Choosing between direct code access and bulk storage
NOR Flash and NAND Flash are two Flash memory architectures that retain data without power. NOR is particularly useful for fast random reads and direct code execution. NAND emphasizes high storage density and a lower cost per bit, especially at large capacities.
Those are design tendencies, not exclusive roles. Both can store programs and data. Choosing nonvolatile memory means deciding how the processor will access the contents, how updates will be performed and who will manage the media.
The difference starts inside the chip
NOR cells connect to the internal bit line through parallel branches. NAND places multiple cells in a series string, helping reduce the array area needed for a given capacity. This describes internal cell connections, not the external communication bus.
There are both SPI NOR and SPI NAND devices. An eight-pin package or SPI label therefore does not identify the storage architecture. Check the full part number and datasheet.
The main differences at a glance
| Characteristic | NOR Flash | NAND Flash |
|---|---|---|
| Reading | Well suited to fast reads of small data units at selected addresses. | Array reads generally load a page into an internal buffer. |
| Code execution | Can support XIP with a suitable processor and controller. | Code is commonly loaded into RAM for execution. |
| Programming | Byte, word or page/buffer programming, depending on the device. | Page-based programming with device-specific partial-write limits. |
| Erasing | Sector or block operations. | Block operations affecting multiple pages. |
| Density / cost | Common for small and medium code-storage requirements. | Typically offers lower cost per bit at large capacities. |
| Error handling | Some products include internal ECC and additional protection. | ECC and bad-block management are important design requirements. |
| Typical uses | Boot code, embedded firmware and lookup tables. | SSDs, eMMC, memory cards and large data stores. |
“Which is faster?” needs a workload. Random-read latency, sequential throughput, programming and erasing are different measurements. Compare capacity and performance together with the interface, controller, caching and access pattern.
Why NOR is useful for executable code
Running software involves fetching instructions from different addresses, especially when branches occur. NOR's random-read capability suits this pattern. Execute-in-place, or XIP, allows instructions to be fetched from Flash without first copying the entire program into RAM.
A generic SPI connection does not automatically provide XIP. The processor and memory controller must support the required access mode. A system may also copy code from NOR into SRAM or DRAM. NAND boot implementations commonly read the application into RAM before running it; NAND is fully capable of storing that application.
The processor's earliest startup instructions may reside in Mask ROM. That first-stage code can then load or start software stored in external NOR or NAND. First-stage boot code and application storage need not occupy the same memory.
A page read is different from a block erase
Consider an illustrative NAND layout with a 2 KiB main-data page and 64 pages per block. Each block contains 128 KiB of main data. Actual devices can have different page sizes, spare areas and block organizations.
A page read can move data into the chip's buffer before the requested bytes are transferred over the interface. Erasing a block affects every page in that block. Updating one record may therefore require preserving other valid records and placing the replacement elsewhere.
NOR has an important distinction too: reading an individual byte does not mean that byte can be erased individually. In either Flash architecture, fast reads do not remove programming restrictions, erase requirements or completion checks.
Error correction and media management
Error-correcting code, or ECC, detects and corrects a limited number of bit errors. NAND's spare area can hold ECC and management information, depending on the device and system. A bad block does not meet the requirements for reliable use; it may be marked at manufacture or become unusable later.
With raw NAND, the host system supplies the required management. Managed NAND such as eMMC includes a controller that handles ECC, bad blocks and wear leveling. A NAND chip with on-die ECC alone is not necessarily a fully managed storage device.
For endurance and reliability, look beyond the NOR/NAND label. SLC, MLC, TLC and QLC describe bits stored per cell, not serial versus parallel interfaces. Check program/erase endurance, retention and temperature conditions for the selected product. NOR can also include ECC; error correction is not exclusive to NAND.
Applying the distinction to industrial equipment
A PLC module might use NOR for boot firmware and NAND for larger event archives, images or long-term records. An industrial IoT device can use both on the same board. These application examples illustrate different workloads rather than a mandatory architecture.
For small records updated very frequently, FRAM or MRAM may deserve consideration. Include energy consumption, update frequency and the complete system cost in the comparison.
Technical references: KIOXIA — NOR and NAND architecture · Infineon — NOR and XIP · Micron — Raw and managed NAND · Winbond — NAND error management · Infineon — Sector erase · KIOXIA — SLC, TLC and 3D NAND · Micron — Booting from NAND
