Well, news. Not good news.
That first Terminus board I got working decided to give up the ghost.
I built a second one to test with and it never worked. So I built a third and it also never worked, exhibiting the exact same symptoms. Worked on both of 'em, couldn't get either to behave. So I went back to the first one, and suddenly IT started doing the same thing as well, after working just fine under extensive poking and prodding for about a week.
So, I don't know what's going on.
I'd previously asked Optimizer for advice, but now I've asked him if he wants to just design the buck himself. He'd expressed interest in the project's power integration in the past, and as that is his area of expertise, I'm hoping he'll be able to put together something good in not a lot of time and that the compensation I'm offering is acceptable.
If he's willing, that'll free me up to focus on a few other things in the meantime. Every part of the pod except the main regulator worked from the original design back around January, so ironing out that one piece has been a substantial delay that I no longer care to fiddle with. Maybe now I can work out some of the BF16 problems.
|
|
|
I have in stock right now a 24" cable that runs from the 8-pin CPU jack to 2x PCIe 6+2. I don't have any for GPU jacks right now but I can make one in about ten minutes if I know the length.
|
|
|
I'm currently stocking up on PSUs (they're ordered and inbound), but once they arrive I'll be able to sell a 1500W Platinum-rated PSU kit for $150 shipped within the US. That's got a breakout board and 10x 18" cables. One benefit to using a modular setup like that is if something breaks or dies you only have to replace that one piece.
|
|
|
Starting to think four a day's a lowball.
|
|
|
Maybe we should get a sticky-thread disclaimer for that, since we're getting like four GPU question threads a day now.
|
|
|
So 2Pac assembly is going so fast now that tomorrow I'll be out of a job again until I get more chips (don't worry, they're already on the way) so I should be able to spend more time on Terminus R&D.
Now my original plan was to build a pod miner in the 50-100W range, and the Terminus was going to be a half-scale pod in the 30-60W range. But my current prototype is already good for more than 60W. Thing is, the full-size BF16 pod was intended to have 11 chips (I know that's a crazy number but there are reasons) and the recommended clock for 'em is at about 6W each, so an 11-chip pod would push around 75-80W at the top anyway. Shoot, the Terminus can handle that heat already.
So here's what I think I'll do.
I think I'll continue the Terminus formfactor as the defacto pod miner, with an expected power range 30-80 watts. However, it'll take the Amita product name that I've always intended for a pod miner. If you know the source material, it makes more sense for the Compac and Amita to be small and the Terminus and TypeZero (the S1-5 refit boards) to be big anyways.
So I'll look into the Amita pod having 11 chips, and a Terminus with 22 chips in more of a New R-Box type thing running up to 150W or thereabouts. The TypeZero boards will have 33 chips per board, so 66 per miner.
I don't know about y'all but that makes sense to me. Anyways I'll be spending most of this week ironing out the remaining kinks with the BM1384 pod, whatever it wants to be called.
|
|
|
Nope. Eyeboot is out of Hong Kong, Valkir out of Canada, bitshopper.de out of Germany and then the regular US sellers is all I have. Nobody in Australia.
|
|
|
Did you have all the dependencies in the build? I think you were missing something, is it ncurses that causes it? Something like that anyways.
|
|
|
What I can pay will depend greatly on where you are, what with the cost of shipping. Let's take this to PM.
|
|
|
One of the things you've missed is that for the last 2 years I've been working toward being able to build new boards with current-gen chips to mount on S1, S3 and S5 chassis. Those models were before Bitmain decided to basically only support industrial customers who didn't really care as much about noise or, apparently, reliability.
Take some time to sniff around here, run a few keyword searches and you'll find good detailed discussion over most everything you're talking about. I'm not gonna get into it all here. If you look at threads I've started, or anything with GekkoScience in the title, you'll see the kinds of things I'm working on. And no I'm not doing chip design, to do anything competitive requires probably a team of experts and a seven-figure budget minimum.
I honestly don't really mind that my website is out of date. Means the only customers I have to bother with are the ones who try a little harder to find what they're looking for, which weeds out a lot of idiots. Not dealing with idiots all day means I have more time to do the actual job.
|
|
|
To tell the truth, the website hasn't been updated in about three years. Most things are discussed here on the forum. I've reasoned extensively about not using BM1385 or BM1387 for new projects and am actively working on <100W miners.
|
|
|
I use it to test sticks during manufacture, about 15 at a time, and it works fine. I tested Compacs 25 at a time on it.
Well, not the internal-PSU version, the older one, but if anything the new one should work better.
|
|
|
I'm gonna do some playing around with anodize one of these days.
Have you looked at any of my other projects?
|
|
|
Another round of what? Sales are still open. I've already shipped out about 200 batch-2 units.
|
|
|
(1) No there isn't
(2) Because I wanted it to be
|
|
|
Probably just a standard 4-pin header with 5V and ground, should be easy to make that work.
|
|
|
Couple more tweaks to the buck and I just hit 325MHz. It's running off a Pi right now, gonna take it home and see how it fares - see if the fan's annoying, if temps stay reasonable, whatever.
|
|
|
Your edit is pretty ambiguous. You mean you're using VH's Compac build linked in this thread, but to control original 1-chip Compacs?
|
|
|
I think there'll be room for it but no guarantees yet, but I'm gonna look into putting a small buck on there instead of a linear regulator for 5V needs (USB, LEDs etc). That would increase the cost slightly but would also increase efficiency slightly - and the real significance is that would increase the 5V current capacity without something catching on fire, so I could put a header on there that one could use to power a Pi without requiring a second adapter. Be handy for a more standalone setup, especially if your Pi had wireless, which I think Biodom among others has been pushing for.
|
|
|
No, I mean millivolts. The whole thing should have 1.24 to 1.26 volts across it. Make sure there aren't parts busted off around the buck controller in the bottom corner near the USB jack. Sounds like you're not getting power, so if there was damage in shipping it's probably the 220pF capacitor just north of the little trimpot. If your main regulator is working, the two sets of numbers I gave are pretty much the only ones possible (either 620mV or 480mV across node 0). Either that or you're measuring wrong, but the diagram is pretty straightforward.
|
|
|
|