Neural networks and other machine learning processes are often associated with powerful processors and GPUs. However, as we’ve seen on the page, AI is also moving to the very edge, and the BitNetMCU open-source project further showcases that it is possible to run low-bit quantized neural networks on low-end RISC-V microcontrollers such as the inexpensive CH32V003. As a reminder, the CH32V003 is based on the QingKe 32-bit RISC-V2A processor, which supports two levels of interrupt nesting. It is a compact, low-power, general-purpose 48MHz microcontroller that has 2KB SRAM with 16KB flash. The chip comes in a TSSOP20, QFN20, SOP16, or SOP8 package. To run machine learning on the CH32V003 microcontroller, the BitNetMCU project does Quantization Aware Training (QAT) and fine-tunes the inference code and model structure, which makes it possible to surpass 99% test accuracy on a 16×16 MNIST dataset without using any multiplication instructions. This performance is impressive, considering […]
Android no longer supports RISC-V, for now…
Google dropped RISC-V support from the Android’s Generic Kernel Image in recently merged patches. Filed under the name “Remove ACK’s support for riscv64,” the patches with the description “support for risc64 GKI kernels is discontinued” on the AOSP tracker removed RISC-V kernel support, RISC-V kernel build support, and RISC-V emulator support. In simple terms, the next Android OS implementation that will use the latest GKI release won’t work on devices powered by RISC-V chips. Therefore, companies wanting to compile a RISC-V Android build will have to create and maintain their own branch from the Linux kernel (ACK RISC-V patches). These abbreviations can be confusing, so let’s focus on them starting with ACK. There’s the official Linux kernel, and Google does not certify Android devices that ship with this mainline Linux kernel. Google only maintains and certifies the ACK (Android Common Kernel), which are downstream branches from the official Linux kernel. One of the main ACK branches is the android-mainline […]
Yocto Project 5.0 “Scarthgap” released with Linux 6.6 and plenty of changes
The Yocto Project 5.0 codenamed “Scarthgap” has just been released with Linux 6.6, glibc 2.39, LLVM 18.1, and over 300 other recipe upgrades. As a result of the release, the developers have made it available for download (bz2 tarball). The Yocto Project, or Yocto for shorts, is a popular framework used to create custom embedded Linux distributions, and we’ve played with it over the year showing how to create a minimal image for the Raspberry Pi, and last year, we used it again when reviewing two industrial development boards, namely the VOIPAC IMX8M and ADLINK i-Pi SMARC 1200. Yocto is quite a powerful framework/build system with plenty of options that make it highly customizable, but the learning curve is fairly steep. Some other changes in Yocto Project 5.0 include: New variables: CVE_DB_INCR_UPDATE_AGE_THRES: Configure the maximum age of the internal CVE database for incremental update (instead of a full redownload). RPMBUILD_EXTRA_PARAMS: […]
Muse Book laptop features SpacemiT K1 octa-core RISC-V AI processor, up to 16GB RAM
SpacemiT, a chip design company from China with RISC-V as its core technology, recently unveiled the Muse Book laptop based on the K1 octa-core RISC-V chip. Unlike our daily laptops, it has many interesting unique features and is mainly sold to hardware engineers and DIY enthusiasts. This Muse Book runs the Bianbu OS operating system based on the Debian distribution and optimized to run on the SpacemiT K1 octa-core RISC-V SoC. Let’s first take a look at its external interfaces. On the left side of the laptop, there are two USB Type-C interfaces, a USB 3.0 Type-A port, a 3.5mm headphone jack, a microSD card slot, and a reset pinhole. The 8-pin header on the right side of the laptop is quite interesting, and SpacemiT hopes the Muse Book can become one of the most convenient hardware development platforms for RISC-V. In addition to the power pins, users will find […]
CAPUF Embedded CH32V003 RISC-V Dev Kit features USB-C, temperature/humidity monitoring, OLED & more
