Chinese chip designer Espressif has developed a dual core XTensa wireless system on chip for AI in the Internet of Things (AIoT).
The ESP32-S3 uses two 240MHz extended XTensa LX7 cores from Cadence Design Systems with 2.4GHz WiFi and Bluetooth Low Energy 5.0 connectivity. It has 44 programmable GPIOs and supports larger, high-speed octal SPI flash than the previous ESP32 devices and PSRAM with configurable data and instruction cache .
The ESP32-S3 AIO chip supports a 2.4 GHz Wi-Fi (802.11 b/g/n) with 40 MHz of bandwidth support while the BLE subsystem supports long range through Coded PHY and advertisement extension. It also supports higher transmission speed and data throughput, with a 2Mbit/s PHY.
The XTensa LX7 cores have been extended with vector instructions for acceleration for neural network computing and signal processing workloads. Espressif is currently working on the software libraries for AI functions as part of the updated ESP-WHO and ESP-Skainet development tools.
ESP32-S3 has 44 programmable GPIOs, ten more than the ESP32 and supports all the commonly-used peripherals, such as SPI, I2S, I2C, PWM, RMT, ADC, DAC and UART, SD/MMC host and Espressif's Two-Wire Automotive Interface (TWAI). In total, 14 GPIOs can be configured as capacitive touch input for HMI applications. Apart from all these peripherals, however, ESP32-S3 is also equipped with an ultra-low-power (ULP) core that supports multiple low-power modes in a variety of such use-cases.
The AIoT chip supports AES-XTS-based flash encryption and RSA-based secure boot with a digital signature peripheral and an HMAC module, which provide functionality that is similar to the hardware secure element, protecting the private or symmetric key from software attacks and identity theft. ESP32-S3 also has a “World Controller” peripheral that provides two fully-isolated execution environments, which enable the implementation of a trusted-execution environment or a privilege-separation scheme.
With processor cores with vector extensions the development environment is key, and Espressif’s ESP-IDF platform that already powers millions of devices on the market.