English
Time: August 24th, 2026
Browse: 5,489
If you are confused about whether to use an ESP32 or ESP32-S3 board for your project, this comparison will help you decide. Although their names are similar, choosing the wrong board can cause compatibility problems, extra development work and unnecessary cost. The better option depends on your project requirements. This article compares their practical differences, performance, software support, published project examples and important limitations.

Figure 1. ESP32 and ESP32-S3 Development Boards
The terms ESP32 board and ESP32-S3 board refer to development boards based on Espressif microcontrollers. They are not single board models. Manufacturers offer different versions with varying modules, memory capacities, connectors and pin layouts.

Figure 2. ESP32 Development Board
An ESP32 development board is built around an original ESP32 chip or module, such as the ESP32-WROOM-32. It combines the wireless microcontroller with supporting components for power, programming and connection to external circuits. Its input/output pins are normally exposed through headers, making the board suitable for developing and testing embedded systems, wireless devices and Internet of Things projects.

Figure 3. ESP32-S3 Development Board
An ESP32-S3 development board is built around an ESP32-S3 chip or module. The ESP32-S3 is a newer microcontroller designed for wireless embedded systems, USB devices, human-machine interfaces and workloads involving signal processing or small machine-learning models. These boards provide the supporting circuits needed to power and program the microcontroller and connect it to external hardware. Most available GPIO pins are exposed through headers for breadboard prototyping and custom circuit development.
|
Specification |
ESP32 Chip |
ESP32-S3 Chip |
|
Main
microcontroller |
Original
ESP32 series |
ESP32-S3
series |
|
CPU
architecture |
Xtensa
32-bit LX6 |
Xtensa
32-bit LX7 |
|
CPU
cores |
Usually
dual-core; some ESP32 variants are single-core |
Dual-core |
|
Maximum
CPU frequency |
240
MHz |
240
MHz |
|
Main
internal SRAM |
520
KB |
512
KB |
|
RTC
SRAM |
16
KB |
16
KB |
|
ROM |
448
KB |
384
KB |
|
Flash
memory |
Depends
on the installed module and board |
Depends
on the installed module and board |
|
PSRAM |
Optional;
often included on WROVER-based boards |
Optional;
commonly available on higher-memory boards |
|
Wi-Fi |
2.4
GHz IEEE 802.11b/g/n |
2.4
GHz IEEE 802.11b/g/n |
|
Maximum
Wi-Fi rate |
Up
to 150 Mbit/s |
Up
to 150 Mbit/s |
|
Bluetooth |
Bluetooth
4.2 BR/EDR and Bluetooth Low Energy |
Bluetooth
5 Low Energy |
|
Bluetooth
Classic |
Supported |
Not
supported |
|
GPIOs
implemented by the chip |
Up
to 34 |
Up
to 45 |
|
GPIOs
available on the board |
Depends
on the module and board layout |
Depends
on the module, memory configuration and board layout |
|
ADC |
Two
12-bit SAR ADCs, up to 18 channels |
Two
12-bit SAR ADCs, up to 20 channels |
|
DAC |
Two
8-bit DAC channels |
No
integrated DAC |
|
Capacitive-touch
inputs |
10 |
14 |
|
UART
controllers |
3 |
3 |
|
I²C
controllers |
2 |
2 |
|
I²S
controllers |
2 |
2 |
|
General-purpose
SPI controllers, excluding flash/PSRAM interfaces. |
2 |
2 |
|
TWAI
controller |
Supported;
external transceiver required |
Supported;
external transceiver required |
|
Native
USB |
Not
available on the chip; boards normally use a USB-to-UART bridge |
Full-speed
USB OTG and USB Serial/JTAG supported |
|
USB
connector |
Depends
on the board; Micro-USB and USB-C versions exist |
Depends
on the board; some boards provide separate native USB and USB-to-UART ports |
|
LCD
and camera interface |
No
dedicated parallel LCD and camera controller |
Dedicated
LCD and camera interface |
|
Vector-processing
instructions |
No
ESP32-S3-style vector extensions |
SIMD
instructions for signal-processing and neural-network workloads |
|
Antenna |
PCB
antenna or external-antenna connector, depending on the module |
PCB
antenna or external-antenna connector, depending on the module |
This table compares chip-level specifications. Flash, PSRAM, connectors, exposed GPIOs and power consumption vary by development board.

