Topic: GekkoScience BM1384 Project Development Discussion - page 42. (Read 146713 times)

Also. I spent a bit of time talking with Novak about the most badass things to make miners shaped like. His suggestions were mostly aerospace-related (which makes sense), like the J58 engine housing from an SR-71, or a F1 engine from the Saturn V rocket. Unfortunately those things aren't exactly rectangular.

So we're gonna proceed with making something look like a big V8. I'm working on an idea for a supercharger with internal fans as the air intake for the topside heatsink. Probably go with a 4x8 string where each cylinder is a node of 4 chips. 400MHz expected 700GH at about 350W (plus fans and extra crap). We're thinking what extra crap to add now, gauges and lights and stuff. Not as a priority of course, TypeZero and whatnot would still be the focus. But a lot of the design of this thing would derive directly from TypeZero specs so, you know.

You truly deserve the "Hero Member" rank.

Same here! The BE200 was a breeze to work with, it just used too much power.

-a[g

I don't think you understand. What we intend to do is built it well enough that it won't break or fail, and then sell it such that you can actually turn a profit. If the only way to build it good enough to meet our standards is to make it so expensive it's not worth buying, we'd rather not build it at all. We'd rather not get sales than sell something we don't think is good enough to basically put on a lifetime warranty. If I've built it well enough, it would be the customer's fault it catches on fire - something about cranking it to maximum power, taking off the fan and then stuffing it under a couch cushion.

If I built something like the Dice I'd use a matrix of more manageable chips instead. Unless you wanted to pay me at least $500 a week and sponsor some new equipment for the dev, I ain't messing with Spondoolies chips.

The Avalon nano is easy to improve on, since it was built with their previous-gen chip. I expect them to have a next-gen chip in a month or two that I'll probably take a look at.

Also, the shipping yard is in China so we can't really just drive there. I'd rather have the machinery than the money back, because if I got the money back I'd just spend it on the machinery again and that would probably mean paying the $750 shipping fee again, which ain't gonna happen.

Well, you're right, we don't need the extra com ports, or multi power rails for redundancy/load balancing..
Its the customers fault that it catches fire, they turned the voltage up that 1mV to high

Yeah, fare enough, spondoolie i expected to be a lot of work to even get one chunk o' silicon to get running. just looking at pictures of the layout looks insane! looks to be 6/8 rails going into one chip? and then cooling the 100w odd chip.. i would like to see your run of the "bitcoin Dice," but the r&d would take some time

And avalon would be a fun little chip to play with, would be nice to see if you could improve over their "Avalon nano v2"

Well, that sucks some, get into a F100, go to the shipping yard, borrow their forklift, load up and go!
or say "Screw you Guy! my money back Now!" ether way someones head should roll over it :|

well show that scanner who the boss is! i wanna see this entertainment!

Why not both? One of our fallbacks is "when in doubt, overbuild". We don't want anything we build to ever fail, that wasn't because the owner is a complete idiot and/or ran over it with a truck. But it's also gotta be affordable. What that means is building something really well and then knocking down the profit margin to make up the difference.

Regarding other chip makers, I talked a bit to Avalon a few months ago when we were hoping to get chips from Bitmain but wanted a fallback. Spondoolies came to me asking about some of our ideas in about February, but their new chip was looking to be the same as their old chip (100W BGA calling for a ~200A multiphase buck per chip), which I refuse to work with because it makes literally every step of the design and manufacture process unnecessarily expensive, complex and inefficient. SFARDS new chip, in addition to having similar features as Spondoolies, is bearing an extra price because of the dual-mining benefit and it would be very difficult for me to care less than I already do about scrypt. If I retool to anything, it's going to be another strain of chip that runs at most 15W per so exotic/complex cooling and power systems aren't required. I mean really, take a look at the inside of a Spondoolies box, consider for a moment that our design philosophy is "Simple, Durable, Reliable" and ask how likely it is we'd work with something they make.

