I've got an idea. What if I set up two addresses, one for sales and one for loans. Once I have prices set on whatever gets made and can open up sales, any orders made would be paid into the sales address and you'd get stuff as soon as it's ready. Loans can be made anonymously to the loan address, and as sales money comes in past what's needed to buy materials, it'll pay back transactions from the loan address - same amount returned to the sending address (or a different address if you can prove it's your transaction).

Looks like I already have a good source for chips pretty much donated (cost of shipping only), so that removes a lot of supply chain headache.

I could also point to that transaction activity as evidence of community support for even better products if Bitfury needs more convincing.

I wonder if the autotune firmware would freak out if you pulled the PWM line and ran the fans at 100%?

Hosting is a flat 8.9 cents per kilowatt-hour, or about $65 per kilowatt per month. No hidden fees, just power. I don't require prepays or multi-month contracts. Bills go out the first of the month for the previous month's use and must be paid within 7 days to avoid late fees; after 14 days your miners are turned off. I take BTC and PayPal.

DPS1200 PSUs can be cranked down to a low voltage and the 54-chip board underclocked to get something like 4.1TH at around 900W, or cranked up like he has it to get over 5TH at still better efficiency than a stock 4.7TH 135-chip S7.

Maybe louder than the quietest airplane ever made. The S3 is probably the quietest miner yet built, with the possible exceptions of what, New R-Box and maybe Avalon4 in certain configurations? Leastways it's the quietest of the S-series and can be made almost silent.

I like the 15pac better than the 4, for sure, but I also like the 2. There are enough stickminer enthusiasts to make it worthwhile.

I'll probably start with the 2 because I know I can make the hardware work. Technically I've already done it just while testing various things last year, I just don't want to promise something will work until I've given the driver guy a chance to figure out if he can make it work.

And even if I do decide to jump on a batch, there's gonna need to be money upfront for materials so I'll have to either let orders stack up or find someone to spot some to be paid back out of the first sales, and the rest of the sales (and sales of other stuff, I make PSU kits and stuff too) would go to keeping the lights on and shoring up the Bitfury project until that gets rolling.

So, right now we're waiting on the coder to let me know what he thinks he can do, and I guess for seed money to start rolling on materials?

Only if physics is broken, or that's the world's most inefficient power supply for some reason. The DC power consumed by the device will be the same regardless what the input voltage to the DC power supply is; the only way the power consumed would change is if the PSU runs at a different efficiency at a different input voltage, and that's rarely more than about 2% difference.

One S7 using 2600W would burst into flames in a matter of seconds.

PCIe 6-pin spec only requires two wires for power; 8-pin requires all 3. If you're only using 18AWG wire (which most consumer PSUs do), 150W is a practical upperbound for a 3-hots 6-pin.

I don't know if I'll do a 4-chip right now. It's kind of a tossup; I'd prefer to focus dev on the 2-chip stick, and even that is going to depend on what the coder can do for driver support. I mean I could build a 2-chip stick right now that would run just fine on cgminer, but last I checked both chips got the same work so half the shares were repeats.

If I do a 4-chip, I was figuring on it being more of a quiet desk miner kind of thing (Jalapeno-type product as BaldToasty mentioned, or the U3 is a better comparison) so USB-controlled but not USB-powered given the chips can draw 10W each plus buck losses.

Okay, so here's something. I really want to do the 2-chip Compac. If I shoot for a small standalone (U3 type) thing, what power range do folks think is best? 40W? 60W? 80W? If I do something like a 3x5 of chips clocked at 200MHz and 3.3V to the string you'd see 165GH and it'd run off a 12V5A brick. Bottom clock of 120GH ~40W which puts it close to Avalon6 and 135-chip S7 for efficiency. This is the kind of thing that would have sold really well a year ago.

So what if I made a 2-chip Compac at $35 and a 15-chip pod at $60? Those are just ballpark prices, especially the pod since I'll have to find a heatsink and fan for it. If it's possible, they'd run off cgminer and have hardware-adjustable voltage just like the Compac. Nothing fancy at all, I know.

The regulator will be able to safely handle at least 12V. There's no reason to build a device capable of drawing 50W and make it run on 5V only, that's foolish. Biggest problem with 5V power will be the fan.

Sure, if you want to rig it up to a USB port for power but then you couldn't super-overclock it to 50W if you wanted. Even a moderate hashrate of, say, 60GH would want 5A at 5V.

Hey toptek, I've got an internal control board from an Avalon6. Pretty sure it works but I'll hook it up and find out if you want it.

Already talked to vh and I'm checking out some other resources, so hopefully we can find out pretty readily how difficult it'll be to talk to multiple chips.

I'll have to do some digging, but I might also have a handful of original Compacs up for grabs next week.

DPOT adds a layer of complexity I would rather avoid for a quick-turn project, since that means tying a microcontroller into the signal path when a CP2102 would handle all the chip comms.

So you want me to design the board to fit that form-factor? Or just that the general design is a good idea? Because I don't really feel like designing for an existing form-factor, or I'd probably pick something like the U3 instead.

Sorry to be terse, but this is just another incredibly busy day in a week of probably minimum 12-hour days. Those BFL giblets are a lot fancier than I care to make for a short-batch fundraiser.

I've got some BFL cubes. Don't really care.

I was thinking of just finding one of those cheap 4x4cm heatsinks with a built-in fan and tucking four chips underneath it. Probably just in case, put a temp sensor under there that kicks off the power if it reaches a factory-preset threshold. The whole thing would be fairly small, maybe 8x8cm to have room for power and node-level stuff. If anyone wants it, anyways.

I'm waiting to hear back from the codemaster about how hard it'd be to extend the Compac driver to handle multiple chips. I don't think it should take much, but then I don't have any data to back that up.

Heck if I know what BFG supports. Never really used it. And whether or not it's compatible with current BF will depend on how you interface to the chips - they don't exactly use a standard protocol. How does BFG talk to the 28nm, and how do you know that because I've never seen any of their 28nm in the wild?

I'm not sure that question has an answer. The original Compac didn't have a "target" hashrate, that's the whole point of making it with adjustable voltage and frequency. Set it where you want it.

I can say that for the ~10W it took the one-chip to generate 20GH, this one should get closer to 30. The 5W hashrate, instead of around 13GH, should be more like 18.

Hadn't really thought about it. I guess Compac SE is pretty boring compared to 2pac.