IBM x345 Dual 2.8Ghz Xeon poor performance

Wojciech Puchar wojtek at wojtek.tensor.gdynia.pl
Tue Jun 2 15:22:23 UTC 2009


>> constantly doing ECC that's why it was damn slow.
>>
>> anyway - can hardware give any info for OS about how often ECC corrects
>> errors?
> This feels like the right track, I'll run memtest86 on it later tonight.
>
if i'm right memtest86 will not detect anything as too - all errors get 
corrected.

for example - on your DIMM with 72-bit bus (64+8) one pin is dirty and is 
not connected well. then you'll get a single bit error every few 
reads/writes, and all will be corrected.

If you can disable ECC - do it, and then run memtest so it will detect 
errors.

If you can't, remove all but one DIMM, check if speed improved, if so, 
remove this and put other DIMM etc.. until you'll find what is bad.

Or maybe it will then work fine because it's just contact problem.


More information about the freebsd-questions mailing list