GekkoScience BM1384 Project Development Discussion

[Unacceptable]

Would the thing I already mentioned about putting 5TH worth of Prisma-refit boards in a 3U case count? It would have forced-air cooling over conventional heatsinks. Or do you mean how the innards of the S4 are designed? Because we probably won't do that. We've got one hardware guy and one software guy and zero time to go designing a jillion different boards. If, however, we design two boards and they're both good they both can be used for any number of different form factors. Like a small board which can be run one or two at a time for a quiet desk miner or four strapped on an S1 chassis, and a larger board which can be mounted on a Prisma chassis or put several in a rack case to satisfy larger-scale people. There's no reason to bust our butts even further to make a jillion really application-specific designs when we can make one or two really good flexible designs and then let people do whatever the heck they want with 'em.

I will make a one-chip stick miner. I might make a two-chip stick miner. I will make a board sized at one half an S1 board. I might make a board sized at a Prisma board. I very likely will not make any other boards for any specific use cases. Those should be good enough to do just almost everything anyone actually wants.

That sounds good to me.  I actually like the idea of getting bare boards that I have to build into a useful form factor myself more than a totally out of the box ready miner.  Gives me something to play with...

And saves money on shipping (no case,no PSU=less weight),assembly by the factory (saves on labor costs & time)  

Keep it simple & be able to use off the shelf parts that are easily bought or created by us.That's a win win!!!!!!!!!!!!  

Rock on sidehack!!  

Oh BTW,scrypt is dead...unless the power consumption is MASSIVELY reduced 

[sidehack]

I'm quite confident if I'm stepping on any toes, he will gladly tell me.

I think we can all agree sidehack likes telling people they are wrong.
--
novak

If I call you out on what appears to be a boneheaded comment and it seems unfair, explain yourself and if I'm being a jerk unnecessarily I'll gladly apologize.

If, however, it's justified... well, you're on your own. I probably tend to expect more out of people than most folks do, so keep in mind before asking questions that it's pretty likely (given the amount of information we've tossed around so far) that your question has already been answered, and if it's already been answered I'll find it for you in a fairly blunt manner. I like being helpful but I don't like helping useless people, or people unwilling to try on their own. Curmudgeon, you know.

We're very much in favor of making solid miners that can save people money on purchase price and shipping. We're also in favor of making miners that require the owner to actually get his hands dirty to get it up and running. Closed-box solutions you just plug in and walk away from are boring as heck. That's why we're wanting to make even our stick miners hardware/software fiddlable, so even the folks that only get stickminers still have a decent learning platform that they're encouraged to actually interact with.

[Spotswood]

...

Closed-box solutions you just plug in and walk away from are boring as heck. That's why we're wanting to make even our stick miners hardware/software fiddlable, so even the folks that only get stickminers still have a decent learning platform that they're encouraged to actually interact with.

^Yeah, I really miss designing and building frames/cases to mount all of the various mining boards.  

[sidehack]

That's the whole reason I got started with bitcoin, was because the hardware was fun to play with. I laid down about $500 on an AM Blade back when and was replacing parts on it within two hours of getting it out of the box. I'm kinda appreciating boring hardware now with 50KW of hosting to manage, but that just makes time for other random stuff like resurrecting stuff in the museum and building some new projects.

[pekatete]

If I call you out on what appears to be a boneheaded comment and it seems unfair, explain yourself and if I'm being a jerk unnecessarily I'll gladly apologize.

......

Why would anyone have to explain themselves to (you) anyone? This is a public forum and anyone registered on it is free to express themselves, even ask the so called boneheaded questions, without (fear of) being subjected to your slurs. If you can not tolerate others' contribution(s) on the basis that its a thread you've started, then you should have opened a self moderated thread whence you could remove the offending post, barring which you'll have to grin and bear it! Keep your house rules to your house .... this is a public forum.

As a side note, interesting project re-inventing the bitmain wheel .....

[sidehack]

What I mean is, if I offend you and you have good reason to be offended, I will apologize because that's not nice. But if I offend you because you asked a question which has been answered already and you were too lazy to find that answer, or because you asked a question based on bad information or assumption and you were too lazy to find the correct information, I stand by my opinion. If I do the same, I fully expect someone to call me out on it. Sometimes politeness is nice, but I'd rather be corrected when something I think I know is wrong than let polite people allow me to persist in incorrectness. It's not helping anyone at all to stay silent and let wrong people stay wrong, nor is it helping anyone to offer information handouts to people too lazy to look for them first.

Additionally, there's plenty of room for improvement in Bitmain's current offerings. We're not reinventing anything. We're taking what they're doing and trying to make it better, and then make it affordably available to home users instead of shipping everything to rich guys so they can get even richer. Yes, it is an interesting project.

