON and Single/Double Precision Floating Point; 32KB/32KB L1 Cache, 1MB L2 Cache
  • Real-Time Processor – Dual-core Arm Cortex-R5F with Single/Double Precision Floating Point; 32KB/32KB L1 Cache, and TCM
  • FPGA
    • 930,300 logic cells
    • 13 Mbit distributed RAM
    • 1080 Block RAM blocks
    • 38 Nvit Block RAM
    • 80 UltraRAM blocks
    • 22.5 Mbit UltraRAM
    • 4272 DSP slices
  • Memory – 256KB On-Chip Memory w/ECC
  • Standard interfaces – 214x PS I/O, UART,  CAN,  USB 2.0,  I2C,  SPI, 32bit GPIO
  • High-speed interfaces – 4x PS-GTR, PCIe Gen1/2, SATA 3.1, DisplayPort 1.2a, USB 3.0, SGMII
  • RF – 8x 14-bit DAC, 8x 14-bit ADC
• System Memory – 6GB DDR4; 4 GB DDR4 for PS, 2 GB DDR4 for PL.
• Storage
  • 8 GB eMMC flash
  • 512 Mbit QSPI flash
  • Optional NVMe SSD via M.2 Key-M socket
  • MicroSD card slot
• RF signals
  • 8x ADC connectors; 14-bit resolution, sample rate up to 5 Gsps
  • 8x DAC connectors; 14-bit resolution, sample rate up to 9.85 Gsps
  • Up to 6 GHz RF input
• Clock inputs
  • SYNC connector – Synchronization signal
  • CLK-P1/CLK-N1 connectors – 156.25 MHz differential clock for the MGT (Multi-Gigabit Transceiver) section
  • CLK-P0/CLK-N0 connectors – 125 MHz differential clock for the MGT section
  • CLK-IN connector – External high-precision clock input (if not used, it defaults to the on-board 19.2MHz OCXO)
• Video Output – Mini DP (PS)
• Networking and transceivers
  • Gigabit Ethernet RJ45 port (PS)
  • 2x QSFP28 cages connected to 100 Gbps GTY transceivers
• GNSS – GPS and Beidou with PPS signal
• USB – USB 3.0 Type-A port (PS)
• Serial – UART/RS232 connector (PS)
• Debugging – JTAG/UART USB-C connector
• Expansion – 32-pin GPIO header with 20x I/Os at 3.3V
• Misc – Reset key, boot mode selection switches (JTAG, QSPI, SD card, or eMMC)
• Power Supply – 12V/3A via DC jack or internal header
• Dimensions
  • Board – 180 x 175 mm
  • Enclosure – 188 x 183 x 52 mm
• Temperature Range – -40°C to +85°C

Puzhi promises that backers will have access to FPGA logic, Arm Vitis, PetaLinux, RF, and PYNQ demos (see video below for the latter) to help them get started, and the PDF version of the schematics will be released after the crowdfunding campaign is complete. In the meantime, they have released a user manual, which mainly provides hardware details.

Last time we wrote about an AMD Zynq RFSoC platform was in 2022, when we covered the Avnet XRF16 SoM based on Zynq UltraScale+ Gen3 ZU49DR RFSoC and associated carrier board. The ZU49DR is similar to the ZU47DR used here, except it doubles the number of ADC and DAC channels to sixteen. At the time, the SoM sold for a cool $24,995, and the carrier board for $4,995.

By comparison, the Puzhi PZSDR P047 feels like a bargain with a pledge of $6,699 through Crowd Supply. The reward includes the board itself, a metal enclosure, a 12V/3A power adapter, a GPS antenna, 23 RF cables, a USB Type-C cable, a two-sided USB 3.0 cable, a Gigabit Ethernet cable, a 3-pin XH2.54 cable, a 32 GB microSD card, and an SD card reader. The price includes worldwide shipping, and deliveries are scheduled to start by November 2026. I’m not sure why the company decided to go with Crowd Supply here, since they already have a similar ZU47DR board listed for $5,545.41 on AliExpress, including shipping.

The post PZSDR P047 RF-ADC and RF-DAC high-end SDR board is based on AMD Zynq UltraScale+ ZU47DR RFSoC (Crowdfunding) appeared first on CNX Software - Embedded Systems News.

Synaptics has introduced the Astra SRW1500 Series of AI-native Edge AI MCUs designed for smart IoT edge devices. The family features an Arm Cortex-M52 core, an optional Arm Ethos-U55 Neural Processing Unit (NPU), and tri-band Wi-Fi 7 or Wi-Fi 6, Bluetooth 6.0 LE, and 802.15.4 connectivity, in a single package. It targets low-latency on-device AI workloads such as voice-trigger detection, sound-event classification, and AI-enhanced Wi-Fi sensing.

The series currently comes in four variants offering different wireless and USB options. Since sensing, AI processing, and wireless connectivity are integrated into a single chip, this approach helps reduce system costs, simplify PCB design, and lower power consumption in devices such as smart home hubs, building automation systems, and home appliances.

Synaptics SRW1500 Series specifications:

• CPU
  • Arm Cortex-M52 application core up to 200 MHz with Arm Helium (MVE) vector extensions for DSP and ML
  • Arm Cortex-R4 WiFi core with 960 KB SRAM and 1408 KB ROM
  • Arm Cortex-M4 BLE/Thread core with 512 KB System RAM and 1024 KB ROM
• AI – Arm Ethos-U55 NPU @ 200 MHz; up to 50 GOPS ( SRW1563 and SRW1573 only)
• Memory – 1 MB SRAM, 128 KB ITCM, 128 KB DTCM
• Storage
  • 1408 KB ROM, 32 KB on-chip anti-fuse OTP
  • Octal SPI (xSPI) interface
  • SDIO v3.0 Host/Target interface (backward compatible with v2.0)
• Wireless connectivity
  • Wi-Fi – Up to tri-band (2.4/5/6 GHz) Wi-Fi 7/802.11be (SRW1571 and SRW1573) 1×1 SISO; Wi-Fi 6/802.11ax on SRW1561 and SRW1563
  • Bluetooth – Bluetooth 6.0 with Low Energy (LE) long-range support
  • 802.15.4  Radio for – Thread and Zigbee support with full Matter compliance
  • Synaptics Smart Co-Ex technology for optimized 2.4 GHz Wi-Fi and Bluetooth coexistence
• Audio
  • I2S interface supporting simultaneous RX/TX
  • PDM interface supporting up to 2x digital microphones
• USB – USB 2.0 dual-role (DRD) interface supporting LS/FS/HS up to 480 Mbps (SRW1563 and SRW1573 only)
• Other I/O
  • Up to 32x multiplexed GPIOs
  • 3x UART (2x with flow control, 1x dedicated for debug)
  • 1x SPI Host (up to 25 MHz), 1x SPI Target
  • 1x I2C Host, 1x I2C Target
  • 6x PWM channels supporting Motor Control and LED Brightness
  • 8-channel 12-bit General-purpose SAR ADC (Up to 1.536 Msps)
  • General-purpose DAC
• Security
  • SynaPROT Secure Island Root of Trust (PSA Certified Level 3)
  • Hardware Cryptographic Accelerators – AES, DES, 3DES, RSA, SHA, MD5, and ECC
  • Arm TrustZone, Secure Boot, Firmware encryption, TZC/MPC/PPC memory isolation, and Anti-rollback protection
• Power
  • Operating Voltage – 3.0V to 5.0V VBAT input, with 0.8V/0.9V standard core voltage support
  • Integrated PMU with Buck and LDO regulators
  • Independent ON/OFF power-saving modes for WLAN, Bluetooth, and MCU domains
• Package – 115-lead DR-QFN, 10.5 x 8.6 x 0.9 mm
• Temperature Range – 0°C to +70°C ambient / -40°C to +85°C junction temperature

The SRW1500 Series of single-chip Edge AI MCUs is split into Wi-Fi 6 (SRW1560) and Wi-Fi 7 (SRW1570) families. Specific part numbers dictate the presence of the Ethos NPU and USB hardware.

