IBM’s Error-Correcting Memory Technology to Enhance Server Reliability
N/A (June 04, 1998) – IBM today announced a new family of error – 04 Jun 1998: correcting semiconductor memory modules that provide system memory reliability up to one-hundred times greater than standard memory modules. This capability is similar to reliability features that IBM developed for the Mars Pathfinder mission. For the first time, this level of reliability will be available to enhance PC server systems.
In the networked computing environment, the need for server reliability is essential. Memory faults, including complete chip failures, have the potential to cause significant system “down time” and data loss. IBM’s new error-correcting products offer a level of “off-the-shelf,” on-board memory fault protection
previously available only in the mainframe environment.
Advanced failure-protection modules become especially important as memory
sizes in server systems increase to one gigabyte and beyond. The standard
“single-error-correct (SEC) error-correction code (ECC)” found on most of
today’s PC servers may not be enough to overcome fatal memory faults. Servers
with large amounts of memory and standard error-correction are projected to
have the same noncorrectable error rate as previous parity-only systems.
Therefore, as memory needs increase, so does the potential for errors.
These failure-protection DIMMs are among the first to provide the self-contained
capability of correcting real-time, multibit DRAM errors, such as complete
DRAM failures. With the use of a high-performance IBM custom chip, enhanced
multibit error detection and correction is incorporated into the module and made
transparent to the system. IBM’s failure-protection modules also allow a PC
server system with existing error-correction capabilities to be upgraded to
support mainframe-class, multibit failure correction.
IBM’s error-correcting memory modules exemplify the company’s leadership in
the area of “on-the-fly” error correction code capabilities. This new family of
products complements IBM’s broad range of offerings, including
high-performance, value-added memory technologies such as ultra-high speed
SRAM, stacked DRAM and SDRAM, SGRAM, VRAM, high-density DIMMs,
ECC-on-board SIMMs, and low-power SO DIMMs.
Memory Modules with Failure Protection: Technical Features and Benefits
Up to 100 times improvement in system memory reliability.
Transparent multibit error protection in an existing single error correction
(SEC) system.
168-pin JEDEC-standard, Dual In-Line Memory Module (DIMM).
Extended Data Out (EDO) Mode, Read-Modify-Write Cycles.
High-speed IBM ASIC technology used to perform “on-the-fly” error
correcting function, while maintaining system access and cycle time
requirements.
Serial (unbuffered) or parallel (buffered) presence detect.
Source: IBM