[pekatete]

You are set in your ways so I'd say a self moderated thread is the way for you to go ... but that's just my opinion.

Good luck with the improvements on bitmain gear .... you'll most likely need it!

EDIT: I doubt you can make it affordable in any reasonable quantities without bankrupting yourself though .... but I stand to be proven wrong. Flogging the single / duo chip sticks at ridiculous prices (though low in the grand scheme of things bitcoin rigs), is no where near any definition of affordable that I have chanced upon, but presents the only possible voluminous line for you.

[sidehack]

Yep, I'm pretty firmly entrenched in the opinion that helping people is a good thing, but helping people that don't try to help themselves is a waste of time (give a man a fish, etc etc). I'm pretty firmly entrenced in the opinion that truth and correctness are important things to keep track of. I considered setting up a self-moderated thread, but I don't want to delete other people's stuff. That's even worse than ignoring it, and again, it helps nobody.

Also, I'm hoping to have a good buck circuit designed for the stickminer tonight, and a simple PCB to test that circuit by Monday. For the Compac (one-chip) I'm giving it a voltage range of 0.55 to 0.75V and the whole thing should be good for 8A if you can keep it cool. The stock setting will probably be ~0.6V ~8GH but if you can keep it from catching on fire you should be able to get twice that out of it. I'll have a row of test pads at the top for measuring core voltages and such, make it easier for folks to calibrate. Undervolt lockout on the regulator at 4.75V, so it'll allow a 5% sag in your USB voltage before cutting out - which is basically a safety feature to make it harder to destroy whatever this stick is plugged into if you start cranking it way up.

The Amita (two-chip) will probably use the same regulator, but an output voltage range of 1.10-1.5V so each strung chip gets the same Vcore range as the Compac. I'm not sure of the stock setting yet, but it'll probably be set to draw around 3W. I know that's over USB spec, but how many folks have hubs capable of outputting more than 0.5A to a port? Additionally the increased output voltage will increase the power efficiency of the regulator slightly, but any efficiency improvement is a benefit if only for making cooling easier.

We're shooting for a maximum stickminer price of $20 for the Compac, which if you can find 8GH worth of stickminer for $20 anywhere else you're looking in places I haven't. I think the NF6 is the only other single-stick miner good for that hashrate, at three times the power, and can sometimes be found for $30 or more. $20 is expensive for the hashrate, but on a stickminer scale it's pretty good especially for new gear. If we can undercut current S5 pricing for a four-set of TypeZero boards, or even match it on a $/GH basis, will be nice. I think we can. The key is, our gear would be comparably power-efficient at stock settings and software-undervoltable meaning it'd have a lot longer viable life than the S5's fixed operating curve. An additional key is low profit margins. Bitmain already demonstrated they don't mind padding an additional hundred bucks of margin onto their machines, on top of what's already there. But for the last year we've done just about everything we could to keep prices low on everything we've sold, which includes keeping margins pretty low and, for the most part, not getting paid for the 60-odd hours we put in most weeks to get stuff taken care of. Given the choice between going bankrupt and competing on price, we'll opt to scrap the project entirely. We do not sell at a loss, and we avoid incurring debts wherever possible.

[pekatete]

Yep, I'm pretty firmly entrenched in the opinion that helping people is a good thing, but helping people that don't try to help themselves is a waste of time (give a man a fish, etc etc). I'm pretty firmly entrenced in the opinion that truth and correctness are important things to keep track of. I considered setting up a self-moderated thread, but I don't want to delete other people's stuff. That's even worse than ignoring it, and again, it helps nobody.

That is completely contradictory if not grossly convoluted (if ever there is such a thing). Surely there can not be a but when it comes to helping ... even if you throw in a fish or two!
Well well, I'll have to leave it at that and hope for the sake of your stick miner customers that your design improvements to the bitmain gear do not exhibit the same characteristics.

[sidehack]

Even assuming there's an unlimited number of fish (but limited time to distribute them), would you rather give them to a guy who is trying to fish but they're not biting, or give them to the guy sitting on the shore with a dry line and a cardboard sign saying "please give me fish"? I'd rather help a guy with half a solution find the other half than help a guy who isn't even looking by giving him the whole thing. He learns nothing for himself that way. I want to help people learn how to do things for themselves, and I'm not overly fond of handholding.

If folks would rather I ignore questions which have already been answered instead of frankly directing them to the answer, or would rather I ignore misinformation rather than frankly correcting it, I can do that.