Figure 4. ESP32 Functional Block Diagram

Figure 5. ESP32-S3 Functional Block Diagram
The official block diagrams show which functions are built into each chip. The original ESP32 includes Bluetooth Classic support, two DAC channels and an Ethernet MAC, but it has no native USB controller. The ESP32-S3 adds USB OTG, USB Serial/JTAG, GDMA, dedicated parallel LCD and camera interfaces, and additional security blocks such as HMAC, RSA digital-signature support and permission control. Its LX7 processor also includes SIMD instructions, although these are not labelled separately in the diagram. However, the S3 does not include the original ESP32’s DAC or Ethernet MAC. Compare these hardware blocks carefully because missing functions may require external components or software changes.
Espressif reports 1,079.96 CoreMark for the dual-core ESP32 and 1,329.92 for the ESP32-S3, both at 240 MHz. The ESP32-S3 score is therefore 23.1% higher, although this does not guarantee the same improvement in every application.
The ESP32-S3 uses the newer Xtensa LX7 architecture and includes SIMD instructions for supported signal-processing and machine-learning operations. These instructions provide an advantage only when the code or library has been optimized to use them.
The ESP32 contains 520 KB of internal SRAM, while the ESP32-S3 contains 512 KB. These totals are not the amount of SRAM available to an application because firmware, wireless stacks, caches and system services also use internal memory. Some boards add external PSRAM, which increases capacity but is generally slower than internal SRAM and may not support every type of data or operation.
Both chips support 2.4 GHz IEEE 802.11b/g/n Wi-Fi with a specified maximum PHY rate of 150 Mbit/s. This value is not guaranteed application throughput. Actual speed and range depend on the antenna, board layout, interference, obstacles, protocol overhead and access point.

