https://bitcointalk.org/index.php?topic=1764803.msg17678108#msg17678108

Picture and explanation of where to measure. You should see 1.24-1.26V across the two chips, with 620-630mV per chip. If that's the case, there's an IO line broken somewhere. If there's an imbalance causing the string to lock up, your Node 0 voltage will be about 480mV. If that's the case, turn up the voltage a bit, unplug, wait about 20 seconds and plug it back in.

Once I have samples of the new PCB in hand and they're tested to work, I may do a "limited first run" sale of them, maybe a dozen or so, to put toward a full batch.

Good thing about the Terminus design is, about 80% of the components transfer over from the 2Pac. It's basically if you wired 4 2Pacs together, the only thing you'd need to replace is the main Vcore regulator. So I already have enough of most things to get started at least. The biggest hindrance, as always, is chips, but people have turned out with S5 boards lately that put me well over the top for the 2Pac batch and, if the really big one finally works out, I'd have enough chips for several hundred Terminus already. Since 2Pac manufacture delays were traced pretty much directly to ASIC availability and reliability of installation, both of which are now solved, Terminus manufacture should go much much faster overall.

Once Bitfury 16nm stuff starts talking enough for a Compac, it'll be almost trivial to port over to a Terminus as well. The most complex things will be communication and power; power is being solved right now, and communication will be a direct copy from the stickminer.

If the Terminus prototype can be made to run 300MHz (132GH) stable and reliable off VH's new freq-step code, I'm calling it good. Though I should probably run it in a quiet environment and see if the fans are annoying. That's something people care about, right?

I'd try turning up the voltage a bit. If you are able, measure the total voltage and also the voltage at each chip when it's plugged in. "0 chips" can be caused by a couple things, but the most likely is a voltage imbalance at the chips causing one to drop out which locks up the whole thing. If upping the voltage doesn't do the trick, it might be a data line issue.

If you up the voltage and it still doesn't detect chips, let me know and I'll see about fixing it.

It's these kinds of things that we check for while doing burn-in tests. You might not believe me but I have a box with probably about 70 sticks in it that passed initial tests but failed burn-in for various issues ranging from too many HW errors to data lines dropping out.

I was thinking more like metalwork, but hey you works with what you gots and my background is in construction.

Case anyone cares, 250MHz means 110GH. http://www.kano.is/worker.php?a=1BURGERAXHH6Yi6LRybRJK7ybEm5m5HwTr_terminus2

Hopefully I can do some more tweaking tomorrow and get it up to full speed.

As soon as I had a working Compac, I built like ten of them and took them home to run overnight where I could observe performance. It was fun having that many flashy lights in the room all night.

I meant that too though, that the flashy light part was the fun part. Power was fairly straightforward (when I finally got it working, the production version was by no means the first draft) and chip comms were very straightforward, but the flashy lights part was fun. It's actually got a double NFET inverter triggered off the RX line, where the second FET has a parallel RC on the gate to hold it high longer and that's what kicks the R and B elements in the RGB LED to make the white-ish flash. Without that the flash would be imperceptibly fast since a share return is like 5 bytes at 115kbaud. So it's not complex at all, but I like to think it's at least clever.

The Terminus uses a similar circuit, but instead of the RGB LED it's got a discrete blue LED and discrete white LEDs that flash. That way there's no weirdness going on with white-balance. Keeps it consistent. And also dead sexy.

Additionally, to stay on the subject of selling 2Pacs and whatever, I shipped out more orders today. We're operating out of standing stock right now so orders are shipped "immediately". I put that in quotes because I only actually send out packages on Tuesday and Friday.

I'm working, but dev time keeps getting kicked by manufacturing time. I started the 2Pac project to raise funds for BF16 dev but ended up taking over a month longer than planned to get the first batch built. Now all the kinks are worked out and I'm flat flying, but still haven't had time to get BF16 anything working. Way way behind schedule with new product development, but it's getting there. A lot of groundwork is laid, so as soon as we get chips talking the rest will fall into place pretty fast.

But that's not really this thread's project. I don't know what's up.

Further tests are required to probe the limits and - more specifically - the reasons for those limits, but my version-2 Terminus is currently on Kano.is pushing 250MHz. I've touched 275MHz on it, where the hacked model can see 325MHz from the same voltage, so there's still some limitation in the regulator that I need to play with. VH hooked me up with some improved stepping code that smooths things out a bit, and I did some jimmying on the hardware itself.

Oh and I should probably test the overtemp protection too one of these days.

If I can get this guy stable and reliable at 300-325MHz, I'll send off for what I hope is the final version PCB. That one will have mounting holes for if someone conjures up a case, means for better potentiometer access, stuff like that. Some of that's gonna have to wait a bit though because manufacturing and also manufacturing.

However - right now, I'm very pleased.

Not sure if it's my crappy upstream provider, but I've been unable to connect to anything ckpool, even the website, for a couple days now. I had been burning in freshly-manufactured stickminers on solo.

It can be done.

Truth be told, designing the blinky light circuit was my favorite part.

Those 49-port hubs are pretty decent. Klintay hooked me up with one for free in exchange for a positive review/endorsement. Normally I would refuse a deal like that, but it worked out in this case because I actually do really like the thing and would endorse it regardless. All but a few hundred of the ~3000 Compacs and every of the ~1300 2Pacs were tested on it, usually drawing about 4W per stick and up to 30 sticks at a time, with no problems.


Also, I think when Windows versions older than 8 are no longer allowed to work will be the last day I use Windows. Microsoft has been going in an increasingly irritating direction with their software since about 2002. Point of fact, pretty much everyone has been.

Hey, at least cgminer doesn't have a cartoonish GUI and burns multiple gigabytes of RAM for no apparent reason. Gotta love CLI.

Additionally, glad to see you got everything working out.

But I wouldn't have sold to you.

Mac paid upfront for the whole order, and paid shipping to Germany, and any import taxes, and it looks like he's also providing a nifty padded case for them, plus that sales price includes shipping.

Are you typing in zero-two, or oh-two? Because that's not a zero it's a capital "o". An invalid flag could cause the compiler to error out like that.

My laptop runs XP (custom-upgraded Thinkpad T61). I only downgraded to 7 on my workbench machine a couple years ago or increased RAM support since, you know, everything is bloatware. I'd consider Linux on it but my experiences with massive multiscreen aren't great and this thing has six heads. I still favor Win2K, it had exactly zero cartoonish elements to it, but XP with minor tweaks is pretty much as good.

Especial good luck if you're using Windows 10, because it's probably the worst possible OS. Not just for mining, I mean, but like in general. For anything. The only tech support I'll ever give for a Win10 application is "try again with a better OS".

As of right now, ALL ORDERS have been shipped and I am completely caught up.

I'm not sure that's ever happened with stickminers before.

They've been delivered. Scheduling is out of my hands now.

Yep, and all current orders will be shipped out tomorrow. Starting from now, I should be able to meet most orders from standing stock with minimal delays.

I know. You told me five times over a weekend and then paid the wrong amount. Next time BE PATIENT.

I'm smack in the middle of the US. As much as I like whole miners, distance and border crossings are gonna make weight really hurt so yeah I'm fine with just hashboards. Those are the things I'd get immediate use from anyway, and it's not like I don't already have 40-50 S-series cores laying around already.

Get a count, drop me a line and we can start talking numbers.