CAPUF Embedded CH32V003 Dev Kit is an all-in-one development board with a USB-C interface, onboard sensors (temperature/humidity), an OLED display, SPI NOR Flash, and ample I/O options. Additionally, it features a Qwicc connector, an RGB LED, and a 3-pin header to connect the WCH-link programmer providing further flexibility for your projects. After the initial release of this “10 cents” CH32V003 RISC-V microcontroller, we have seen $1.5 development boards and an open-source GGC toolchain, as well as Arduino support for the WCH RISC-V microcontrollers. Very recently, CNLohr even managed to transmit LoRa packets with this MCU. CAPUF Embedded CH32V003 Dev Kit Specifications: MCU – WCH CH32V003F4U6 32-bit RISC-V2A microcontroller up to 48 MHz with 2KB SRAM, 16KB flash (QFN20 package) USB – 1x USB-C Port for Power (5V) and Serial Interface with onboard USB to UART converter I/Os – 20 Pin MCU IOs with USART, I2C, SPI, ADC Program & Debugging – […]
ESP32-H4 low-power dual-core RISC-V SoC supports 802.15.4 and Bluetooth 5.4 LE
Espressif Systems has formally announced the ESP32-H4 low-power dual-core 32-bit RISC-V wireless microcontroller with support for 802.15.4 and Bluetooth 5.4 LE portfolio after having unveiled it at CES 2024. It’s the first Espressif chip to support Bluetooth 5.4 LE with previous models such as ESP32-H2 or ESP32-C6 only supporting Bluetooth 5.0/5.2. Besides BLE 5.4 support, the new ESP32-H4 dual-core RISC-V WiSoC is an evolution of the ESP32-H2 single-core chip with PSRAM support (up to 4MB built-in), additional GPIOs (36 vs 24), touch sensing GPIOs, and some extra security features such as a power glitch detector also found in the recently announced ESP32-C61. ESP32-H4 specifications: CPU – Dual-core 32-bit RISC-V core (at up to 96 MHz) RAM – 320KB KB SRAM, optional PSRAM up to 4MB Storage – 128KB ROM, External flash support Wireless connectivity IEEE 802.15.4 radio with Zigbee and Thread support, Matter protocol Bluetooth 5.4 (LE) radio designed in-house, […]
Imagination launches the APXM-6200 RISC-V “Catapult” CPU for cost-sensitive consumer and industrial applications
Imagination has expanded its Catapult product portfolio to include a new RISC-V core, the Imagination APXM-6200 CPU. The APXM-6200 is a 64-bit, in-order application processor with an 11-stage, dual-issue pipeline. There isn’t much information on the new Imagination RISC-V core on the product page but we know it offers “best-in-class” performance density, a minimal silicon footprint, and industry-standard security features. The CPU is targeted at intelligent consumer and industrial applications and delivers a 2.5x improvement in performance density and a 65% improvement in normalized performance over comparable Arm Cortex-A53 and other cores on the market. It’s also faster than the Cortex-A510 Armv9 core in SpecINT2k6. Imagination claims that combining the APXM-6200 CPU with their GPUs will ensure a 2x increase in bus utilization and a 2x reduction in memory traffic. It also comes with RISC-V vector extensions, and AI compute libraries and supports fast data coupling with AI accelerators for […]
u-blox ALMA-B1 and NORA-B2, Bluetooth 5.4 LE modules are based on Nordic nRF54H20 and nRF54L15 SoCs
Wireless communication solutions provider, u-blox has added two new modules to its Bluetooth LE portfolio with the ALMA-B1 and NORA-B2 modules built upon the latest low-power wireless nRF54 chips from Nordic Semiconductor. Both modules come in a portable, power-efficient layout and support Bluetooth 5.4 and 802.15.4 (Thread, Matter, Zigbee). The ALMA-B1 and NORA-B2 BLE modules are powered by the low-power nRF54H20 and nRF54L15 SoCs, respectively. This enables them to provide IoT devices with sufficient processing power for edge computing and machine learning without the need for external components. u-blox claims that the ALMA-B1 module provides “more than twice the processing power of previous Bluetooth LE modules” and can replace general-purpose microcontrollers in compact solutions. Also, the NORA-B2 reportedly “consumes up to 50% less current compared to previous generations of Bluetooth LE modules,” leading to longer battery life or smaller batteries in end products. Both modules are classified as open CPU […]