On the software side, the SRW1500 uses the Astra SDK (also referred to as SR SDK for the MCU series) with support for FreeRTOS and Zephyr RTOS, along with pre-integrated Wi-Fi 6/7, Bluetooth 6.0, 802.15.4 connectivity stacks, and Matter support. The platform includes a PSA Level 3-certified security stack with secure boot and hardware-based cryptography. For AI workloads on the Arm Ethos-U55 NPU, Synaptics provides the SyNAP toolkit for model optimization and quantization (including TensorFlow Lite). More information can be found on the Synaptics Developer Portal and the Astra Documentation Hub.

Previously, Synaptics introduced the Astra SR series based on the Arm Cortex-M55 with the Ethos-U55 NPU for edge AI, and also introduced the Veros SYN4383 / SYN43756(E) wireless SoCs with Wi-Fi 6E, Bluetooth 5.4, and 802.15.4. The SR series focused on AI processing but lacked built-in wireless, whereas the Veros chips provided connectivity but required an external host processor. The new Astra SRW1500 series combines both into a single chip, reducing the need for separate compute and wireless components.

The company did not mention pricing for the new Astra SRW1500 series. More information can be found on the product page and in the press release.

The post Synaptics Astra SRW1500 Cortex-M52 Edge AI MCU features Ethos-U55 NPU, Wi-Fi 6/7, Bluetooth 6.0, 802.15.4 connectivity appeared first on CNX Software - Embedded Systems News.

FriendlyELEC’s NanoPi M6V2 is an update to the NanoPi M6 Rockchip RK3588S SBC, whose main change is a 4-pin connector for dual analog microphone input. The RAM is also now fixed to 8GB (no more 4GB, 16GB, or 32GB options), some buttons have changed, and the company has stopped offering an enclosure with a built-in 3.5-inch display.

The rest of the specifications remain the same, with LPDDR5 memory, support for microSD, eMMC flash, or NVMe storage, HDMI 2.1 and MIPI DSI display interfaces, MIPI CSI camera inputs, Gigabit Ethernet and optional M.2 WiFi/Bluetooth connectivity, a few USB ports, and a 30-pin GPIO header.

NanoPi M6V2 specifications:

• SoC – Rockchip RK3588S
  • CPU – Octa-core processor with 4x Cortex-A76 cores @ up to 2.4 GHz, 4x Cortex-A55 cores @ up to 1.8 GHz
  • GPU – Arm Mali-G610 MP4 GPU compatible with OpenGL ES 3.2, OpenCL 2.2, and Vulkan 1.2 APIs
  • VPU – 8Kp60 video decoder for H.265/AVS2/VP9/H.264/AV1 codecs, 8Kp30 H.265/H.264 video encoder
  • AI accelerator – 6 TOPS NPU
• System Memory – 8GB LPDDR5 at 2400 MHz
  
• Storage
  • eMMC module socket supporting HS400 mode
  • MicroSD card slot with SDR104 mode support
  • M.2 M-Key socket with PCIe 2.1 x1 for NVMe SSDs
• Video output
  • HDMI 2.1 port up to 7680×4320 @ 60Hz; RGB/YUV(up to 10bit) format
  • 2x 4-lane MIPI-DSI connectors
• Optional Display – 3.5-inch touchscreen LCD with 800×480 resolution and capacitive touchscreen connected to one of the MIPI DSI connectors
• Camera input
  • 4-lane MIPI-CSI V1.2 connector
  • 4-lane MIPI-CSI D/C PHY connector
• Audio
  • 3.5mm jack for stereo headphone output
  • 4-pin connector for dual analog microphone input
• Networking
  • Gigabit Ethernet RJ45 port via RTL8211F PHY
  • Optional WiFi and Bluetooth via M.2 E-key socket with PCIe 2.1 x1 and USB 2.0 host
• USB
  • 1x USB 3.0 Type-A port
  • 2x USB 2.0 Type-A ports
• Expansion
  • M.2 Key-M socket for NVMe storage
  • M.2 Key-E socket for WiFi/Bluetooth (PCIe/USB)
  • 30-pin 2.54mm GPIO header with up to 1x SPI, 6x UART, 3x I2C, 1x SPDIF, 4x PWM, 20x GPIO
• Debugging – 3-pin header for serial console
• Misc
  • 2x user LEDs
  • 2-pin RTC battery input connector for low-power HYM8563TS RTC IC
  • Buttons – MASK, User (replaces Reset), Recovery, Power
  • 5V fan connector
• Power supply – 6V~20V input via USB-C with PD support
• Dimensions & Weight
  • SBC – 90 x 62 mm (8-layer PCB) | 52 grams
  • Metal case – 94.5 x 68 x 30mm | 252 grams
  • Metal case + LCD – 99 x 68 x 31mm | 275 grams
• Temperature Range – 0°C to 70°C

Some connectors and buttons have also been moved around as shown in the comparison above. The company doesn’t provide any accessories for the new 4-pin dual analog microphone connector, so you’d have to find something on your own and wire it yourself.

FriendlyELEC provides a long list of supported operating systems based on Linux 6.1, namely Alpine Linux 3.23, Android 14 Tablet/TV, Buildroot – Weston (Wayland), Debian 12 Desktop XFCE (X11),  Debian 13 Core, Debian 13 Desktop GNOME (Wayland), FriendlyCore 20.04 (Qt5), FriendlyWrt 25.12, OpenMediaVault 8.0.6, Proxmox VE 8.2.7, Ubuntu 22.04 Desktop XFCE (X11), and Ubuntu 24.04 Desktop GNOME (Wayland). The board is also supported by Armbian with Platinum support (Ubuntu 26.04/Debian 13), so you may consider that one if you are just getting started. Information to get started can be found on the wiki.

The NanoPi M6V2 sells for $172 on the FriendlyELEC store as a bare single board computer, and the optional metal case adds $15. Other accessories include a 4K MIPI camera module ($15), an RTL8822CE WiFi module ($6.90), a 32GB microSD card ($15.99), and 64GB or 256GB eMMC flash modules ($23/$61). In the good old times (2024), the NanoPi M6 started at  $70 with 4GB of RAM…

The post NanoPi M6V2 RK3588S SBC gains support for dual analog microphone input appeared first on CNX Software - Embedded Systems News.

OpenCV 5 open-source computer vision library has recently been released with a brand-new DNN (Deep Neural Network) engine that provides better ONNX coverage and enables LLM/VLM support.

The fifth version of the popular CV library also adds support for Intel, Arm, Qualcomm, and RISC-V hardware acceleration, improved 3D vision, and various new core features such as new data types, real N-dimensional and scalar support, and performance improvements.

OpenCV 5’s DNN Engine

OpenCV 4.x supports about 22% of ONNX operators, and the new DNN engine in OpenCV 5 brings coverage to over 80%.  That means models with dynamic shapes that used to fail on OpenCV 4.x, should now work, as the 5.x engine was rebuilt around a typed operation graph with proper shape inference, constant folding, and operator fusion.

The table below shows the main difference between OpenCV 4.x and OpenCV 5

Since it’s quite a big change, to make sure nothing breaks, you can select four engine options:

• ENGINE_CLASSIC (1) – Force the old 4.x-style engine. Supports non-CPU backends and targets such as CUDA and OpenVINO.
• ENGINE_NEW (2) – Force the new graph engine, with fusion and dynamic shapes. Important: the new engine is CPU-only for now, although work on GPU acceleration for inference and non-CPU HAL for pre-/post-processing has started for the new DNN
• ENGINE_AUTO (3) – The default. Try the new engine first, and fall back to the classic engine if the model fails to load.
• ENGINE_ORT (4) – Use the bundled ONNX Runtime wrapper. ONNX models only, and the build must be configured with WITH_ONNXRUNTIME=ON.

While the CV in OpenCV stands for Computer Vision, OpenCV 5 can also run large language models (LLMs) and vision-language models (VLMs) directly inside the DNN module, with no separate runtime thanks to a native tokenizer and KV-cache for autoregressive decoding. Qwen 2.5, Gemma 3, PaliGemma, and the GPT-2 / GPT-4 family models work through the same Net API used for a YOLO model.

