TBH, I would not waste time on MB, system ram, or even PSU. (in terms if variables, there is nothing to tune, and little to gain, (assuming you have a stable rig right now)
The key point with PSU is, they are a big portion of your asset cost, so ensuring they run at optimal efficiency vs number of GPU is a quick and easy calculation, that's it, there is nothing else to tune there, no more variables to consider, (other than maintenance and mtbf)
Ok, I'm going to through my MB question. I started mining with whatever I had lying around. In my case, I have an AMD PHenom II X2 545 CPU with DDR2 ram and . I've got a PCIe x1 splitter to run out extra cards.
I've got 7 cards in total, split across two machines.
However, I would ideally run these on a single machine, so I have less CPU/RAM/HDD power overhead. But, the problem I have found is I can't run more than 5 cards on this machine, or the other one (AMD A4-4000, with DDR3 ram). The 6th card would detect but give that error-33 (32?). I swapped many cards around, tried cables, risers, rechecked mb bios settings, but I started to get the feeling that these old CPUs just don't have enough PCI lanes to run this many cards, or else there is a similar MB limit. So I ended up giving up on a single machine setup, and have the two running.
I looked at a mining motherboard, etc. but the saving in power overhead doesn't outweigh the cost of hardware for over a year. Which still makes these "free" motherboards (they were just lying around home) still appear attractive.
So, in summary, do these old boards/CPUs have a GPU limit?
(P.S. sorry to hijack the claymore thread for these kind of questions, but most people here actually seem to contribute constructively to questions!)