Anyway. Semantic arguments and opinionated browbeating aside, I'm genuinely hoping that our design improvements are improvement enough to merit the time and resources invested. But please note that we're not taking in any money from anyone for this work until the designs have proven themselves. We're not asking for any help at all in any way. If it's not an improvement, you have lost nothing except a bit of time on this thread, and I will have had a heck of a lot of fun with design work regardless.

[pekatete]

Semantic arguments and opinionated browbeating ..... !?
You've comprehensively demonstrated over the last few posts that you do not mind repeating yourself, even when not prompted (in the real world I'd say you love the sound of your own voice, though in this case not too far off that). You can also now testify to your own contradictions by labelling them sematic arguments. So one, two even six fishes (if you want) in hand, rudeness is just that, not lack of politeness, and has no place in a public forum.

On the other hand, flogging your stick miner for $20 (includig a respectable margin) is nigh on the impossible, bearing in mind (if I'm to go by your assertions) that you'll be fabricating in the US. Of course, I'd like to be proven wrong, but I know you can't even if you fabricate in China, as you can not have the quantities to justify that.

[sidehack]

Yep, fabricating in the USA. If we can make it for less than $20 we'll probably drop the price. If we can't sell it for the price we're asking, we won't make 'em. Simple as that. The primary reason we're moving forward on making them is because plenty of people have stated they want them, and from the beginning we've said we were shooting for a maximum price of $20. I could make it fifty cents cheaper if I wanted to use a low-end fixed-voltage regulator or a tiny boring heatsink but the point is to make it an awesome learning tool instead of a boring toy. You gotta consider that the per-chip complexity is an order of magnitude greater than for a full miner, given that a full miner will have one VRM and one UART bridge and 18 chips, where this has one VRM and one UART bridge and one chip.

What do you consider a "respectable margin" on a product like this?

[novak@gekkoscience]

As a side note, interesting project re-inventing the bitmain wheel .....

I think maybe you didn't read any of the technical details.  No one has ever designed a miner power topology like this so I seriously doubt you could say we're reinventing anything.  That's not really true for the stickminer, but even there we're creating something that doesn't currently exist- a current gen stickminer with voltage adjust (has any stickminer actually had voltage adjust built in?).  Hopefully these will be good for home miners.  At any rate it should be a fun project.   

--
novak

[sidehack]

Sure enough. I stuck around the shop until midnight working on the Compac regulator and I think I got it finished. I'll probably tackle a PCB starting Sunday afternoon so it can go on Novak's etch batch on Monday and we can start testing it out as soon as parts come in. Novak can attest to how giddy I am working on all this stuff. We really are doing it because it's awesome and fun, not because we want to get rich.

Also, I'll work on being more polite - or at least more neutral. Less grouchy in any case.

[TheRealSteve]

If we can make it for less than $20 we'll probably drop the price.

That probably depends on how you're planning the run.  The following is more for others' info than you guys

There is an unfortunate gap between small batch (<100) and large batch electronics (1000+) that isn't currently being served well.
PCBs are certainly not the major cost - whether you go with an outfit that delivers 3-10 boards or with a company that throws you entire panels, PCBs for these things are cheap as long as you can avoid 4-layer+.
Components are also not the issue - plenty of reputable suppliers with good prices on components qty 10+, and given that most of these components (certainly the passives) get used multiple times, you can generally just buy in bulk.
Pick-and-place is where most of the cost tends to be.  If you're doing small batch, you might as well order a stencil to go with the PCB, pull up a few chairs, and make a mini assembly line with a buddy.  Cost is equal to whatever you value your time, but since most of these projects tend to be in-house and generally on downtime, it's not even equal to billable hours.  The problem is that once you go beyond small batch, you really need to start looking at automated pick-and-place.  While there's possibly a few makerspaces/fablabs that have the equipment, they're going to be far and between - which means that often you'd end up outsourcing it to a third party, and those tend to not be cost-effective unless you're doing large runs.  So you're stuck in the limbo between cost-effective and time/sanity-effective.
It's a space that some people are trying to fill - e.g. SBA - but it's still something that you have to look at twice.

has any stickminer actually had voltage adjust built in?

Probably rhetorical, but nope.  Plenty of mods (pencil/replace resistors/etc), but nothing out-of-the-box.  The Avalon Nano does do built-in throttling based on temperature - and wiring something across the pot to effectively get the same effect would be something I'd play with on the Compac

We really are doing it because it's awesome and fun, not because we want to get rich.

FUN > ROI?

Quick Q: I know it's early days, but are you targeting specific dimensions/outline?

