Espressif's New ESP32-S3 Adds AI Features for IoT Devices

Dual-core LX7 32-bit MCU with vector instructions for AI applications.

The very last day of 2020 came with great news for ESP32 fans. The ubiquitous IoT microcontroller is getting a new family member — the ESP32-S3! Espressif's latest introduction includes an updated dual-core MCU, expanded GPIO, AI-focused instructions, and hardware-accelerated security features.

"ESP32-S3 is the result of an overall optimization in terms of connectivity, AI computing power, security, IO availability, and peripheral interfaces."

The ESP32-S3 features a dual-core LX7 32-bit processor operating up to 240 MHz. (The previous -S2 was single-core.) On-chip is 512 KB of SRAM and 384 KB of RAM with support for octal SPI to connect to external memory devices. The RF capabilities include WiFi 2.4 GHz b/g/n, legacy Bluetooth, and Bluetooth Low-Energy 5.0.

Artificial Intelligence applications can take advantage of the ESP32-S3's vector instructions in the MCU. These provide acceleration for neural networks and signal processing loads.

Compared to the base ESP32, the ESP32-S3 has significant hardware security support. These features are similar to the previous single-core ESP32-S2. Security features include an HMAC controller for message authentication. The Digital Signature (DS) module provides hardware-accelerated message signing. Also available is a secure boot capability that verifies the RSA-PSS signature of an application image before executing it. Additional encryption includes "Flash Encryption" for off-chip flash memory. Fully-isolated execution environments are possible with the "World Controller" peripheral. One use of this feature would be to isolate security-sensitive data tasks from the rest of the application.

Not only is ESP32-S3 fast and smart, but it is also flexible. For peripherals, the ESP32-S3 supports up to 44 programmable GPIOs for GPIOs, SPI, I2S, I2C, PWM, RMT, ADC, DAC, UART, SD/MMC host, and TWAI.

Espressif's ESP-IDF platform already provides support for the ESP32-S3. Though, software library support for the AI features may not be immediately available. Watch for updates to the ESP-WHO (face detection) and ESP-Skainet (voice assistant) libraries' GitHub repositories.

For more information, check out the ESP32-S Series SoC product page on Espressif's website. It appears the initial modules incorporating the chip will come in the form of a WROOM and a new S3-DevKit.

baldengineer

Electronics enthusiast, Bald Engineer, and freelance content creator. AddOhms on YouTube. KN6FGY.

Latest Articles