In practical terms, that means various Detection, segmentation, vision-language, and generative AI models are now supported in OpenCV 5.

Performance also looks fairly good, as shown in the table below comparing OpenCV 5’s DNN to the ONNX runtime on an Intel Core i9-14900KS machine running Ubuntu 24.04 LTS.

You’ll find more benchmarks on various platforms in the wiki on GitHub.

New Core features

Besides deep learning changes, the OpenCV core also got improvements:

• New data types. OpenCV 5 adds first-class FP16 (cv::hfloat, CV_16F) and BF16 (cv::bfloat, CV_16BF) types, plus bool, 64-bit integers, and more.
• Real N-dimensional and scalar support. cv::Mat can now represent 0D (scalar) and 1D arrays; OpenCV 5 also adds broadcasting and first-class N-D operations like transposeND and flipND
• Better performance. Up to 2x improvements on mathematical workloads, and the same code now runs across CPUs and accelerators without modification.

Some cleanups were made on the language side:

• The legacy C API is officially deprecated
• C++17 is now the minimum recommended standard, with C++20 modules planned for later 5.x releases.
• Python – NumPy 2.x support, deeper integration, and named (keyword) arguments for C++ algorithms, so you can write cv.someAlgorithm(threshold=0.5) instead of memorizing positional order.

Hardware acceleration

OpenCV 5 benefits from a redesigned Hardware Acceleration Layer (HAL) that relies on Universal Intrinsics 2.0, a single vector codebase that maps to SSE, AVX2/512, NEON, SVE, RVV, and more. It supports the following accelerators:

• Intel IPP (IPPICV) – Restructured from the original x86/x64 acceleration path. A free subset (ICV) ships by default, dispatching to SSE/AVX-optimized kernels for filtering, color conversion, and geometric transforms.
• Arm KleidiCV – HAL for AArch64 that accelerates core image processing and DNN kernels using NEON, SVE, and SME, validated on AWS Graviton 4 and Cortex-A chips. Enabled automatically on supported operations; up to 3-4x speedups measured on Arm operations like resizing and warping.
• Qualcomm FastCV – Acceleration on Snapdragon targets through the Hexagon DSP and NPU.
• RISC-V Vector (RVV) – Scalable vector support mostly enabled by the OpenCV China organization.

It’s the same high-level code for all four, and OpenCV automatically uses the best path on the hardware it runs on.

Better 3D Vision

OpenCV 5’s 3D vision supports multi-camera calibration, point cloud and mesh I/O, dense RGB-D fusion. It implements three modules:

• 3d: basic 3D geometry and vision, including I/O, geometry primitives, algorithms like ICP, and parts of SLAM.
• calib: camera calibration, including single-camera calibration and a refactored multi-camera pipeline.
• stereo: depth from stereo.

The developers claim it’s a meaningful upgrade for people working on structure-from-motion, robotics, or any kind of reconstruction.

You’ll find fore more details about OpenCV 5 in the announcement and the revamped documentation website, while the source code is hosted on GitHub.   The announcement mentions that the “pip version of OpenCV 5 will be released on 8th June”, but I can only see “4.13.0.92” so far, so you’d still need to build it from source for now.

The post OpenCV 5 release – New DNN engine with enhanced ONNX and LLM/VLM support, Intel, Arm, and RISC-V hardware optimizations appeared first on CNX Software - Embedded Systems News.

Altera’s wideband Agilex 9 Direct RF-Series AGRW039 SoC FPGA is designed for high-performance Radio Frequency (RF) systems in aerospace, defense, and advanced communications systems.

It is the fourth device in Altera’s Agilex 9 Direct RF-Series, and compared to the previous generation AGRW027, the new AGRW039 delivers about 45% more logic resources, 45% more DSP resources, and 43% more block RAM. It also integrates a 64Gsps wideband RF, a hard quad-core Cortex-A53 processor, and supports DDR5 and LPDDR5 memory.

Agilex 9 Direct RF AGRW039 SoC FPGA key features and specifications:

• Hardware CPU – Quad-core Arm Cortex-A53 processor @ up to 1.41 GHz
• FPGA fabric
  • 3.9M logic elements (LEs)
  • 1.3M adaptive logic modules (ALMs)
  • 18,960 M20K memory Blocks
  • 370 Mb M20K memory Size
  • 65,280 MLAB memory count
  • 40 Mb MLAB memory size
  • 24,600 18×19 multipliers
  • 12,300 DSP/AI blocks
  • Single Precision – 18.4 TFLOPS
• Memory I/F – DDR5, LPDDR5, DDR4
• Peripherals
  • PS – 2x USB 2.0, 3x Gigabit Ethernet (EMAC), 2x UART, 4x SPI, 5x I2C, 7x general purpose timers, 4x watchdog timers
  • 2x 8-channel ADC/DACs
    • 64 Gsps sample rate, 10-bit resolution
    • RF Frequency Range: 0.1–36 GHz
    • ADC Rx Max IBW: 32 GHz
    • DAC Tx Max IBW: 6.4 GHz
  • Up to 24x 58G transceivers
  • 2x F-Tiles
    • PCIe hard IP controller (4.0 x16) or Bifurcatable 2x PCIe 4.0 x8 (EP) or 4x 4.0 x4 (RP)
    • Transceiver channel count – 16 channels at 32 Gbps (NRZ) /12 channels at 58 Gbps (PAM4) – RS & KP FEC
    • Advanced networking support:
      • Bifurcatable 400 GbE hard IP block (10/25/50/100/200/400 GbE FEC/PCS/MAC)
      • Bifurcatable 200 GbE hard IP block (10/25/50/100/200 Gbs FEC/PCS)
    • 300G Interlaken
    • IEEE 1588 support
    • PMA direct
• Secure device manager (SDM)
  • AES-256/SHA-256 bitstream encryption/authentication
  • Physically unclonable function (PUF)
  • ECDSA 256/384 boot code authentication
  • Side channel attack protection
• Dimensions – 56 x 45 mm
• Manufacturing process – Intel Fab 7

The Agilex 9 Direct RF-Series is supported by the Quartus Prime Pro Edition design software, and the company also provides example designs, IP blocks, and a development kit for which we have limited information, except for the cool photo below.

We’re also told that the AGRW039 is fully compatible with designs targeting Altera’s Agilex 7 M-Series devices, so customers can preserve designs while moving to the Agilex 9 Direct RF-Series platform.

You’ll find a few more technical details on the product page, and a nice summary on the blog, but I’ll spare you the link to the utterly useless, quasi-misleading press release…

Thanks to TLS for the tip.

The post Altera Agilex 9 Direct RF-Series AGRW039 SoC FPGA targets high-performance RF systems appeared first on CNX Software - Embedded Systems News.

GEEKOM sent us a review sample of the A7 2026 Edition mini PC powered by a 4.9 GHz AMD Ryzen 5 7545U hexa-core/12-thread processor, paired with 16GB DDR5 SO-DIMM memory and a 500 GB NVMe SSD.

It comes with two HDMI 2.0 ports and two USB4 with DisplayPort Alt mode to enable four-monitor setups, 2.5GbE, WiFi 6E, and Bluetooth 5.2 connectivity, an SD card reader, a 3.5mm audio jack, and a few USB-A ports. As its name implies, it’s a variant of the GEEKOM A7 (Ryzen 9 7940HS) mini PC we reviewed in 2024 with Windows 11 Pro and Ubuntu 22.04/24.04, and we’ll keep that in mind to compare the more affordable 2026 Edition to its older sibling in our three-part review.

In this article, we will cover the full specifications, unbox the device, go through a teardown to check the hardware design, and boot it to the pre-installed Windows 11 Pro OS for a first look. We’ll then test the AMD Ryzen 5 7545U mini PC in depth with Windows 11 and Ubuntu 26.04 in the next parts of the review.