As for the pick-and-place, well that's becoming an adventure in ineptitude. We paid for it on June 8th with the expectation (and by expectation, I mean it's exactly what we were told) that it'd ship within three days. So it shipped about two weeks later. The shipper specifically requested our contact info so it would go with the shipment in case customs needed it. Apparently customs needed it, but it wasn't actually filed with the shipment at all. So it spent two weeks in customs holding in Ohio while we tried to get ahold of anyone capable of figuring out what the heck was going on. We finally got someone at DHL who knew what information was needed, and gave him that information, and was told he'd be back in contact with us by the next day 5PM with answers. So he got back to us seven days later with the news (we had already discovered) that the whole crate had been returned to the manufacturer a couple days after we gave them the information they needed to clear it with customs. But the manufacturer claims they don't have an import license, just an export, so they can't actually reclaim it from the shipper. So either the manufacturer and shipper need to agree to get it back to us, or it's apparently going to be deemed "abandoned". And we sure as hell aren't gonna pay for it to be shipped again, since it was their screwup that got it hung up in customs and the shipper's [lazy/incompetent customer service representative's] screwup that got it shipped back. So we're approaching five weeks after expected delivery of seven thousand dollars worth of machinery trapped in limbo. Which, by the way, represents probably 20-25% of the business' total income for the year to date. So I can say "we bought a pick-and place" but I can't actually say "we have a pick-and-place" quite yet.


Also, AJRGale, I got an engine-shaped miner sketch for you but apparently the scanner doesn't like me.

well, i was going overboard with it, i've never seen watercooling for longevity, even engines needs a good flush and waterpump change after 100,000kms (or 5 years of driving). so simple watercooling for electric components, running a waterpump and fluids for 24/7, just seems like a job to take care of every 3 months.

but hey, you know you want to go overboard


Joys of being small and flexible dev team.
Large companies would look at what you're doing and may copy, if not they'll just ignore you. im almost tempted to say there are some people with many bitcoins and their thumb stuck up their butts wouldn't even care, they would be content with a warehouse full of dense miners.. but im not in the best mind frame to say that...


all i can say is i'm content with my 160GH/s pulling 170W and only getting 0.0003btc a day, so im not the type to whinge about getting my return investment. the way i see it, don't skimp out on making a miner cheap so people RoI, make it so it doesn't catch fire after 3 months of use.

laugh at me with this, but the U3 is built better then the NRB, for starters, it has 3 other ways to talk to the chips if something goes wrong. (rs232, i2c, and direct chip coms).

i do have a question, have you talked to the other 2 chip fabs? avalon and Guy/spondoolie? if all else fails, would it hurt to retool to one of them?
If not, im happy to start a Open source bitcoin ASIC core (i've basically watch a few hours of people explaining Verilog, that thats my grasp of hardware description language, all i could do is start the movement)


Well done getting the P&P+Oven, the sweet smell of cooking solder, why do you want to sleep though that?


that it is, but i think his PR department should be sacked, they suck!

holy hell, i opened this to reply at about 11pm, it is now 12:40am.. need more beer

Except for the watercooling part (and double-sided chips), the thing you just described is pretty similar to the thing I was thinking about last night.


With the exception of the large chip purchase (which, recall, was partially funded by community loans), the total budget for the project so far, which has produced a single-chip test board, a dual-chip test board capable of parallel or string, an 18-chip base string board (with no integrated power or control, but all node-level and inter-node requirements met) and prototypes of two revisions of the stickminer, has been less than $1000. We can work pretty lean, which we do because there ain't a lot of money.

That said, we don't have a lot of interest in getting money from a venture capitalist. Maybe if I already had a demonstrably working final product and only needed money for production, but didn't want to coordinate a group preorder that size (which is riskier) we'd talk about it.

I'm pretty sure I can finish out a BM1384-based functional 18-chip and 30-chip board (as planned), probably also with a functional 4-chip pod as a dev step on the way, for about another thousand bucks. If by the time I have that working someone's gotten us samples of a chip worth playing with, it shouldn't be difficult to retool for the new chip. If we go Bitmain, likely the protocols would be basically the same so the only requirement would be redoing footprints and routing, maybe adjusting power systems a bit. Avalon uses a similar primitive IO with a different control structure, which means new footprint, swapping the IO chip and redoing drivers. In either case it wouldn't be difficult. I'm willing to risk a thousand bucks and work time over the next month to finish the BM1384 project on the assumption that I'll be able to get new-gen chips and end up with the best miner money can buy, and then I'll be able to get someone with money to buy it.

