Here is some performance data over many days for my 3 A6's:
1) Power supplies: Antec HCP Platinum 1300W Power Supply (runs either 120 or 140 VAC).
2) Antec outputs (DVM): 12.3 to 12.4 volts at load. These are plugged into 240 VAC.
3) Hash rate: Two units 3600+, one unit 3900.
4) Each A6 has it's own RPI.
5) All units have the latest MM and RPI firmware.
6) Each Antec has 4 12V power supply buses. Each PCI connector is hooked up to a different bus.
7) They are all hooked up to kano.is.
The A6's display 0.1 to 0.2 volts lower reading than a DVM power supply reading.
yeah out of my curious mind if you flip the two antec's does the hash rate flip?
I had this same issue with my two avalon's and two 1200 watt seasonic platinums.
I end up with ½ of each seasonic on each side of the two avalons
Don't do this if you do it wrong you can fry both psu's and both avalon'sonce I did this the avalon hash rates balanced out.
This was why I ended up with the 240 volt server psu's
Phil,
I tried your suggestion earlier and maybe, there was a 50Ghz difference.
I also tried a unique feature that the Antec's have called:
OC Link™ - Allows 2 HCP Platinums to work in tandem to power the most demanding systems. This cable hooks up to some sort of special com port and allows two HCP Platinum 1300's to function as one 2600W supply. I didn't see any change in the collective hash rate.
What I did see a change in was giving each A6 there own RPI. I had two chained together and this added about 100 or 200 Ghs to the lowest performing A6 when the low performer was on it's own. This is consistent with the earlier observation in this thread that there is some sort of compromise when operating an A6 chain. I think what the farms probably do is have a standard setup to eval incoming individual A6's and try to match A6's with similar performance characteristics and put the similar one's together on the same chain.
64