GEEKOM A7 2026 Edition specifications

• SoC (one or the other)
  • AMD Ryzen 5 7545U 6-core/12-thread Zen4/4c processor up to 4.9 GHz with 16MB L3 cache, AMD Radeon 740M Graphics @ 2800 MHz; TDP: 28W (cTDP: 15-30W)
  • AMD Ryzen 5 7535U 6-core/12-thread Zen3+ processor up to 4.55 GHz with 16MB L3 cache, AMD Radeon 660M Graphics @ 1900 MHz; TDP: 28W (cTDP: 15-30W)
• System Memory – 16 GB DDR5 upgradeable to 64 GB via 2x 262-pin SODIMM sockets
• Storage
  • 500 GB (7545U) or 1 TB (7535U) NVMe SSD upgradeable up to 2 TB NVMe PCIe Gen x4 4 SSD
  • Full-size SD card reader
• Video Output
  • 2x HDMI 2.0 ports up to 4Kp60
  • 2x USB-C ports with DisplayPort Alt. mode
  • 4x independent displays supported
• Audio – 3.5mm audio jack, digital audio via HDMI and USB-C ports
• Connectivity
  • 2.5GbE RJ45 port via a Realtek RTL8125BG-CG controller
  • WiFI 6E and Bluetooth 5.23
• USB
  • 3x USB 3.2 Gen 2 Type-A ports
  • 1x USB 4 Gen3 Type-C port
  • 1 x USB 3.2 Gen 2 Type-C port
  • 1 x USB 2.0 Type-A port
• Misc
  • Power button with LED
  • IceBlast 3.0 Silent Cooling
• Power Supply – 19V (120W) via DC jack
• Dimensions – 112.5 x 112.0 x 37 mm

The mini PC comes preloaded with Windows 11 Pro. The specifications are basically the same as the GEEKOM A7 (2024 Edition), except for a more efficient processor with lower TDP (performance to be determined) and lower memory and storage capacities (16GB/500GB vs 32GB/2TB), likely due to the current RAM and SSD market situation.

GEEKOM A7 2026 Edition unboxing

The mini PC shipped in a sturdy retail box marked “GEEKOM” and “GEEKOM A Series”. The sticker on the bottom side provides basic configuration details: A7 model AMD Ryzen 5 7545U CPU, 16GB DDR5 SO-DIMM memory, and 500GB M.2 SSD storage. Just as expected.

The GEEKOM A7 2026 Edition mini PC ships with a 19V/6.32A (120W) power adapter and power cord, an HDMI cable, a VESA mount with a screw pack, a Thank You card, and a user guide. The VESA mount was not included in the previous generation A7 model, so that part is new.

What has not changed at all are the ports. The front panel comes with two USB 3.2 ports (left one with USB PD output), a 3.5mm audio jack, and a power button.

The rear panel features some ventilation holes, a 19V DC jack, a 40 Gbps USB4 port with DisplayPort Alt mode, two HDMI 2.0 ports, a 2.5GbE jack, a 10 Gbps USB 3.2 Type-A port, a USB 2.0 port, and a 10 Gbps USB 3.2 Type-C port with DisplayPort Alt mode.

The new model still comes with a full-size SD card reader.

GEEKOM A7 2026 Edition teardown

Time for a teardown of the GEEKOM A7 2026 Edition mini PC. We’ll need to take out four rubber pads and loosen the screws under each to start the unboxing.

Be very careful when removing the bottom cover since there’s an antenna attached to it, and we’ve previously broken the antenna wire in mini PCs with a similar design. No problem this time though. We’ll need to loosen four more screws to remove the cooling metal plate.

At this stage, we have access to the motherboard featuring two SODIMM RAM slots, but fitted with only one, meaning single-channel memory which may impact performance of graphics and AI tasks. The right side features a Kingston M.2 SSD that is cooled with a thermal pad attached to the metal cover.

We removed the NVMe SSD and the memory module to have a closer look. The mini PC is fitted with a 500GB Kingston NV3 PCIe 4.0 NVMe SSD and a single 16GB TWSC TDS5DSAG08-48MA40C PC5-4800 DDR5 SO-DIMM memory module.

There are two notable chips on the motherboard: Realtek RTS5452H USB-C power delivery controller and Genesis Logic GL3590 USB 3.2 Gen 2 hub controller.

We’ll also find the Azureware AW-XB591NF wireless module there based on a MediaTek MT7922 Wi-Fi 6E and Bluetooth 5.2 chipset. The company used the same part as in the 2024 Edition GEEKOM A7 here.

First boot to Windows 11 Pro

We connected the GEEKOM A7 2026 Edition to an HDMI monitor, a USB RF dongle for a wireless mouse/keyboard combo, and the power adapter to boot the mini PC. The first time, it entered the initial Windows 11 Pro setup process with the usual language selection, network connection (Ethernet or Wi-Fi), and user account setup.

It went smoothly, and we reached the Windows 11 Pro desktop with an active internet connection over Wi-Fi.

We then went to System -> About and confirmed we had an A7 computer equipped with a 3.20 GHz (base frequency) AMD Ryzen 5 7545U processor with Radeon 740M graphics, 16GB of RAM, running Windows 11 Pro 64-bit version 25H2 installed on a 466 GB storage device.

That will be all for today. In the next part of the review, we will test the GEEKOM A7 2026 Edition mini PC on Windows 11 in detail, and follow up with an Ubuntu 26.04 review to evaluate the AMD Ryzen 5 7545U mini PC under Linux.

We’d like to thank GEEKOM for sending us the A7 2026 Edition mini PC for review. The model shown in this review, equipped with an AMD Ryzen 5 7545U processor paired with 16GB DDR5 RAM and a 500GB M.2 SSD, sells for $599 on Amazon US or the GEEKOM store.

CNXSoft: This article is a translation – with some additional insights – of the original review on CNX Software Thailand by Suthinee Kerdkaew.

The post GEEKOM A7 2026 Edition (AMD Ryzen 5 7545U) mini PC review – Part 1: specifications, unboxing, teardown, and first boot appeared first on CNX Software - Embedded Systems News.

GL.iNet Mudi 7, codenamed GL-E5800, is a battery-powered 5G NR WiFi-7 travel router with 2.5GbE, up to 4.67 Gbps downlink cellular speed, up to 5.76 Gbps WiFi link rate, and 700 Mbps VPN transfer rate using OpenVPN-DCO (600 Mbps with WireGuard VPN).

It’s the third model since the company introduced the Mudi portable 4G LTE WiFi router in 2019, and followed up with the Mudi V2 (GL-E750V2) 4G LTE router in 2024. The Mudi 7 is a serious upgrade with 2.5GbE, 5G cellular, and WiFi 7 connectivity, and a large 5380 mAh battery good for a full day. It also features two USB-C ports for power and tethering, a 2.8-inch touchscreen display, eight internal antennas, and two TS-9 connectors for external antennas.

Mudi 7 specifications:

