This is a discussion on Chipkill protection and ECC error logging - SUN ; When using an Opteron, How do I ensure that the processor's ChipKillEccEn is set, and that logging is enabled in Solaris? Is it handled solely by the OS or is BIOS support required? Does the OS log every corrected soft ...
When using an Opteron, How do I ensure that the processor's
ChipKillEccEn is set, and that logging is enabled in Solaris? Is it
handled solely by the OS or is BIOS support required? Does the OS log
every corrected soft error, or only hard errors or uncorrectable
errors? Does the log distinguish between bit errors detected in SECDED
(plain ECC) mode and symbol errors detected in SSCDSD (chipkill) mode?
Has Sun tested the OS logging functionality by injecting hardware ECC
errors into a running system?
The Ultra 40 M2 offers chipkill memory, but the Ultra 20 M2 offers
just non-chipkill ECC memory. How could the Ultra 20 M2 provide ECC
but fail to provide chipkill protection, other than by using x8 DRAM
chips instead of x4 chips? If I put two DIMMs each with 18 x4 chips in
an Ultra 20 M2 and set ChipKillEccEn, will the system have chipkill
Is memory scrubbing enabled by default by either system's BIOS, or is
enabling it the responsibility of the OS?
Is Nehalem going to have integrated chipkill functionality in its
memory controller, and ECC error logging, and automatic memory
scrubbing, like Opterons do? How long will it take for Sun to produce
a Nehalem-based workstation, assuming Intel ships the chips on