Worldwide

Memory Sparing technology

You are here: Home / Products & Services / Servers & Storage / Memory Sparing technology

Memory sparing uses a spare online bank to provide DIMM fail-over capabilities when a pre-defined threshold of single-bit correctable errors is reached. The pre-defined threshold is set in BIOS and cannot be change by the user.


DIMM capacity and configuration requirements:

  • Memory is 4-way interleaved and as a result the DIMMs must be installed in banks consisting of 4 DIMMs per bank.
  • The DIMMs in a bank must be of the same capacity, manufacturer, and have the same SPD characteristics. No mixing of memory DIMMs is allowed.
  • Bank #1 and bank #2 are defined as the "Active banks". Bank #3 is defined as the spare bank. For memory sparing to function, DIMMs must be installed in the active and spare banks.

The table below lists the configuration for memory sparing:

  Bank #1(Active Bank) Bank #2(Active Bank) Bank #3(Spare Bank)  
4 DIMM configuration
Installed Not installed Not installed Not Available
8 DIMM configuration
Installed Not installed Installed Available
12 DIMM configuration Installed Installed Installed Available

NOTE. The spare bank will be removed from the memory pool and held as a spare; for instance, if 12 x 1 GB DIMMs are installed, only 8 GB of that 12 GB is available for use by the operating system.

How memory sparing works:

In normal mode of operation, the system reads/writes memory data to the active bank(s) (bank #1 and bank #2). The system does not read/write memory data to the spare bank (bank #3). Bank #3 does not store memory data before memory data sparing begins. The memory sparing function moves memory data from the bank registering errors to the spare bank, if the correctable error count exceeds the threshold value. After moving the memory data, the system de-activates the bank registering errors and records an error log. Further, the system status LED will blink.

Memory sparing does not recover non-correctable errors. Therefore, if the system detects non-correctable errors, the system will assert fatal error status, and the system will halt.



The system is not able to use memory mirroring and memory sparing concurrently.