....but while two of the PSU's are cheapo ones, but I've cut all the molex connectors and directly wired them in. The wires don't even get warm. Maybe its a conflict, but why do they restart all the time? All 3 blades do it, and they are each on different PSU's... hmm.
I messed around with PC power supplies, found they had inconsistent behavior and were way too complex. Here is the solution, $27 delivered, gives 30A clean 12V, no other voltages, no hassles. The screw down block even has 3 screws for + and 3 screws for ground (labeled V-).
You get these babies, some 16 or 14 gauge stranded, use one power wire and one ground wire per blade.
I have three of these running 9 blades.
http://www.ebay.com/itm/261198916470?ssPageName=STRK:MEWNX:IT&_trksid=p3984.m1497.l2649By the way I've got a lab power supply that tells me the draw from these blades at LOW speed is 4.5-5.5 A, less than what's been claimed...
Spendulus, that is a great link!
I am definitely going to pick one up, However it will take several days to get to me, and we know what happens in a week in bitcoin! arg!
These units will definitely clean up the power, But I feel pretty confident some of my problems are not power related, I just brought in a new power supply, Its a 700W out of my triple GPU rig. Its a great unit and I plugged it into my 3rd blade (at 1.05V stock) and the blade is doing the exact same as the other two. It hashes at 50% speed on the pool. But shows no hashrate or accepted shares on the blade config page. Is that why it restarts? no accepted shares being reported back?
What the heck could it be?!
Yeah all this chatter about hooking up pc power supplies is a pursuit of the damned, for additional pain and misery. Just get real power supplies and be done with it.
I do not know how these units were designed, but let's assume for a sec they have a thermal cutoff. Then some of the sections may shut down if they get too hot. Inconsistent fan on your three blades could cause that. Laying them flat on a surface with no airflow to one of the sides could do that. Too close together, etc.
To determine if it is a hardware problem, swap the little board with the RJ45 connector between known good and suspect. Then swap the power board. Then swap the RJ45 cables, which can definitely go bad.
If you move a part around and the problem follows the part, you have a bad part. The only exception would be special handling of that part's IP address by the router. Obviously each of the blades needs a discrete IP and it must be in the addressing range of the router. They all came as 192.168.1.254, iirc.
I set mine to 192.168.1.200/201/202...
Note there is a reset switch between two pins, it reset back to default IP.