Figure 6. ESP32 and ESP32-S3 Wi-Fi and Bluetooth Differences
The main difference is Bluetooth support. The ESP32 supports Bluetooth 4.2 BR/EDR and Bluetooth Low Energy, while the ESP32-S3 supports Bluetooth 5 Low Energy only. The S3 provides a 2 Mbit/s PHY, coded PHY modes and extended advertising. However, it cannot use Bluetooth Classic profiles such as SPP and A2DP. Projects that depend on Classic Bluetooth must use the original ESP32 or migrate to BLE or another communication method.
Espressif lists typical chip-level values of approximately 0.8 mA in light sleep and 10 µA in deep sleep for the ESP32. For the ESP32-S3, the specified 240 µA light-sleep value assumes that Wi-Fi and VDD_SPI are powered down and the GPIOs are in a high-impedance state. Its deep-sleep current is approximately 8 µA when RTC memory and RTC peripherals remain powered, or 7 µA when only RTC memory remains powered.
These sleep-current figures were measured under different operating conditions and are not a direct board-to-board comparison. They also exclude the complete development board. Voltage regulators, USB bridges, power LEDs, PSRAM and connected components can increase current consumption. Wireless activity and processor wake time can also have a much greater effect on battery life than sleep current alone.
A simple estimate is:
Runtime in hours ≈ usable battery capacity in mAh ÷ average current in mA
Example: a 2,000 mAh battery powering a system with a measured average current of 10 mA provides approximately 200 hours, or 8.3 days, of ideal runtime before regulator and battery losses.
This formula assumes that capacity and current are measured on the same voltage side. Regulator losses, battery cut-off voltage, temperature and aging reduce the usable capacity. For the most reliable estimate, measure the average battery current during the finished system’s normal sleep, processing and wireless operating cycle.
ESP32 and ESP32-S3 development boards are not pin-compatible. Header positions, exposed GPIOs, strapping pins, USB connections and memory-reserved pins vary by board. Always check the exact board schematic before transferring existing wiring.
|
Pinout
difference |
ESP32
DevKitC V4 |
ESP32-S3-DevKitC-1 |
|
Header
layout |
Two
19-pin headers; 38 positions |
Two
22-pin headers; 44 positions |
|
USB
connections |
One
Micro-USB port connected through a USB-to-UART bridge |
Separate
USB-to-UART and native USB OTG/Serial-JTAG ports |
|
UART0
header pins |
TX
on GPIO1 and RX on GPIO3 |
TX
on GPIO43 and RX on GPIO44 |
|
Native
USB pins |
None |
USB
D− on GPIO19 and USB D+ on GPIO20 |
|
Strapping
pins |
GPIO0,
GPIO2, GPIO5, GPIO12 and GPIO15 |
GPIO0,
GPIO3, GPIO45 and GPIO46 |
|
Memory-related
restrictions |
GPIO6–11
are connected to flash and should not be used. GPIO16–17 depend on the
installed module. |
GPIO26–34
are not exposed. GPIO35–37 are also unavailable on variants using Octal flash
or PSRAM. |
|
Onboard
RGB LED |
No
addressable RGB LED |
GPIO38
on board revision v1.1; GPIO48 on v1.0 |
Sources: Espressif’s official ESP32-DevKitC V4 guide and ESP32-S3-DevKitC-1 v1.1 guide.
Both the ESP32 and ESP32-S3 are officially supported by the Arduino-ESP32 core and ESP-IDF. In Arduino IDE, select the correct board and configure its flash, PSRAM, partition and USB settings. Basic Wi-Fi, GPIO, UART, I²C and SPI source code can often be reused, but the program must be compiled separately for each target.
In ESP-IDF, select the target before building:
• ESP32: idf.py set-target esp32
• ESP32-S3: idf.py set-target esp32s3
Source-code compatibility does not guarantee hardware compatibility. Board pin assignments, strapping pins, available memory and built-in peripherals can differ. A library may compile for both targets but still require different pin definitions, wiring or configuration. Code that depends on chip-specific hardware must also be changed.
Always confirm that third-party libraries support the selected chip and framework version. If compatibility was tested directly, record the exact Arduino-ESP32 version, ESP-IDF version, board model and library versions used. Do not add version numbers unless those versions were tested.
Fanariotis, Orphanoudakis, and Fotopoulos (2024) compared an ESP32 DevKit TTGO and ESP32-S3-DevKitC-1 using four identical ESP-DL models. Each test used one processor core, a 3.3 V supply, and a Keysight 34465A digital multimeter.
With the same 32 KB data cache, the two-stage human-face model took 419,556 µs on the ESP32 and 131,160 µs on the ESP32-S3. Their standby-subtracted energy values were 8.558 µWh and 3.166 µWh, respectively. Cat-face detection required 146,592 µs and 3.277 µWh on the ESP32, compared with 22,388 µs and 0.609 µWh on the ESP32-S3.

