Tri-bonded? Have you tried this? What MPI stack and/or other interconnect software do you plan to use?
My understanding is that kernel support for bonding has been vastly improved. This is only through anectodtal accounts of others. I have not yet tried it myself.
Instead of using MPI, I am writing my own message-passing protocol based on UDP Multicast with Forward Error Correction. I will be using the cluster to run a semantic web-crawler with some of the features found in IBM's WebFountain. So, the entire cluster will be rack-mounted in a data center.
However, the older non-PCI-Express versions of that MSI K8N Master2-Far board all connected only ONE of the two Opterons to the DIMMs. Thus the 2nd Opteron has to do all its memory access via the 1st Opterons HT link, so the 2nd Opteron sees more memory latency, and probably more important for you, the total aggregate memory bandwith is only 1/2 what you'd get with a real server-grade dual Opteron board. I've no idea whether or not that is still the case with the current MSI K8N Master2-Far or not, but it's something you'll want to check carefully when considering those sorts of motherboards...
I wasn't aware of the memory bandwidth issues with the MSI K8N Master2-Far. This is good to know. Thanks for the tip. I am interested in dual-opteron boards such as the Asus K8N-DL primarily for the cost-savings they offer over higher-end boards. I am using 4U rackmount chasis', so form-factor is not a concern. I've also been looking at the MSI Master3-FA4R because it has 12 x DDR slots. However, I have read reports about boards shipped with innoperative DDR slots. _______________________________________________ Beowulf mailing list, Beowulf@beowulf.org To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
