He asks, literally a year and a day since the last post in this thread.

That's pretty much what they've always been for, learning tools and lottery tickets. I've never disguised the fact these will never pay for themselves - outside of the improbability of a solo block.

Not likely, unless he compiled it with a filesystem and ran a bitcoin wallet off it.

Bitcoin miners never actually see the bitcoins they generate. They just process transactions - and most of the time they don't actually see those transactions, just a compressed version of a bundle of transactions already being processed. A transaction assigns bitcoins to a particular address. The only way to access those bitcoins is to have the private key corresponding to the address, which will be stored in a file associated with your wallet software on a computer, completely separate from any mining being done.

Interestingly enough, the first search result for "pod" is this thread.

Solid white light usually means there's a problem. Compile the proper cgminer as instructed in the 1st post and if it detects the device but doesn't mine with it, you might have to contact the seller for warranty replacement.


Yeah, in the "news on pod miners" thread. This is the stickminer support thread.

Oh yeah, I forgot about that guy. I believe it was scrypt, not X11. Yeah that and gridseeds were dual-algorithm chips. Good call.

Honestly, I didn't really like that chip anyways. Big CPU-style BGA really makes a lot of unnecessary headache for integrators. Spondoolies was really the only group to do it well, exactly one time, and then went back to array of small chips for the SP50.

But that's a whole different argument.

Ah, that would be why you're talking about 60db fans instead of about 85db.

High pressure is required to move adequate air volume against high resistance. The ridiculous solidity of the cross section of an Antminer presents a lot of resistance. High volume is required to move the ridiculous amount of heat from the ASICs running 100% power, otherwise the die temperatures hot enough to boil water would start getting more like hot enough to melt solder, or start glowing and burst into flames.

So yeah, high CFM at high static pressure is sorta essential to the process.

If you wanted to do it with an ASIC, it'd be a matter of building both cores into the same chip and then integrating a controller that could switch between the two. It's possible, but nobody's done it.

Then there's the FPGA option, where you could program the chip with one type of core and after a while reprogram it with a different core. But then you'd have to build into your miner whatever hardware was required for rewriting the FPGA, and FPGAs are probably going to be more expensive and less efficient.

It's technically possible, but if the major manufacturers can get more money from doing less work (selling simple special-purpose machines rather than more complex general-purpose machines, you buy two miners instead of one) there's no real incentive.

Dang, really? I figured ckpool would be a big hit.

Well, that was useless.

Some recent good news - I have acquired and will soon have set up and running a CNC drill for working up heatsinks. I have another pick-and-place robot on the way, and made some design and process changes to streamline manufacture. This week I also got some additional part-time help. The end result should be manufacture won't be kicking my butt as hard as it has for the last few months.

Which means I should get some more R&D time and get back to ongoing projects. Specifically the Terminus, the USB hub, and most excitingly Bitfury projects.

I'd really like to have a bitfury miner out in time for Christmas. Worst case, the Terminus BM1384 pod is in mass production by then. It's not nearly as good as a BF pod but still better than nothing.

If you're thinking of dropping the clock anyway, you should look into hacking the voltage as well and drop the power draw quite a bit more. On most 135-chip S7s it's possible to get around 4TH around 900-1000W.

Of course then your shipping fees would be all over the place, might end up costing more than customs. Who would you rather give money to, a foreign corporation or your own government?

There's tricks with running two instances of cgminer with a different frequency for each one and limiting which sticks each instance enumerates.

That's weird. Stock setting is about 1.26 and they're tested, twice, to run stable at 100MHz at that voltage.

Unless you're talking about the seconds I sold last month, I don't remember who bought them all. If that's the case, then yeah they would have gone out with voltages up to 1.40. That's why they were discounted.

In any case, there's no performance guaranteed past stock specs.

Nobody said that HW error rate is okay. When I test them, if they put up more than 2 errors per hour they don't get sold. With all those errors, I bet your effective hashrate is in the toilet.

You're running it at a higher than stock frequency. It therefore needs a higher core voltage to operate stably.

Yes, and how to do so is explained at least once in this thread.

There's also the consideration of turning up the core voltage a bit, if you have available power. That's also explained how to do.

Weird, it was never a problem before. Including four months ago when I started this thread. Had that account for the better part of a decade, rarely gets used, but my upload at the shop is low enough that if I host photos on my own server they don't load right on here half the time. Guess I'll go back to doing that. Grumble grumble free service grumble grumble.

The temp sensor on this miner is right in the center of the board, underneath the heatsink with the chips, so it should read more accurately than, say, the S5 temp sensor. The cutoff is also imiplemented in microcontroller firmware on the board, certainly more failure-proof than a check box on a cgi webconfig jumping through various loops to an OS interfacing with half a dozen hardware something-or-others.

Hopefully that image is visible.

So I've verified that all the hashing parts of the current version board still work, which is good because I moved basically everything around.

In the front left corner you can see holes for a 4-position header. That's where 5V comes out to power a Pi, and it'll be replaced with some kind of USB jack.

You can also see the empty pads for the 8-SOIC where the microcontroller would go if I hadn't bypassed it on this one. That guy would control fan speed and automatically power-cycle the chips if the string locks up or overheats.

Free space on the board is getting pretty tight so it'll be fun figuring out the last few tweaks. All the jacks and such are final positions, and the board is final size, so this is a decent representation of the form-factor. The future Bitfury pod will be mechanically compatible.

I'll be sending a few out to established resellers and my coders as samples, working up about a dozen more that'll probably be for sale on here in a few weeks. Whenever I have time, things are pretty backed up here.

But that's the present state of things.