One thing that's going to help people ride out the halving is getting a miner well beforehand that can run profitably at top clock, and can also run both profitably and longevitously at lower clocks afterward. A fixed-operating-point miner like the S5 couldn't do it. My TypeZero is designed specifically to meet that need. If it behaves how I think it'll behave, the top-end efficiency should be a few percent less than the best-case with a string miner, but the bottom-end efficiency should be a few percent better than a typical re-voltable miner, and that it can make all those adjustments in software isn't terribly common. Spondoolies does it, sure, but I'm pretty sure my boards will do it better - more efficiently and a whole lot cheaper.

Being able to build a good chip would be pretty great, but we've zero VHDL experience and don't know anyone capable. If the community, or someone friendly to the community, can pull off a chip I'm more than willing to help wherever I can.

Margins on the Compac aren't amazing but if we end up selling all 1000 sticks there'll be more money available than is required to finish out the project. One thing that helps cut costs is "vertical integration" - we were able to scrape up enough for a pick-and-place and decent oven so we'll be doing all the assembly at almost zero cost instead of hiring out at probably several dollars per stick. It also helps timelines because we can work 20-hour days to get stuff done if we need to (the week I fetched 31 Tubes for hosting customers I worked 110 hours and slept about 20 in six days, and Novak has many similar stories) and most hirees wouldn't be willing to do stuff like that for free.

I think chips in September and having a product by Christmas is certainly doable.




Also - Bitmain did respond to email overnight. No chips to sell at the moment, but they may let me know if anything comes up. I'm not terribly optimistic but that's certainly better than nothing.

i just had a baaaad idea, v8 shaped container, the daughter boards in between the head and blocks, daughter boards have dual sided chips, massive cables, for power/data, hid inside the "exhaust" tubes, using the block and heads to allow water, Yes, Water cooling over the chips, and a 120mm/140mm fan strapped to the same sized radiator.

At the back of the "v8" is control board with gauges/led light bars/ with throttle control etc

à la Conley Precision Engines

the 1st heatsink was from http://www.overclockersonline.net/?p=5000106
it was made back in 2001 for the old AthlonXP 1.7GHz, rated at 75W of heat.
 'prolly could makes some off old stock of AMD sinks and 45 degree cut the fins.

the next is the Scythe Grand Kama Cross
http://www.scythe-eu.com/en/products/cpu-cooler/grand-kama-cross.html
this thing is Huge! 177x140x137 mm / 6.97x5.51x5.39 in with a 140mm fan strapped to the top

The good news is they do have some chips already bought.  So hopefully they are very successful on selling those.  I personally would like to see GekkoScience prosper, and be able to do another product or line of products.

I do not know their profit margins but hopefully it allows for even more chips to be ordered next time.  If they can order more each product and grow, it would be great to see this.

Only thing weird is bitmain not even responding to them about chips.  That scares me a little about that partnership.

Prudent.

1) I believe you'll easily move 800-1000 stickminers globally.
2) I also believe that there is a substantial market for S1 replacements.
3) If it were me, I'd reserve a larger quantity (300-400) for multi-chip dev. As it's a skill set (multi-chip dev) that few others have.
4) Bitmain "promised" to release their new product line in July (I'll share the email if desired, think I already shared that with Novak in a PM), anyway, we're sneaking up on August . . . .
5) Since late May antpool's hash rate has gone from high 50's PH/a to 80+ Ph/s and last time I checked they're the second largest pool.
6) Our contacts in the 3M immersion cooling world have indicated to us that the Spondoolies work well immersed . . . How would they know that, unless it's been tried.
7) Getting a VC to put up 1/4 mill for outdated chips would be a very hard sell even with a brilliant board design.
Getting a VC to fund a competitive startup with an aggressive plan not so much. i.e. a company that designs hash chips, builds boards, and either hashes with those boards and/or sells some.
9) If the big 4 release new chips in Sept and whomever acquiesces to selling you chips, I'm betting you'll have a product in time for Christmas. Not pinging on you, just stuff takes time.

GekkoScience et. al., I can think of no other I would rather have designing/implementing boards, you are always at the fore front of our minds in that regard.

