CrazyGuy's Minera RPi image is the way to go with the AM Prisma.  Especially if you have more than 1 unit.  I had very little issues with a single Prisma on a BE Controller used in conjunction with BFG proxy.  As soon as I added my 2nd Prisma - I had nothing but issues. 2 units wouldn't hash for more than 1hr - 1hr15min, had a hashboard that refused to hash 90% of the time.  The only way I could get 2 units to hash half way reliably was to run the 2nd unit with 3 hashboards - the other hashboard would cause the BE Controller to "freakout". Even with running 7 hashboards - I still experienced wide variations in the poolside hashrate.

This morning I moved the 7 board setup to CG's Minera RPi. Started hashing great; poolside hashrate very solid & consistent (still varies a bit but nowhere like it did on the BE Controller).  Efficiency/Reject rate has dropped from 6.25% + to less than 4.15%.

When I got home this evening I connected the 8th hashboard, after a little over 2.5hrs the unit stopped hashing - 1 indicator poolside was a drop in work difficulty (same reflected by cgminer and when on BE Controller); had seen this before with the BE Controller - the controller & Minera both still showed normal poolside hashrate - but the pool showed the unit dropping down to 0 hashrate. But this time Minera Acepted/Rejected/Errors graph showed the drop also (not sure why it still reflected the poolside hashrate).

What I did to get around this on the BE Controller was to use a pool port set for difficutly 2048 (for S4), doing this seemed to correct this issue with BE Controller, and now also with the Minera RPi.  It's like the Prisma doesn't handle vardif very well when more than 1 unit.

Since changing to a fixed difficulty the 8 boards have been hashing for over 4hr40min






http://i.imgur.com/wN1PfEr.png