Figure 7. ESP32 vs ESP32-S3 Machine-Learning Energy Comparison. Source: Figure 5, Fanariotis, Orphanoudakis, and Fotopoulos (2024)
As shown in the figure, the ESP32-S3 used less processing energy across all four models when using SIMD/vector instructions. Its average power was higher, but the SIMD instructions completed inference much faster. However, the boards had different PSRAM configurations, and the tests excluded cameras and wireless communication. Trial counts and measurement variation were also not reported, so the results apply only to the tested workloads.
|
Project
requirement |
ESP32 |
ESP32-S3 |
Recommended
chip |
|
Bluetooth Classic |
Supports Bluetooth 4.2 BR/EDR |
Not supported |
ESP32 |
|
Analog DAC output |
Two integrated 8-bit DAC channels |
No integrated DAC |
ESP32 |
|
RMII Ethernet |
Integrated MAC; external PHY required |
No integrated Ethernet MAC |
ESP32 |
|
Native USB |
Requires a USB-to-UART bridge or
external USB controller |
Supports USB OTG and USB Serial/JTAG |
ESP32-S3 |
|
Bluetooth 5 LE |
Limited to Bluetooth 4.2 LE |
Supports Bluetooth 5 LE features |
ESP32-S3 |
|
Camera applications |
Possible, but with less dedicated
support |
Dedicated camera interface and stronger
image-processing support |
ESP32-S3 |
|
LCD applications |
Uses general-purpose peripheral
interfaces |
Dedicated parallel LCD interface |
ESP32-S3 |
|
Machine learning |
Suitable for smaller or basic models |
LX7 cores and SIMD/vector instructions
accelerate optimized workloads |
ESP32-S3 |
Development-board memory, exposed pins, USB ports and connectors vary. Check the exact board schematic before making the final choice.
Technical References
• Espressif Systems. ESP32 Series Datasheet, Version 5.3, July 2026.
• Espressif Systems. ESP32-S3 Series Datasheet, Version 2.2, March 2026.
• Espressif Systems. Arduino-ESP32 Programming Guide.
• Espressif Systems. ESP-IDF Programming Guide: ESP32 Getting Started.
• Espressif Systems. ESP-IDF Programming Guide: ESP32-S3 Getting Started.
• Espressif Systems, Arduino-ESP32 Supported SoCs.
• Fanariotis, A., Orphanoudakis, T., and Fotopoulos, V. (2024). “Reducing the Power Consumption of Edge Devices Supporting Ambient Intelligence Applications.” Information, 15(3), 161.
Espressif reports a dual-core CoreMark score that is 23.1% higher for the ESP32-S3 at 240 MHz. However, the actual improvement depends on the workload. Applications optimized for its SIMD/vector instructions may gain more, while other applications may show a smaller difference.
No. The ESP32-S3 supports Bluetooth 5 Low Energy only. Projects requiring Bluetooth Classic profiles such as SPP or A2DP should use the original ESP32 or another compatible chip.
Not if the sensor outputs a 5 V signal. Both devices use 3.3 V logic, so a voltage divider or level shifter is required for 5 V outputs. Also ensure that I²C pull-up resistors connect to 3.3 V, not 5 V.
ADC2 shares hardware resources with Wi-Fi on both chips. ESP32 ADC2 access can be blocked, while ESP32-S3 ADC2 readings may fail during Wi-Fi operation. Use ADC1 for reliable Wi-Fi-connected sensor measurements.
No. Twelve-bit resolution provides 4,096 output levels, but noise, reference-voltage variation, attenuation and circuit design affect accuracy. Use ADC calibration, averaging and input filtering, or an external precision ADC when measurement accuracy is important.
PSRAM increases memory capacity but is generally slower than internal SRAM. It is useful for camera frames, display buffers and machine-learning data, while time-critical code, DMA descriptors and frequently accessed data should remain in internal memory.
Both support secure boot and flash encryption, but the ESP32-S3 adds features such as hardware-assisted digital signatures, HMAC and improved memory protection. These protections must be correctly configured; choosing the S3 alone does not make a device secure.
CAP CER 7.9PF 25V C0G/NP0 01005
CAP CER 39PF 50V C0G 0201
IC MCU 8BIT 4KB FLASH 14SOIC
IC OPAMP GP 2 CIRCUIT 8SOP
SP124104CFAE Freesca
MRFIC1830DMR2 TSSOP
ASAHI SOP6
IC REG CONV I.MX6 12OUT 56QFN
ATHENAV9.38R ST
ICS1575-01G ICS
K4S643232E-TC-60 SAMSUNG
IC GATE DRVR HIGH-SIDE DSO8-907