But designing/implementing a board without chip availability is a moot point, regardless of the targeted consumer community.
It boils down to a simple supply chain management problem. A problem that I can imagine has multiple solutions.

I think it's safe to say that the BE300 is a dead horse, and I for one am gonna' quit kicking it.
Having said that, I think it's time to pickup the BE300 style torch and design a new chip.
How that would happen, exactly, I'm clueless, but would jump on the supportt/participation bandwagon in a heartbeat.
Just to be clear, I/We are not a Richie Rich and could not fund a 1/4 mill effort.
But we do have contacts that would consider that sum to be "pocket change".

Used to think the small hobbyist miner was a dying breed. I don't hold that opinion anymore.
I believe, at the latest, when the next block reward reduction happens most if not all the small to medium sized miners will go the way of the dodo. That includes us.
I think this time next year the panorama will be very polarized, the extremely large and the extremely small.
So positioning to manufacture and sell to the small has merit.
 

That's why so many of us love you thread and your work.  No matter what it is you seem to go for a challenge

I should be asleep instead of thinking about this stupid awesome miner.

Maybe I'll custom-machine a topside heatsink shaped like the top half of an engine, with a stock rectangle heatsink on the underside. The heads of the topside heatsink would have a pair of 60-72mm fans per side, running at partial speed (probably thermally throttled) with a turbo button that revs them up to full RPM. The back end has a pair of steel legs coming down from it like an engine stand, behind which is the voltage adjustment steering wheel and gauges for core voltage, VRM output current and temperature.

With that kind of setup, I kinda feel bad only running 8 chips. Maybe I should build 16 into it but call it the B8 anyway? 4x4 grid of chips and a doubly-badass 100A four-phase buck, pushing over 400GH at the super-overclocked top end?

And then I'll use it for 8 months before listing it for sale as new equipment.

Indeed.... i forgot to mention "Used for 8 months before being sold, because no-one believes in good ol' fashioned new devices sales"

Hehehe... ooh and the B8+ has a cast iron backplate 4mm thick just to add some weight

But... but none of that stuff is actually awesome. Also everything on there describes a product I'd never voluntarily sell to someone, but I think a few weirdos might actually be willing to pay for a real B8 Interceptor so if it's going to exist it needs to be a friggin' tank.


Okay, what if the board is actually mounted vertically with angled heatsinks on both sides (so it pulls top and bottom heat) shaped like halves of a V8? Maybe the turbo is heatpiped from a VRM cooler on one end. Tomorrow when I'm not heading to bed and have a decent mouse maybe I'll draw up some concept art. But it definitely needs a steering wheel and gauges. That's a must.

How about a purely vaporware B8 Interceptor... so vaporware that it only exists on renders!?

That's pretty badass. Product information link? If the footprint is big enough to house two rows of four chips it might be enough.

The copper heatpipes would look better in a dark stainless or something though. I wonder if I could plate them?

Gonna have to think about that a bit. It's definitely kinda engine-shaped, but what about the turbocharger?

What if the chips are on daughter boards attached to the main board at an angle, with a straight heatsink mounted on top and 4 small fans on it? Look kinda like cylinder heads. But that still doesn't solve the turbo. Hmm... Maybe I'll worry about figuring out the power systems and AJRGale, you can keep whittling away at the aesthetics. Making things not look like plain rectangles is not something I'm terribly good at.

i thought this..


but then i remembered this:

add 8 little fans across the top, 4 little blowers across the bottom side, aiming slightly down.
each heatpipe clamped down onto 2 chips across the banks.

Alright, how about this. A badass dual-phase 50A VRM pushing power for a 2x4 string, top and bottom heatsinks each with color-coordinated dual ducted fans, pushing the chips to 500MHz (estimated 16W each) and zero warranty because let's face it, you're going to blow it up. And a toggle switch underneath that you have to flip or it won't turn on. The voltage potentiometer will be a little steering wheel. Maybe throw some arbitrary analog gauges on there to monitor power draw and chip temp and stuff. It'll be obscenely expensive but ultimately sexy.

The B8 Interceptor. Such a stupid idea but I gotta admit, I'm starting to like it.