Flash memory

Introduction

ESP8266 flash memory sizes vary from 512Kbytes on the ESP-01 up to 4Mbytes on the ESP12F. Up to 16MBytes are supported for custom designs. Rp2040 has similar support, and the Esp32 has enhanced VMM (Virtual Memory Management) hardware.

Sming version 4.3 introduced partition tables to support multiple architectures, different hardware variants and custom flash layouts without restriction. It is binary compatible with the Esp32 IDF partition tables but with a consistent API across all architectures.

See Hardware configuration for details.

A typical layout for a 4MByte Esp8266 device might look like this:

Address

Config variable

Size

Source filename

Description

(hex)

(if any)

(KB)

(if applicable)

000000

1

rboot.bin

Boot loader

001000

4

rBoot configuration

002000

ROM_0_ADDR

rom0.bin

First ROM image

102000

ROM_1_ADDR

rom1.bin

Second ROM image

200000

RBOOT_SPIFFS_0

300000

RBOOT_SPIFFS_1

3FA000

4

Partition table

3FB000

4

blank.bin

RF Calibration data (Initialised to FFh)

3FC000

4

esp_init_data_default.bin

PHY configuration data

3FD000

12

blank.bin

System parameter area

Note

This was the previous layout for a 4MByte flash device:

Address

Config variable

Size

Source filename

Description

(hex)

(if any)

(KB)

(if applicable)

000000

1

rboot.bin

Boot loader

001000

4

rBoot configuration

002000

ROM_0_ADDR

rom0.bin

First ROM image

100000

RBOOT_SPIFFS_0

202000

ROM_1_ADDR

rom1.bin

Second ROM image

300000

RBOOT_SPIFFS_1

3FB000

4

blank.bin

RF Calibration data (Initialised to FFh)

3FC000

4

esp_init_data_default.bin

PHY configuration data

3FD000

12

blank.bin

System parameter area

The actual layout in use can be seen by running make map.

Speed and caching

Flash memory is accessed via an external SPI bus, so reading it takes about 12x longer than reading from internal RAM. To mitigate this, some of the internal RAM is used to cache data. Part of this is managed in hardware, which means if the data required is already in the cache then there is no difference in speed. In general, then, frequently accessed data is read as if it were already in RAM.

Bear in mind that every time new data is read via the cache, something else will get thrown away and have to be re-read. Therefore, if you have large blocks of infrequently accessed data then it’s a good idea to read it directly using flashmem_read(). You can get the address for a memory location using flashmem_get_address().

See Program Space for details of how to store data in flash, and access it.