• SoC – Unnamed Qualcomm quad-core processor @ 2.2GHz (potentially IPQ9574 quad-core Cortex-A73)
• System Memory – 2GB LPDDR4X
• Storage – 8GB eMMC flash
• Display – 2.8-inch color LCD with touchscreen
• Wired Networking – 2.5Gbps Ethernet port
• Wireless – 5G and WiFi 7
  • Chipset – Qualcomm Dragonwing MBB Gen 3 (based on Snapdragon X72)
  • WiFi 7 
    • Protocol – IEEE 802.11a/b/g/n/ac/ax/be
    • Wi-Fi Speed – 688 Mbps (2.4GHz), 2,882 Mbps (5GHz), 5,764 Mbps (6GHz)
  • 5G / 4G LTE cellular
    • 3GPP Rel-17 NSA/SA operation, Sub-6GHz
    • LTE Category: Cat 20 (DL) / Cat 18 (UL)
    • Max speed – Downlink: 4.67 Gbps; uplink: 3.5 Gbps (from X72/X75 specs)
    • Frequency bands
      • GL-E5800NA (North America variant)
        • 5G NR NSA: n2 / n5 / n7 / n12 / n14 / n25 / n26 / n30 / n38 / n41 / n48 / n66 / n71 / n77 / n78
        • 5G NR SA: n2 / n5 / n7 / n12 / n13 / n14 / n25 / n26 / n29 / n30 / n38 / n41 / n48 / n66 / n70 / n71 / n77 / n78
        • LTE-FDD: B2 / B4 / B5 / B7 / B12 / B13 / B14 / B17 / B25 / B26 / B29 / B30 / B66 / B71
        • LTE-TDD: B38 / B41 / B42 / B43 / B48
      • GL-E5800EU (Europe/Global variant)
        • 5G NR NSA: n1 / n3 / n5 / n7 / n8 / n20 / n26 / n28 / n38 / n40 / n41 / n75 / n77 / n78
        • 5G NR SA: n1 / n3 / n5 / n7 / n8 / n20 / n26 / n28 / n38 / n40 / n41 / n75 / n77 / n78
        • LTE-FDD: B1 / B3 / B5 / B7 / B8 / B20 / B28 / B32
        • LTE-TDD: B38 / B40 / B41 / B42 / B43
        • WCDMA: B1 / B5 / B8
    • 2x Nano SIM card slots
    • Onboard eSIM
  • Antennas 
    • 8x internal antennas (6x cellular antennas, 2x 2.4GHz & 5GHz & 6GHz Wi-Fi antennas)
    • 2x TS-9 ports for external antennas (not provided by default)
• USB 
  • USB 3.2 (10 Gbps) Type-C OTG port for power input & output, USB tethering
  • USB Type-C port for power input and output
• Misc
  • Power button
  • Internal Reset button under battery cover
• Power Supply
  • 3.85V/5380mAh (20.72Wh) built-in removable battery good for a typical 13.5 hours on a full charge
  • Charging – USB PD/PPS 5-12V, 30W Max
  • Power Consumption – < 30W
• Dimensions – 157 x 75 x 22.8 mm
• Weight – 300 grams
• Temperature Range – 0 ~ 40°C

The Mudi 7 router ships with a battery pack, a USB-C cable for charging, a travel pouch, and a user manual. The system runs a fork of OpenWrt 22.03 with Linux 5.15. You’ll find documentation to get started on the GL.iNet website. It’s still based on Admin Panel v4.0, so the user experience when configuring the router with the web interface will be similar to the GL.iNet Spitz AX 5G router we reviewed about three years ago. The advantages of the new Mudi 7 are that it’s portable, features a touchscreen display for quick configuration/settings, and should deliver much better performance thanks to a faster SoC and an enhanced 5G/WiFi 7 sub-system.

It was unveiled at CES 2026, started shipping from the GL.iNet store in April 2026, and is now more broadly available, so we already have some reviews from sites or YouTubers such as NAScompares and Jabber Tech. Reviews are generally positive, although both complain about the lack of MLO (Multi-Link Operation) support in the Mudi 7.

One thing that has not changed since I reviewed the Spitx AX 5G router is that 5G hardware is not exactly cheap. The Mudi sells for $420 on Amazon, AliExpress, or the GL.iNet store. By comparison, the Mudi V2 4G LTE + WiFi 5 router (433 Mbps) now goes for $155, so if all you require is basic connectivity on the router, the Mudi 7 might be overspec’ed and overpriced.

The post Mudi 7 (GL-E5800) 5G NR Wi-Fi 7 travel router supports up to 4.67 Gbps link rate appeared first on CNX Software - Embedded Systems News.

The Armbian community has just released the Armbian Imager 2.0 GUI program to easily flash pre-built Armbian-built Ubuntu or Debian images for over 338 boards from 64 SBC vendors. The new version features a slick user interface rewritten from scratch and implements custom user profiles in the settings with username and password, SSH key, Wi-Fi network credentials and country code, timezone, locale, and shell. That means the board is ready to use after flashing.  In some ways it’s similar to the Raspberry Pi Imager 2.0, except it covers a much broader ecosystem of single board computers.

To be honest, I had no idea Armbian had an imager so far. The last time I used an Armbian image, I downloaded it directly from their website and used USB Imager or another tool to flash it to a microSD card slot. So it’s a good opportunity to check out the Armbian Imager 2.0 program by downloading and installing it on my laptop.

It’s available for Linux x64/arm64 (Raspberry Pi), Windows x64/arm64, macOS x64 and Arm (Apple M1-M4). I installed it on a laptop running Ubuntu 24.04. Some UI features include dark theme selection and eighteen languages automatically selected from your locale.

The first step is to select the board’s manufacturer, and the platinum vendors, who sponsor the Armbian project, come first. I went with Radxa.

We’re now shown a list of boards from the manufacturer, again listed by level of support with four levels: Platinum, Standard, Community, WIP. I selected the Rock 5B Plus SBC.

In the next step, we can select the operating system, again with a list of images that can be filtered by “Stable”, “Rolling Release”, “Apps”, and “Minimal”. The “Apps” filter will show specific images such as Kali Linux, Home Assistant, OpenMediaVault, etc. I went with the Armbian 26.2.6 GNOME image above.

Up next is storage device selection with the system device hidden by device. My 16GB microSD card was properly detected, and after selecting it, I was presented with a summary, and click on Erase & Flash to start the process.

The utility will download and uncompress the OS image, then write it to the microSD card and verify it.

Sadly, it failed due to a problem with that specific microSD card:

09:31:40 ● flash::linux::writer: Image size: 6140461056 bytes (5.72 GB)
09:31:40 ● flash::linux::writer: Unmounting device partitions...
09:31:46 ● flash::linux::writer: Writing image...
09:35:54 ● flash::linux::writer: Write error at byte 301989888: No space left on device (os error 28)
09:35:54 ● operations: Flash failed: Failed to write at byte 301989888: No space left on device (os error 28)
09:35:54 ● board_queries: Device(s) removed: ["/dev/sda"]
09:35:54 ● operations: Cleaning up failed download

The uncompressed image size is 5.72 GB, and we’re told there’s no space left on a ~16GB microSD card, so there’s clearly an issue, which the kernel log confirms:

[387400.206648] sd 0:0:0:0: [sda] tag#0 FAILED Result: hostbyte=DID_OK driverbyte=DRIVER_OK cmd_age=0s
[387400.206657] sd 0:0:0:0: [sda] tag#0 Sense Key : Not Ready [current] 
[387400.206661] sd 0:0:0:0: [sda] tag#0 Add. Sense: Medium not present
[387400.206663] sd 0:0:0:0: [sda] tag#0 CDB: Write(10) 2a 00 00 08 1e 00 00 00 f0 00
[387400.206665] I/O error, dev sda, sector 531968 op 0x1:(WRITE) flags 0x4800 phys_seg 30 prio class 2
[387400.206675] Buffer I/O error on dev sda, logical block 66496, lost async page write
[387400.206682] Buffer I/O error on dev sda, logical block 66497, lost async page write
[387400.206685] Buffer I/O error on dev sda, logical block 66498, lost async page write
[387400.206687] Buffer I/O error on dev sda, logical block 66499, lost async page write
[387400.206689] Buffer I/O error on dev sda, logical block 66500, lost async page write

While the program detected the issue in the console, the UI didn’t show any error message, and instead, I got a mostly blank window.

I tried again with an 8GB microSD card, and I got the same result as above in the user interface, but it failed silently in the terminal.

I have enough experience to know that when it’s not the storage device itself that fails, it might be the card reader instead. Since I don’t have a working card reader with me, I installed the Windows version of Armbian Imager 2.0 on the Khadas Mind 2 mini PC and tried again using the card reader in the Mind Graphics 2 dock. I could confirm the 16GB microSD card was indeed dead, but I managed to flash Armbian 26.2 successfully on the 8GB card.

What we haven’t checked out so far is the custom profile option, so let’s do that now by going to the Settings, Profiles section.

Clicking on “+ New profile” will show Network options for Ethernet and Wi-Fi…

… as well as localization, root account, first user options…

… as well as Advanced options, and a preview of the config file stored in /root/.not_logged_in_yet. Clicking Reveal will show the passwords. If you set passwords, you may not want to share the resulting image with third parties since passwords are stored in plain text in the file. I suppose it’s deleted after the first login, but it’s usually possible to mount such an OS image as a loop device to access all files.

We can now click save and create as many profiles as we want. The profile can be selected on the “Confirm Selection” window right before starting the flashing process.

The Armbian Imager program is open-source, and you’ll find the source code on GitHub.

