Thanks mibo. ECC on consumer boards such as biostar are SECDED and I expect that single bit error get corrected transparently, while uncorrectable (ie multiple bit) raises an immediate kernel halt, whatever the OS.
ECC logging is a nice functionality which allows to detect a failing/weak RAM module and take preventive action before a hard error occurs which halt the system, but it is not strictly mandatory.
Regarding server vs consumer grade boards, on AMD64 the support should be the same independently of the chipset, once ECC operation has been configured (normally by the bios). Logging may require chipset-dependant software support (I'm not sure), but at least standard ECC operation should be supported in any case (transparent correction of single bit error and raised interrupt on uncorrectable error leading to a kernel halt)
I plan to use OpenSolaris or FreeBSD, still not decided.
ECC logging is a nice functionality which allows to detect a failing/weak RAM module and take preventive action before a hard error occurs which halt the system, but it is not strictly mandatory.
Regarding server vs consumer grade boards, on AMD64 the support should be the same independently of the chipset, once ECC operation has been configured (normally by the bios). Logging may require chipset-dependant software support (I'm not sure), but at least standard ECC operation should be supported in any case (transparent correction of single bit error and raised interrupt on uncorrectable error leading to a kernel halt)
I plan to use OpenSolaris or FreeBSD, still not decided.
Zuletzt bearbeitet:

) I had a close look at 785G motherboards with an AM2+ socket and was actually quite disappointed. But I had the opposite feeling you've got, I expected 785G AM3 to be more efficient, both because of the platform itself (with it's processor, memory, ..) and the fact manufacturers improved the boards (with fancy terms as DrMos).


