I saw 15 S7LN and didn't have a single bad board in 'em, but I did have three bad PSUs. One was a shot fan so I swapped it out but the other two were straight up gone. Even considering only electronics failures in PSUs that's still 2/15 for 13%

I'd rather see solid heatsinks on both sides even if one has to be milled for clearances. Bitmain would rather go the cheap (and demonstrably less reliable) route, which does nicely keep node-level power differences isolated so that's a benefit. Course I'd also rather see lower power density than 4x the heat per volume of the S3. It's great if you're limited on shelf space, but for most folks space is probably the cheapest part of the equation. It's not the super-powerful ASICs that's the problem (I think BM1382 generated more heat per chip), it's cramming 60 of them on a board the size of a paperback book.

I also expect PayPal will realize quickly there are issues with Bitmain and chargebacks. Right now their most failure-prone product through PayPal will probably be the S7LN PSU, which in my experience are lemons (20% failure/DOA) but maybe they already culled out all the garbage ones by now?

Also, I don't think going after BitmainWarranty (the Colorado-based entity) will do you any good. They're not actually legally associated with Bitmain last I checked, and also do their best to pick up the slack when Bitmain screws over US customers.

I believe that ferrite of some sort would be the inductor on the regulator supplying the node-level LDOs for IO voltages at the top of the string. No guarantees on that but it's my guess.

I would also guess they're more upset about the corner ASIC which was more firmly bonded to the heatsik than the PCB so when the heatsink was ripped off it took the chip with it and probably damaged pads (or "pins"?) on the PCB.

As a comp sci, oh man I hate ladder logic. Ladder logic sucks balls for complex tasks. But yeah for reacting to sensors it actually kinda makes sense.

Surely some of that would work, yeah. I think chigiuretor (or however he spells it, been doing business with him for like two years you'd think I'd remember by now) has been setting 'em up for remote turnon/off for a while. I wonder what he does.

But anyways none of those DC-side solutions would work if there were an issue with the mains feed, in which case you'd want control of an AC disconnect and that gets a bit more complex.

Or that, yeah. It's a completely different solution to a breakout board with an external turnon input pin. My boards take an active-high signal so that wouldn't work.

Guy in my hosting has some he's unloading for around 0.52BTC if you're in the US.

Especially if you use a good breakout board with an external turnon input pin.

I don't know about your main breakers, but mine would take a very sturdy servo to toggle 'em over. Kicking off isn't too difficult, but kicking back on has a fair bit of pressure on it.

What about using a normally-open relay driven off some networked GPIO? Someone on here was talking a month or two about making his own networked switchable PDU and I think he got laughed off, but that seems to me a lot less jankety than wiring a servo to a cellphone and attaching it to a breakerbox.

Appears like one of my hosted S9 has a dead board, right side looking from the intake. I think it had red light when I was looking over the machines because I couldn't immediately distinguish which board was dead, but when I took it down for testing the red LED never lit. It draws full power and the buck output is the right voltage but it doesn't register in the stats. Guess we'll see what the owner wants to do about attempting warranty.

Which, since bmminer is apparently closed-source, could be in there on purpose. In any case it makes finding and patching the hole a lot more difficult.

Have they ever explained anywhere what that PIN is? I would guess it's a MacGuffin they made up to have a universal justification for denying warranties in otherwise passable cases.

As long as it's compatible - pinout, signals, physical size etc it should be fine. The 1200W rating is the maximum power it'll output, not a fixed requirement that will blow up your miner with too much juice.

External voltage adjustment would probably have only minimal effect on overall efficiency. Perhaps increase the efficiency of the board-level bucks by one or two percent depending on particular implementation, but that'd be the extent of it. Unless the board-level buck output exceeds 11V typically in which case it'd be forced to operate at a lower voltage, but that seems unlikely.

If Phil is unable to replace his S9 board effectively and sends it to me, I'll look into fixing it at getting enough information from it to develop a voltage hack like the S7. I've had an idea for a while, but not been able to experiment with it yet, about a plug-in hardware solution that doesn't require disassembly of the miner.

I was also thinking about awesome lasers a bit, but I only know the classroom math there instead of actually playing with them on a daily basis; be assured I am darn jealous.

One KWh is 3.6 megajoules, right?

At the risk of seeming unnecessarily pendantic, but for the sake of correctness:


Watts are the amount of work; Joules are energy. Work is energy per unit time. The electric bill is actually based on energy; you are billed in accumulated kilowatt-hours (equivalent to 1 kilowatt draw [at 240VAC, about 4.17A] for one hour), which since watt = (joules / time), watt-hours = watts * time = (joules / time) * time = joules. I'm not entirely sure what to make of "the rate of use is Watts/time".


24 hours * 1.3KW = 31.2KWh -> there is no unit division, but unit multiplication. The proper unit is kilowatt-hours, not kilowatts per hour.

The ceiling on my 49-port hub is about 280GH, so however many sticks at whatever frequency to do that.

And you verified that both fans are spinning proper speed?

That post about a different machine from a different manufacturer solved your SP10 problem? I am skeptical.