The post Armbian Imager 2.0 release supports over 300 boards from 64 SBC vendors, custom user profiles appeared first on CNX Software - Embedded Systems News.

Qualcomm has introduced the Dragonwing Mobile Broadband Multimedia (MBM) family, a new lineup of 4-nm SoCs that integrate 5G connectivity, Wi-Fi, multimedia processing, and on-device AI into a single platform for “interactive broadband” devices.

The family currently features the top-tier Dragonwing MBM715 and the lower-cost Dragonwing MBM415, designed for smart displays, retail kiosks, service terminals, and other connected devices that require both high-speed connectivity and local processing. Unlike mobile broadband (MBB) or fixed wireless access (FWA) platforms, the MBM series combines connectivity, display, camera, and AI into one chip, so no separate processor is needed.

Qualcomm Dragonwing MBM715 & MBM415 Specifications:

SoCs	Dragonwing MBM415	Dragonwing MBM715
CPU	64-bit Qualcomm Kryo with 6x efficiency cores, up to 2 GHz (Qualcomm Kryo)	Qualcomm Kryo 64-bit, up to 2.8 GHz
GPU	Qualcomm Adreno; OpenGL ES 3.2, OpenCL 2.0, Vulkan 1.1	Qualcomm Adreno; OpenCL 2.0 FP, OpenGL ES 3.2, Vulkan 1.3
Video Capture & Playback	HEVC/H.264 support, 720p@120fps slow-mo, lower than 4K resolution	Up to 4K@60fps, 1080p@120fps slow-mo, supports HLG/HDR10+/AV1/VP9/HEVC, plus advanced processing features
ISP	12-bit dual ISP	12-bit triple spectra ISP
AI	No dedicated on-device Generative AI engine	Qualcomm AI Engine with Hexagon processor (scalar, tensor, and vector accelerators), dual-core Sensing Hub, and Qualcomm AI Hub support
Memory	LPDDR5 and LPDDR4x speed not specified for MBM415	LPDDR5x up to 4200 MHz
Storage	Not Specified but should support flash/eMMC memory
Display	Up to FHD+ @ 120 Hz (900×1600), no listed HDR or external display	Up to WQHD+ @ 144 Hz (on-device), 4K @ 60 Hz external, 10-bit color with HDR support
Camera	Up to 64 MP (single), dual camera support, basic features with limited processing	Up to 200 MP (single), triple camera support, with HDR, AI auto-focus/exposure, EIS, and low-light
Audio	Qualcomm Aqstic codec + smart speaker amp MBM715 Adds aptX Adaptive, Lossless, Voice; MBM415 supports Standard aptX Audio
USB	Not listed	USB-C interface via USB 3.1 Gen 2 (Only on MBM715)
Cellular	Sub-6 GHz 5G (up to 2.5 Gbps / 900 Mbps up), multi-SIM, standard power and signal optimizations	Sub-6 GHz 5G (4.2 Gbps), wide legacy support, advanced features like PowerSave 2.0 and Smart Transmit 2.0
Wi-Fi	Wi-Fi 6E / 5, up to 6 GHz bands, lower max performance	Wi-Fi 7 (up to 5.8 Gbps), supports 4/5/6 GHz, advanced features like 4K QAM, MU-MIMO, OFDMA
Bluetooth & NFC	Bluetooth 5.3	Bluetooth 6.0 with BLE Audio; NFC supported
Location	GPS, GLONASS, Beidou, Galileo, NavIC, QZSS	GPS, GNSS, GLONASS, Beidou, Galileo, NavIC, QZSSFreq: TripleAccuracy: Lane-level, Sidewalk-levelFeatures: Sensor-assisted, concurrent satellite systems, Location Suite
Security	Not listed	Platform Security Foundations, Trusted Execution Environment (TEE), Type-1 Hypervisor, Wireless Edge Services (WES), Trust Management Engine, 3D Sonic Sensor Max (Fingerprint)
Charging	Not listed	Qualcomm Quick Charge 5 technology

The MBM415 has no dedicated on-device Generative AI engine, but Qualcomm officially brands the MBM415’s modem-side processing as “Qualcomm AI-Enhanced Signal Boost.” It is not for generative AI, but it is an AI feature listed in the specs.

Also note that the MBM715 supports CBRS (3.5 GHz), which is used for private 5G networks in the US only. That means, if a company is building a private 5G network in the US, the MBM715 complies with the FCC’s CBRS regulations.

So in conclusion the MBM715 is designed for high-end edge devices like smart kiosks, AI assistants, and smart-home hubs, with a built-in Hexagon NPU capable of running AI models locally, along with faster connectivity (up to 5.8 Gbps Wi-Fi 7 and 4.2 Gbps 5G), Bluetooth 6.0, and support for up to 200 MP cameras and 4K video; in contrast, the MBM415 is a lower-cost, lower-power option without dedicated AI accelerator, using six efficiency cores and focusing on basic connectivity with Wi-Fi 6E, Bluetooth 5.3, up to 2.5 Gbps 5G, and FHD+ @ 120 Hz display output, making it better suited for simpler applications like digital signage, ticket kiosks, and industrial panels.

The MBM family (MBM715 and MBM415) supports both Android and Linux, with shared drivers across platforms to simplify development.

Qualcomm says the MBM715 and MBM415 are already available for evaluation, with reference designs for OEMs building devices like kiosks and smart displays. Access typically requires contacting Qualcomm or using their developer portal, as these platforms are aimed at OEMs rather than general developer kits. More information is available on the product page and press release.

The post Qualcomm Dragonwing Mobile Broadband Multimedia (MBM) MBM715 and MBM 415 SoCs combine 5G, Wi-Fi, multimedia, and Edge AI appeared first on CNX Software - Embedded Systems News.

The GEEKOM A6 mini PC, pairing an AMD Ryzen 7 6800H octa-core CPU with up to 16GB DDR5 and a 1TB SSD, is now offered at a 20% discount, dropping the price to $479 from $599 MSRP.

This promotional price is expected to be lower than the A6’s pricing during the June 2026 Prime Day event, making it the best time to buy. To enjoy the discount on the Windows 11 Pro mini PC, simply enter the coupon code CNXGKA620 on the GEEKOM store or Amazon before June 21.

Here’s a reminder of the GEEKOM A6 specifications:

• SoC – AMD Ryzen 7 6840HS 8-core/16-thread processor up to 3.2 GHz / 4.7GHz (Turbo) with 16MB L3 cache, AMD Radeon 680M Graphics; TDP: 45W
• System Memory – 16GB dual-channel DDR5-4800 via SODIMM sockets
• Storage
  • 1TB NVMe 2280 SSD via M.2 Key-M socket (PCIe x4 Gen 4 or SATA)
  • M.2 Key-B socket (SATA only)
  • Full-size SD card reader
• Video Output
  • 2x HDMI 2.0 ports
  • 2x USB-C ports with DisplayPort Alt. mode
  • Support for up to 4x independent displays with 4K resolution
• Audio
  • HDA CODEC
  • 3.5mm audio jack
  • Digital audio via HDMI and USB-C ports
• Networking
  • 2.5GbE RJ45 port via a Realtek RTL8125BG-CG controller
  • WiFI 6E and Bluetooth 5.2
• USB
  • USB4 Type-C port with DisplayPort Alt mode
  • 3x USB 3.2 Gen 2 Type-A ports, including one with Power Delivery (USB PD)
  • USB 3.2 Gen 2 Type-C port with DisplayPort Alt mode and Power Delivery (USB PD)
  • USB 2.0 Type-A port
• Security – dTPM 2.0
• Misc
  • Power button
  • 2x Power LEDs
  • RTC battery
  • IceBlast 2.0 cooling system
• Power Supply – 19V (120W) via DC jack
• Dimensions – 112.4 x 112.4 x 37 mm
• Certifications – CE, FCC, CB, CCC, SRRC, RoHS

The mini PC ships with a 19V/6.32A (120W) power adapter with a power cord, an HDMI cable, a VESA mount with a screw set, and a user guide. While the GEEKOM A6 comes pre-loaded with a licensed version of Windows 11 Pro, the company highlights support for Ubuntu and other Linux distributions.

