>
>
> Is not a bug, is the Opteron architecture, all processors have a path with
> the other for snoops requests, IBM enginners told me they help to AMD with
> redefine this situation trying to cross paths to have redundancy between
> processors.
> >


An unwanted reboot?
This seems more thana bug to me.

I know pretty well the AMD architecture and the hypertransport but this is
the first time I read of this kind of rebbot.
I wish to know more.

May you describe the issue and how to reprodece it?

Any statspack/AWR of your benchmark available?

thanks
Fabrizio