What might be a consideration is that the design is robust and although rated for less on the 12v line it can provide more anyway.
If so you'd be moving the switching supply design very far away from switching supplies usual comfort zone of 50% full load. Expensive switching supplies can maintain efficiency across much of its range.
Still, switching supplies, even out of their 50% comfort zone, are far ahead more efficient than non-switching power supplies.
Wasn't saying that correctly. A 7812 will take an input voltage of at least 15 volts. To get the 15vdc you want to take peak, (15v)(1.414)=21.21v; then for effective, no dropouts, take .86 of that (21.21v)(.86)=18.24vdc.That 18.24vdc will be after a bridge rectifier which is going to drop well over a volt under heavy current. Then you want a step down transformer that will drop the 120vac down to 20vac. At the regulator say it's 30 amps, 18v-12v=6v; (30a)(6v)=180watts dropped at the regulator. Then the rectifier, lets say that's dropping 1½v, will be 45 watts. Then there's the transformer inefficiency but even without its losses you're dropping 225watts for 12vdc@30amps.
So if our C1 is using 818 watts at the wall using an 88% efficient switching supply, our C1 is using 720 watts which is 60 amps at 12vdc. Our non-switching supply example above would be dropping 450watts at 60 amps before transformer losses or 1170watts.
So I've been using 360 watt supplies, one on each blade for S3's, s3+'s, C1's and S5, no matter works on all. The supplies are modified with an added cooling sink to the bridge rectifier, a line filter, a 25 turn pot instead of a 1 turn pot and a digital readout. This have typically a 14 gauge conductor pair to each PCIE-6. I adjust to 12vdc. I've been changing the voltmeter to a pair of leads run to a PCIE-6 connector instead of right off the supply. On a C1 the difference is .2 vdc on each of the 4 supplies so those had been running at 11.80vdc, same with the other miners more or less. The difference being that when having both the first voltmeter and the second measuring at the PCIE, there was an approximate .2vdc difference between the two readouts.
I put a voltmeter on the Poweredge 750 watt. It's stock voltage is 12.15vdc. I calibrate all my voltmeters to a personal standard. On a problem S3 which was running at 12vdc measured at a PCIE on each blade, I replaced the two 360watt supplies with the Poweredge. I was sure I'd see less a drop in voltage under full load, the 360 watt dropping .2 vdc each. Surprise! The 750 watt Poweredge dropped from 12.15 to 11.8 under full load measured at the PCIE's. That's a drop of .35vdc. Even if I were to go in and adjust the Poweredge internally to 12.35vdc no load, that still shows a greater drop than the pair of 360watt supplies. So, I have to assess the pair of 360watt supplies are stiffer than the single Poweredge.
Wait, I just get I'm comparing apples to oranges. The .2vdc is the voltage drop due to the 14 gauge wire length while the .35vdc is the drop from no load to full load. I'll have to fire up the two supplies I replaced to get a no load/full load difference. But not tonight.
--------------
Falling hashrate problem with a C1. It has been running at intentionally lower frequency to baby it. Should be running at 960GH/
[email protected] and after a cold start it does but drops to 850GH/s after a few hours. Up last night I restarted due to 850GH/s hashrate.
This has the digital voltage readout measured at the PCIE connector rather than at the supply. When the miner first fires up, before hashing, before fans run, pump on another circuit, the voltage at the PCIE measures 12.15v-12.17v and after hashing starts it's down to 12.00v so the inexpensive supplies drop ~.15v from no load to full load and there's the .2v drop on the lines so these supplies drop a full .35vdc same as the Poweredge.
With the falling hashrate, when it drops, as each hashing board has a digital voltage readout at one of the two the PCIE, I see the voltage at 12.17v at one of the four hashing boards indicating it's not drawing much current. Tells me exactly which board it being troublesome.
I made a mistake building these leads. Besides the PCIE plugs having a mix of heavier conductors on some and cheaper wire on others, after splicing the 14 gauge stranded to the PCIE wires, which I mistakenly cut overly close to the PCIE, I used heat shrink tubing on the splices and a larger dia. shrink tubing over all the splices. This made the wires stiff near the PCIE. I then put the miner on a cart but the wires came too close to the upper shelf and bending the leads flat put pressure on the PCIE board connections. This is where I hope I'll find the problem. If this isn't it then it's an ASIC that heats and accumulates errors (an internal board) but even when the hashboard isn't drawing much current while the other three are, there are no x's in any chain.