While on the topic of OS support, we reviewed the 32GB/1TB variant of this AMD Ryzen 7 6840HS mini PC with both Windows 11 Pro and Ubuntu 24.04 in February 2025.  Everything worked well, including quad display support, 2.5GbE and WiFi 6 networking, and YouTube video up to 4Kp60 or 8Kp30 in both operating systems. We had an issue with the MT7922’s Bluetooth function on Ubuntu 24.04, but it’s now resolved. One small downside is the usual fan noise under load with this type of actively-cooled mini PC, but the cooling system used helps keep that in check to some extent.

At the time, we noted the good value for a sub-$500 mini PC, and, remarkably, they’ve managed to keep the price at a similar level ($479) in 2026, even though the memory had to be reduced from 32GB to 16GB due to the current memory market conditions. GEEKOM offers shipping from a local warehouse for easy returns, a 3-year warranty, and a 30-day refund & return commitment.

The post GEEKOM A6 AMD Ryzen 7 6800H Mini PC is now off by 20% for a limited time (Sponsored) appeared first on CNX Software - Embedded Systems News.

AweSun has sent me a sample of their “Cloud KVM Q1” 4K KVM over IP solution for review. It’s a compact device with the minimum number of ports for a connected KVM: USB-C for keyboard and mouse emulation, HDMI input for video, and Ethernet for connectivity to the host.

Like other such KVMs, it enables hardware-level remote access even in the BIOS, and the company advertised remote access to server, computer, and mobile phone targets, with the latter requiring an additional USB-C dock for HDMI input. I’ll start the review by going through the specifications, performing an unboxing and a teardown, and testing both mobile and desktop clients with targets like a Raspberry Pi 5 and an Android mobile phone.

AweSun Cloud KVM Q1 specifications

Here are the specifications from the company:

• Video Input
  • HDMI port
  • Resolution – Up to 2560 x 1600 with 15 FPS framerate
• Networking
  • 100Mbps RJ45 WAN Port
  • Bluetooth 5.0 (no WiFi)
• USB – 1x USB-C port for mouse and keyboard emulation
• Misc
  • System LED
  • Reset pinhole
• Power Supply – 12V/1A via DC jack
• Dimensions – 75 x 45.05 x 17mm
• Weight – 63 grams
• Material – aluminum alloy + engineering plastic
• Temperature Range – Operating: 0~40°C; storage: -40~85°C
• Humidity – Operating and storage: 10%~90% RH, non-condensing

No information about the SoC and memory, but we’ll find more details about these in the teardown below.

AweSun IPKVM Q1 unboxing

I received the device in a simple retail package reading “AweSun IPKVM Q1 hardware-level remote access” along with a CN to US/AU plug adapter. I was told the following for the latter:  “our power plugs are the Chinese version, so we’ve included the adapter to make it compatible with US outlets for now”. However, I didn’t have to use it at all here in Thailand.

The package comes with the device itself, quite smaller than the GL.iNet Comet (GL-RM1) KVM-over-IP solution I reviewed last year (partially because it lacks a USB Type-A port), a 12V/1A power adapter, a QC pass sticker, and a card with QR codes to the documentation, as well as WhatsApp and Facebook for support.

There aren’t any cables, so you’ll need your own HDMI and USB Type-C cables.

The front panel only features a power/status LED and the Q1 marking.

All ports are on the rear panel. From left to right: 12V DC jack, USB-C port, Reset pinhole, Ethernet RJ45 jack, and HDMI input port.

Teardown

It took me a couple of minutes to find out how to open it. I initially thought the inner part was clipped to the case, but I eventually found four screws on each corner.

From there, it’s relatively easy to take out the board from the enclosure. The top of the board features a Winbond 25NQ1GVZE1G 1 Gbit (256MB) SPI NAND flash for the OS,  an U&T Unlimited UTH16A10 Ethernet transformer, and a WB800DCS.2 wireless module based on the AIC8800DC 2.4GHz WiFi 6 and Bluetooth 5.2 chipset. People who want to play around with the serial terminal will also find an unpopulated 3-pin UART connector.

The chip under the small thermal pad is a Rockchip RK628H MIPI CSI/DSI to HDMI bridge, a newer variant of the RK628D, to handle up to 4Kp60 video input..

The bottom side has various passive components, and the main chip is under a thermal pad for cooling.

Unsurprisingly, it’s a Rockchip RV1106G3 Arm Cortex A7 camera SoC with 256MB on-chip DDR3L memory.

Initial setup

Let’s reassemble everything together to get started and set up the device. I’ll mostly follow the instructions from the user manual accessible using the QR code on the card in the package.

For the first setup, we’re told to connect the Q1 to an AweSun account to enable remote control, and do so via the AweSun app. I personally prefer desktop control, but fair enough, I’ll go the mobile route first and head over to the download page. The AweSun app is available for Windows (32/64-bit x86 and Arm), iOS, and Android. For the latter, two apps are available: “Control side” and “Host side”. We want the former for now, and we’ll use the host app later to control the phone from a Windows laptop, since there’s no Linux client for my Ubuntu installation. I installed the “AweSun Remote Control” app on an OPPO A98 5G Android smartphone via Google Play.

After a welcome screen, we’re asked to agree to the privacy policy and user license, and enable permissions for location, calls, camera, and Bluetooth. As I understand it, only Bluetooth is actually required for the AweSun Q1 KVM.

After that, we’ll need to register an account with an email or Google account, and upon sign-up, we’re asked whether we want to pay 675 THB per year (about $21) to enable extra features such as high-speed file transfer, 2K HD quality, multi-screen control, and True Color 4:4:4. As I understand it, it’s because the app also works as a remote access solution without hardware.

I just closed this window and reached the “Devices” section on the main page, where I was asked to agree to a user license agreement and privacy policy again… Time to add our device. Tap on the + icon on the top right, and select “Add smart device”. You’ll want to make sure the AweSun Cloud KVM Q1 is connected to power (and optionally to Ethernet) at this stage, with the Blue LED blinking, waiting for pairing.

The device will show as Q1-XXXX. Let’s select it, another scan occurs, and we can click on Add. Bluetooth connectivity should now be established with the LED on the unit turning to solid Blue. If you haven’t done it already, it’s time to connect an Ethernet cable.

Tap Connect via network cable, and after waiting for a couple of seconds, you’ll be asked to enter a device name and a password for remote access. The password requirements are pretty low, with a minimum of six characters and 2 types of letters, so something like Passw0rd would work (weak password). Make sure to select something better, since this can be used to remotely access your computer and, by extension, your whole network.

Once successful, you’ll see the Binding success window, and your new device will be listed in the Devices window with its name and public IP address. You’ll want to keep that secret, since it might also give access to your ISP’s router dashboard. Where I live (Thailand), ISPs don’t take security seriously with default passwords (like admin), open telnet port, and other vulnerabilities… I initially had some issues at the Binding stage, and I first thought it may have been because of the VPN enabled on my smartphone, but I had the same problem after resetting the Q1, and several tries were necessary for binding to be successful. Once the initial setup is complete, the LED on the AweSun Q1 should become solid green.

Testing the AweSun KVM Cloud Q1 with a Raspberry Pi 5 and AweSun Android app

Now that the initial configuration is finished, let’s connect the AweSun KVM Cloud Q1 to a target: a Raspberry Pi 5 housed in a Pironman 5 Pro Max enclosure. I connected my own HDMI cable and USB-C to USB-A cable between the two. I used HDMI1, but it might be better to go with HDMI0.

The Pironman 5 Pro Max enclosure also comes with a MIPI DSI display (800×480 resolution), and it’s better to disable it in Raspberry Pi OS.

That’s because, in a dual-screen configuration, the KVM will only show the secondary HDMI output without access to the top menu (by default). It’s possible to set up mirroring, but since the displays have different aspect ratios, that’s an issue, especially with Wayland…

Two connection methods are available: click on the device itself in the Devices section and enter the “access password”, or go to the Remote Access section and use the Device ID and a user-defined passcode. By default, that password is temporary for security, but you can change that to long-term passcode, Today’s passcode, or One-time passcode.