[Bicknellski]

Yep, fabricating in the USA. If we can make it for less than $20 we'll probably drop the price. If we can't sell it for the price we're asking, we won't make 'em. Simple as that. The primary reason we're moving forward on making them is because plenty of people have stated they want them, and from the beginning we've said we were shooting for a maximum price of $20. I could make it fifty cents cheaper if I wanted to use a low-end fixed-voltage regulator or a tiny boring heatsink but the point is to make it an awesome learning tool instead of a boring toy. You gotta consider that the per-chip complexity is an order of magnitude greater than for a full miner, given that a full miner will have one VRM and one UART bridge and 18 chips, where this has one VRM and one UART bridge and one chip.

What do you consider a "respectable margin" on a product like this?

If it costs you 20$ to make it I think you are being very generous to the community.

Gotta have some return on your design time etc. got no doubt at 20$ they will sell out.

[jekecoin]

Yep, fabricating in the USA. If we can make it for less than $20 we'll probably drop the price. If we can't sell it for the price we're asking, we won't make 'em. Simple as that. The primary reason we're moving forward on making them is because plenty of people have stated they want them, and from the beginning we've said we were shooting for a maximum price of $20. I could make it fifty cents cheaper if I wanted to use a low-end fixed-voltage regulator or a tiny boring heatsink but the point is to make it an awesome learning tool instead of a boring toy. You gotta consider that the per-chip complexity is an order of magnitude greater than for a full miner, given that a full miner will have one VRM and one UART bridge and 18 chips, where this has one VRM and one UART bridge and one chip.

What do you consider a "respectable margin" on a product like this?

If it costs you 20$ to make it I think you are being very generous to the community.

Gotta have some return on your design time etc. got no doubt at 20$ they will sell out.

WOW 20 $ ? tell me about how much to send a pair to Europe

[sidehack]

So far we've done all our assembly in-house, including the surface-mount parts. That is to say, I am our pick-and-place machine. We're saving up for a real one and hopefully we'll have it ready to roll by the time we get to assembling stickminers, but the parts count for the stickminer probably won't be more than our DPS2K board which means if I had nothing else to do I could probably still get most of them assembled in about a week. I really hope we have a pick-and-place.

Something we've talked about doing in the code (at least for the TypeZero) is an "auto-tune" feature where you set the voltage and an initial clock speed and it'll adjust the clock up or down until it's hit a stability threshold. I believe Avalon4.1 has a setting like that, and I've seen some for a Gridseed. It's up to Novak whether that gets implemented at all, and if it's set for the stickminer. I don't believe we'll be able to measure temperature on the Compac so throttling based on that isn't really a stock-circuit option.

I don't have dimensions for the Compac yet, but with the buck circuit designed I can start on an intial PCB to get an idea for it. We're shooting for about 1" wide, not sure on the height yet. The current plan is to keep all parts on the same side, and use a top-mounted heatsink that covers most of the topside of the board. It'll have to be milled out on the sides to allow clearance for parts that aren't the ASIC, and probably three screws in a triangle to hold it down because something stability something three noncolinear points to define a plane something something. The PCB will be a test design based entirely on assumption until we get the breakout boards back and test them, and I get the buck circuit prototyped, but if both of those things work as expected we can immediately send off for stickminer PCB prototypes and test the real thing. I'm gonna try to keep the final board design the same dimensions as the test design even if it requires revision, so I can get the dimensions to a heatsink manufacturer early on and get samples (if at all possible) to play with and approve for full manufacture.

Once we have a few working prototypes I'll probably send some out for review. I've talked to MrTeal and Philipma1957 about hardware testing, might send samples to a couple other folks. The earliest we're likely to even have samples is about three weeks, since we probably won't have our prototype breakout boards until the end of this week at the earliest.

[philipma1957]

So far we've done all our assembly in-house, including the surface-mount parts. That is to say, I am our pick-and-place machine. We're saving up for a real one and hopefully we'll have it ready to roll by the time we get to assembling stickminers, but the parts count for the stickminer probably won't be more than our DPS2K board which means if I had nothing else to do I could probably still get most of them assembled in about a week. I really hope we have a pick-and-place.

Something we've talked about doing in the code (at least for the TypeZero) is an "auto-tune" feature where you set the voltage and an initial clock speed and it'll adjust the clock up or down until it's hit a stability threshold. I believe Avalon4.1 has a setting like that, and I've seen some for a Gridseed. It's up to Novak whether that gets implemented at all, and if it's set for the stickminer. I don't believe we'll be able to measure temperature on the Compac so throttling based on that isn't really a stock-circuit option.

Good work man keep it up.  I will want both styles of sticks and a few boards.
Let me know what I can do to help via pm.

[vapourminer]

great thread and project.

nice to see things done the right way.

sent ya a berger tip.