Looks like it's by no means stable. Thermal cycling is hard on the works.

72 chips across 2 boards means 36 chips per board. Based off what I've seen, 14/16 likes the 0.4-0.5V range, and there's no good way to stack 36 chips into 12V to get this. I wouldn't be surprised if it was a regulated string of 2x18 chips per board with a dual-phase buck in the 8V 25A-per-phase neighborhood. If that's the case, undervolting might be possible - either directly through the config like the Avalon4, or by a hack like S7.

Bathtub failure curve so as long as you're say a month in, you've likely covered most of the initial failures.
[/quote]

You'd like to think that, but I just had a board go down last week on a B1 machine and another two weeks before on a machine from August. The other two dead boards were within the first month.
How many of Phil's miners waited a while before kicking it?

Honestly, I might have to say the ASICMiner Tube. I had a buttload of them (about 30?) that got run in 110F down to about 40F temperatures and I dropped exactly 6 chips. The VRM driver fried, which was an easy replacement. The S1 and S3 were pretty solid, but if we're looking at something with twice the power density those Tubes were tanks.
That's physical reliability, of course. They had their own problems what with the botched stratum implementation on the cheesy controller bricks, but we had workarounds for most of that already figured out before most of North America had received theirs yet.

I've had very few problems with the fleet of Avalon6 miners that have come through. They're more sensitive to heat than S7 is, but in my experience they fail slightly less often. S7 are too power-dense for my tastes, but they do undervolt nicely which, by not running boards as hard, should extend the functional lifetime and reduce overall failures.

If I can build a miner that's as reliable as an AntMiner S3 or ASICMiner Tube, I'll be happy with that.

I've never liked 'em, at least not the rack units. Ran too loud and too hot and always had parts failing. SP20s were a bit better, but I still saw a lot of them drop loops or whole boards, and some arrived with innards so bounced around (they weren't designed well for shipping) the ribbon cables were getting cut. Everything Spondoolies made was unnecessarily complex, which made way too much room for failures, and the fact their chips required 200A 4-phase bucks that ran like 80% efficient didn't really help much either. That's been my opinion since the first Rockerbox machines rolled out.

Why would anyone buy at those prices, especially if it's apparently so easy to mine that a rig gets paid off quickly?

Most S7 won't get down to 0.2J/GH; best you could reliably hope for is around 0.24J/GH which would still get you 4TH at <1000W. Some will do better, but that setting would pretty much work across the board.

But you can't mine 0.003BTC per day with it. You mine something else and then exchange it for BTC. Good in that BTC gets used (the more its value is based on currency utility versus asset investment, the better) but in the end someone's paying an inflated price for a coin that stands a much better chance of being worthless in six months than just mining BTC. Some call it gambling, some call it theft.


Looks like they'll be run like Avalon4 and Avalon6 were, where multiple modules are chained together to a single RPi controller. The single-point failure part isn't so great, but since it's an off-the-shelf part replacement is super easy. Setup and configuration are also super easy because you don't have to track down IPs for every individual machine. Overall a pretty good setup.

Daily profit of an S5 is pretty much dependent on the electricity being free. Sounds like you need to do some reading about what exactly goes into bitcoin mining before spending any money.

To be honest, I have a bit of unfair advantage with PSU cost since I build PSU hardware, but shelf price for what I build is $80 per S7. I see about half a dozen S7 on the front page of this forum's hardware sales board for under $300.

Most altcoins exist for people who only care about short-term profits. People caring only about short-term profits were the principal cause behind pretty much every depression, recession and financial crisis in history. Knowing this, some of us would rather be in it for the long game. Sure there are a few coins actually built for long-term viability, but that's only realized after they've survived being abused by the flock of pump-and-dump circlejerkers.

It's also in less than half the time, and during a lot cooler period overall so it should have been easier on the works.

There's no way I'd build a farm around Spondoolies, especially if I lived somewhere warm. They never ran at listed speed, they're incredibly complex and failure-prone especially because the cooling design isn't that great. I don't think I had a single SP10 or SP3x come through my hosting that didn't break down in some way.

I don't know what shipping is like to Venezuela, but I could get S7 in the US for $350 or less including PSU. That'd get you about the same hashrate from half the power, 1/4 the weight and overall more reliable.

S9 are a reliability nightmare especially if they're running hot, on top of being overpriced.

I believe you are confused. The pod and the S1 refit are, and pretty much always have been, separate designs.

No idea, and no idea. I haven't actually gotten any power curves from Bitfury. If the engineering samples from last winter are representative of production chip efficiency, it'll beat the Avalon7. By price, no guarantees.

Unscrew the old boards from the heatsinks, screw new boards onto the heatsinks. The old controller won't be necessary, the new boards run off USB. Fans will jack directly into the boards.

900W is easy to pull off a 120VAC outlet too, which is nice.

If I can stay in stock of boards and Phil does a group buy, I'll do 8-cable DPS1200 kits for $75 for group buy members who need a PSU.

I've had at least 20 Avalon6 come through hosting (most are still here), with two bad boards total. In about half as much time I've had around 25 S9 come through hosting with 4 bad boards total. I'm honestly surprised it's not higher than that. They do cost more, but there's something to be said about reliability (and, you know, customer service when something does go wrong).

S7 pulls approximately 1.4KW so for a 240V setup (you definitely don't want 120V, too troublesome for high-current loads) you're looking at 6A per miner. Four miners will run fine off a 240V 30A circuit.
If you're wanting to divide that up, I'd recommend a pair of 20A circuits because that gives you room for expansion (20A can support 3 miners).

I recommend server PSUs. For the record, I sell server PSUs - I don't recommend them because I sell them, I sell them because they're generally a more reliable and cost-effective option than consumer ATX supplies. For one miner a good DPS1200 kit will work; if you want more juice, talk to Finksy or Optimizer for heavier-duty supplies that can run two or three miners together.
Going with a server supply kit usually means a bit more hands-on, but it's also a lot more modular so if something does go wrong it's cheaper and easier to maintain than replacing the entire thing like with a consumer PSU.
Bitmain also makes decent supplies; the 1600W APW3 is okay, but I have no experience with the newer bigger one they released a few months ago.

240V has a buttload of different outlets even for the same general power requirement. Depending on the PSU, you're looking at either a C13 or C19 end (both are very standard, very common). Make sure to use outlets for which you can find cables to fit the PSUs.

I'm happy to see another <1KW miner.