Short answer: Move from SPI NOR to SPI NAND when you need around 1 Gb or more for firmware, a file system and resources, and the cost of 256 Mb to 512 Mb NOR is rising. Both use the same 8-pin Quad SPI style interface, but SPI NAND ships with some bad blocks, needs ECC (most parts have it on die) and cannot execute code in place (XIP). Before choosing, fix three things: the capacity, who handles bad blocks and ECC (the device, the controller or the file system), and the I/O voltage (3.3 V or 1.8 V). Many designs combine a small NOR for boot code with a larger SPI NAND for data.

SPI NAND vs SPI NOR

SPI NANDSPI NOR
Typical capacity512 Mb to several GbUp to 512 Mb; above that it gets expensive
Cost at high capacityLowRises quickly
Bad blocksPresent from the factory; needs bad block managementNone
ECCRequired; most parts include on-die ECCUsually not needed
Execute in place (XIP)No; code is copied to RAM firstYes
Random readLoad a whole page into the cache register, then read it outTrue random byte access, low latency
Erase unitBlock, typically 128 KBSector, typically 4 KB
OrganizationPages (typically 2 KB plus spare area) grouped into blocksLinear address space
Typical useFile systems, Linux rootfs, images, fonts, audio, logs, OTA imagesBoot code, XIP, small parameters and fonts

For a broader comparison of the two technologies, see NOR Flash vs NAND Flash.

Common 3.3 V and 1.8 V SPI NAND parts

All of the following are SLC SPI NAND parts on our site. Pinouts at the same capacity and package are broadly aligned, but ECC status bits, bad block marking and some commands differ, so a second source needs driver work (see below).

Unless marked 1.8 V, these are 2.7 V to 3.6 V parts. Every maker also offers 1.8 V versions with a different code, so match the part to the platform I/O voltage before ordering. The JEDEC manufacturer IDs differ: Winbond 0xEF, GigaDevice 0xC8, Micron 0x2C and Kioxia (Toshiba) 0x98.

When to switch from SPI NOR to SPI NAND

  • Capacity: at 256 Mb (32 MB) and above, NOR becomes expensive; from about 1 Gb, SPI NAND gives several times the capacity for similar money.
  • File system rather than boot only: Linux rootfs (UBI/UBIFS, SquashFS), LittleFS or FAT under an RTOS, images, fonts, audio, maps and dual OTA images all suit NAND.
  • No XIP needed: the bootloader or SoC copies code to SDRAM or PSRAM and runs it there. If your MCU must fetch instructions directly from flash, keep a NOR.
  • The software stack handles bad blocks and ECC: Linux MTD and UBI, a vendor SDK driver, or a device with on-die ECC plus a file system with bad block management. Estimate the cost of writing a bare-metal NAND driver before committing.

Selection checks and common pitfalls

  • Bad blocks are normal: NAND is allowed to ship with a limited number of bad blocks, and datasheets typically specify that block 0 is valid. Read the factory bad block marks at first power-up and let the file system or UBI manage new ones. Never treat factory bad blocks as a defective part.
  • Always use ECC: SLC NAND has bit flips that must be corrected. Prefer on-die ECC, read the ECC status bits (corrected, uncorrectable, refresh recommended) and include them in health monitoring.
  • No XIP: reading even one byte means loading a page into the cache register first, which takes tens of microseconds. Do not expect the first instruction after reset to come from SPI NAND unless the SoC boot ROM supports SPI NAND boot.
  • Voltage first: choose 3.3 V or 1.8 V versions to match the I/O. Never connect a 1.8 V device to 3.3 V I/O.
  • Page and block size affect the file system: page size can be 2 KB or 4 KB depending on capacity. Configure UBI/UBIFS or LittleFS page, block, spare-area and ECC layout for the exact device, and update them when you change part or capacity.
  • Do not hard-code one ID or one bad block scheme: the four makers can second-source each other, but IDs, on-die ECC enable methods, status bit definitions and bad block marker positions differ. Let the bootloader and driver identify the device, ideally through a generic framework such as Linux MTD SPI-NAND.
  • Match the controller: check the QSPI clock, continuous or buffer read support and whether the controller handles the device's ECC status.
  • Leave headroom: add up images, file system, dual OTA copies, bad and reserved blocks and wear-leveling margin, then keep 20% to 40% spare capacity.

When SPI NOR is still the better choice

  • XIP is required: MCUs that run code directly from flash, bootloaders and FPGA configuration bitstreams that must be read at power-up. Examples are Winbond W25Q, GigaDevice GD25 and Micron MT25Q.
  • Capacity is small (128 Mb or less): small firmware, parameters, fonts and configuration are simpler on NOR, with no bad blocks and no ECC.
  • Small but critical data: boot images and key parameters in automotive or safety products.

The usual combination is a small NOR for boot and XIP code and a larger SPI NAND for the file system and resources.

Send us your requirements

Send us the capacity, I/O voltage, package, temperature grade, controller and yearly volume. We will suggest SPI NAND and matching SPI NOR options and compare second sources pin by pin. We reply to routine inquiries within 1 business day. Submit your BOM.