According to some documents that's I've seen, such as for example in the following link Digit-Life AMD facts' ">http://www.digit-life.com/arti...mmer-family/index.html, the latency between two adjacent Opterons is 118ns, whereas for a couple of Opterons 2 hops away from each other it's more like 136ns. I assume the extra delay is due to having to travel through an extra processor on the way to the final destination. My question is is this latency figure for initial access only, or is it for all subsequent access?