PC Hardware in a Nutshell, 3rd Edition

Parity and ECC Memory

Two types of memory use extra bits to store checksum information that can detect and sometimes correct memory errors. They are as follows:

Parity memory

Uses nine bits per byte to provide simple error detection, so simple as to be useless on modern PCs. Most early DIP-based PCs required parity memory, and so used nine chips per row. SIMMs, which are much more reliable than DIPs, made parity unnecessary. The Intel Triton Pentium-class chipset and competing chipsets removed support for parity memory. Recent chipsets again support parity memory, but there is no good reason to use it.

Although parity memory is still sold, buy it only if your computer requires it. Some computers do not allow mixing parity and nonparity modules. Before buying parity memory, verify which type is already installed and whether your chipset can mix parity with nonparity memory. Identify parity modules by counting the chips on the module. If that number is evenly divisible by three, it’s a parity module. If nonparity memory is installed, or if parity memory is installed but your chipset allows mixing types, buy nonparity memory. Buy parity memory only if parity memory is already present and your chipset does not allow mixing types.

ECC memory

An enhanced form of error-checking memory that can detect single- and multibit errors and correct all single- and some multibit errors. ECC originated on mainframes, migrated to minicomputers, and is now common on PC servers.

ECC memory requires five ...