I initially had some problems with the “access password” with the app claiming it was wrong, so I did most of the testing with the “Remote Assist” method instead. The Devices method eventually worked for me, so I’m not sure what went wrong initially.

While it’s possible to use the default cursor during emulation, I found out that enabling “touch mode” and “virtual mouse” is much more convenient on a small smartphone display. It brings a draggable mouse with large left and right buttons that makes everything easier to use. It can be enabled through the <> button in the interface.

While you can use the built-in keyboard on your smartphone, there’s also a “custom key combinations” function to enable shortcuts.

I’ve shot a video controlling Raspberry Pi OS, mostly to show the lag and performance of the solution.

I also demonstrated the ability to access the Raspberry Pi 5 remotely even when Raspberry Pi OS is not running. I turned off the Pi 5, at which point the system detects there’s no video signal and shows a color test pattern.

I removed the microSD card to simulate an issue, and I was able to see the output on my phone through the AweSun app.

What I couldn’t do was press the Spacebar to enter the bootloader. It looks to be a common issue for KVM used with the Raspberry Pi and not unique to the AweSun Q1.

Switching to AweSun for Desktop

The AweSun app is not only available for Android and iOS mobile devices, but also for Windows and macOS computers. As an Ubuntu user, I was a little disappointed by the lack of Linux support, and I had to reboot my ASUS Vivobook 16 laptop to Windows 11 to download and install AweSun for Windows.

What I could immediately do in the program is control the Windows laptop from the Android app using the Remote Control device ID and passcode provided in the Windows program. That’s not using the KVM at all, though.

To access the AweSun Q1 KVM, either through the “Device List” or the “Control Remote Device” option, I had to log in using the same credentials used in the Android app.  Just one little problem: the “Continue with browser” button didn’t do anything on my laptop, so I used the “copy link” option from the program to go through the login process. It completed, but wouldn’t go back to the AweSun program, which read “connection failed, try to switch the login page”.

I tried with Firefox, Microsoft Edge, and Chrome browsers, but all had the same result. I communicated with the company for a while about the issues, and we couldn’t find a working solution. In the meantime, I received the Khadas Mind 2 with Windows 11, so I installed the AweSun program on it. I didn’t get “connection failed” anymore and a “Log in” pop-up I had never seen on the laptop showed up.

I could go through the login process normally and found my Q1-CNXSOFT KVM right there.

As a side note, the Mind (2) mini PC also showed up in the Android app. So that means each device running the AweSun app/program becomes part of a remote access “mesh” where each device can access each other, which is pretty neat.

The phone is not part of the Devices list because a separate app (AweSun Host) is required to control the phone from the AweSun program, and I haven’t tried it because it’s out of the scope of this review. What does work directly with the AweSun mobile app is screen mirroring.

Since it’s possible to do remote access without hardware, the KVM is mostly useful for unsupported operating systems like Linux or BSD distributions, and when you need remote access even if the system crashes or to access the BIOS, for example to install or reinstall Windows/Linux remotely. Note that the storage on the Q1 is too small for most ISOs, so you’d need a USB drive attached to the target to install an operating system.

I can now access the Raspberry Pi 5 in Windows 11. There’s some lag similar to what we got in Android, but that’s normal since it goes over the Internet, and it’s still usable…

The toolbar has options for Desktop, notably for resolution (original or scaled), image- or game-optimized desktop mode (I didn’t notice that much difference between the two while browsing), and more. The Keymouse has two options: local mouse and immersive mode. When I enabled local mouse, it was unusable and jittery, but I could disable it with Ctrl+Alt+Enter. Immersive mode means the mouse pointer stays in the target window. To use the mouse pointer on the host again, press Alt+Shift+X.

I tried to play a YouTube video with “Play sound of remote device” enabled. It didn’t work with the Raspberry Pi 5 due to an issue with HDMI audio on Raspberry Pi OS unrelated to the Q1 KVM. However, I could confirm audio worked fine after connecting the Q1 KVM to my Ubuntu 24.04 laptop. The video itself was watchable.

Finally, I did a quick test turning off the Pi 5 and starting it again without a microSD card to show I could still access the bootloader output from the AweSun Windows program.

Conclusion

The AweSun Cloud KVM Q1 can be useful for people who want an ultra-compact KVM device with basic KVM features such as HDMI video and audio, and keyboard and mouse emulation. Besides the hardware, I found the AweSun program/app ecosystem to be interesting, as it creates a sort of “KVM mesh” for all your devices, and any device running AweSun can be a remote access host or target, and the KVM adds support for unsupported OS and provides access to the BIOS and to a system that crashed/hung.

On the downside, I found the software can have rough edges, especially the on-board/logging part for which I had issues in both Android and Windows. But once you’re in, it’s working pretty smoothly. Some limitations to be aware of include the lack of expansion ports such as a USB connector or a microSD card slot, so you can’t do hardware power control with an ATX power board or FingerBot (Wake-on-LAN is supported, however), or copy an ISO file directly to the device like some other KVM solutions. The AweSun Q1 also only works through the cloud, and there’s no LAN access option.

I’d like to thank AweRay for sending the AweSun Cloud KVM Q1 for review. It can be purchased for $69 on the company’s store or on Amazon.

The post AweSun Cloud KVM Q1 Review – An ultra-compact, low-cost KVM over IP solution tested with Windows 11 and Android clients appeared first on CNX Software - Embedded Systems News.

AkiraOS is a Zephyr-based embedded OS that runs sandboxed WebAssembly applications on microcontrollers and lets users deploy and update firmware OTA without reflashing. In other words, it’s similar to Docker containers, but for microcontrollers.

The open-source embedded platform separates the OS from the application. That means the firmware stays stable, while apps are independent .wasm binaries deployable over-the-air without touching the OS, and portable so a single binary works on ESP32-S3, nRF5x, or STM32 MCU boards.

AkiraOS highlights:

• User space
  • Up to 8 wasm apps can be installed
  • Up to two apps can run at the same time
  • Footprint: 50KB to 200KB per app
• Akiraz runtime – Custom WASM runtime
  • App Manager
  • UI Framework with 32 widgets
  • Shell/console
  • 18 API modules
  • WebAssembly Micro Runtime (WAMR) – Two options: Interpreter or Ahead-Of-Time (AOT) compilation with 10 to 50x higher performance
• RTOS – Zephyr RTOS
  • Scheduler
  • Network stack
    • HTTP for OTA updates
    • Bluetooth LE for AkiraMesh
  • Drivers
  • LittleFS file system
• Benefits
  • Update apps in the field without a firmware flash cycle
  • No recompilation for the apps – One binary runs on ESP32-S3, nRF5x, or STM32 — no recompile
  • Device stays up even if a bad app crashes
  • Every app gets only the hardware access it explicitly requested

AkiraOS is supported on the following hardware targets:

• Tier 1 support (best)
  • Espressif Systems ESP32 series
    • ESP32-S3 (LX7) and ESP32-H2/ESP32-C6 (RISC-V)
    • ESP32-S3-DevKitM-1 recommended, and upcoming open-source hardware AkiraConsole V3 on Crowd Supply (we’ll check it once launched with full info including price)
  • Native_sim for fast iteration on x86-64 machines with no MCU hardware needed
• Tier 2 support
  • Nordic nRF54L15 Arm Cortex-M33 MCU with BLE 5.4
  • STM32 Arm Cortex-M microcontrollers
    • B-U585I-IOT02A Discovery kit (STM32U585AI)
    • STEVAL-STWINBX1 SensorTile Wireless Industrial Node Development Kit (STM32U585AI)
    • STMH753/H723 platforms

You’ll find the source and instructions to get started on GitHub, and a separate repo contains an SDK to develop apps for AkiraOS. More details can be found in the documentation on the project’s website, where I notice a mobile app is being developed to retrieve device info & status, manage apps, trigger an OTA firmware update, access the shell/terminal, and browse files on the target device. There’s also a management web interface accessible through WiFi or USB.

The post Docker for Microcontrollers? AkiraOS combines Zephyr RTOS with WebAssembly (WASM) applications appeared first on CNX Software - Embedded Systems News.