BM1384 Pod Miner plus trade-in/recycling - an interest and feasibility poll

[sidehack]

It's very likely that I'll start this project. Even if I don't sell them, it's worth it to me to mine on these pods myself because, since most of the parts are scrounge, the $/GH is better than anything else I could buy and especially better when you consider the combination of efficiency and adjustability.

The real concern is how many I build, which is going to depend heavily on how many other people want to get it on it.

Yes, the deal would be you send S5 boards and I send you working pods. You'd have to pay shipping for the boards and the pods, but as mentioned before, if all shipping is within the US that'd be in the $15 neighborhood. I really wish I was able to ship internationally for cheaper. Even shipping to Canada costs $45 for something I can ship for $15 within the US, and it'd be $65 to anywhere overseas.

[VirosaGITS]

I'm not sure how a device you can get for free or close to free which will quietly generate 100GH from about 35W - which is more efficient than any miner (which I didn't build) that you can buy for less than $1500 - is already obsolete.

I honestly didn't expect to see as many people voting for the "I'd pay $50 for that" option, because as mentioned, the $/GH isn't so great. If I was buying chips straight from Bitmain it'd be even worse. At top clock it's about 0.25$/GH but that's also at 0.5W/GH where the S3 at 0.5W/GH is probably priced comparably but you get a bit more hashrate from a single unit - except it also has no future.

Anyone wanting to run 3KW should probably fill it with S3 or S5 or S7. Monolithic units with less per-unit overhead. As soon as I can (which is to say, as soon as someone has available chips worth using), I'll be building TypeZero boards that run up to 300W each and mount on the S1 chassis. Then I'll meet the needs of people who want to run 3KW or 30KW or whatever. This is more for people who want gifts, toys or lottery machines but want something bigger than stick miners. Honestly, if Bitmain keeps making things on the scale of the S4+ or S7, I really won't be competing in their domain. They're building big KW boxes and jet turbines and I prefer to specifically avoid that.

If I can get BM1385 chips, I'd probably port the 8-chip design straight across. It'd actually be easier to design with the new chips because Bitmain did a heck of a good job with the pinout and integrated features. You might not be able to push the chips quite as far as the 1384 design because the per-chip currents increase for the same power dissipation, so with a good cooler you'd run into power availability limits before you ran into power dissipation limits. But it'd still be a heck of a good little box.

chig, if you can get a board from Venezuela to the US for $5, make sure to send a shipping label because I guarantee I can't get get a board to you for that price. Maybe with the absolute cheapest slow-boat takes-a-month shipping option. Maybe.

They could make great room heater instead of using electricity cooling for this winter.

I said "obsolete" when i should say efficiency is already one generation behind. And next gen chips surpassing the S7 are supposedly already in existence, however not easily available. For example the nice SP chips we were talking about making a 400-600gh pod at under 100w. So in short, a pod using the BM1384 could fall 2 step behind before 2016.

Since the chips are available with the S5, you're competing at that efficiency but your pod would be a great value for people who want to control the voltage without using special PSU's or downsteps. And only the S5 v1.91 board seem to take the downvolt properly.

And for getting them in lots of 6, i don't really see a problem in term of setup. I already run more Antminer S1's than any other miners, so i thought of the pod as a "upgrade" for them. Albeit i suspect your pod wont run on wifi.

The builder himself designing them said 3000 watt filler is not what they are meant for on target audience.   For one reason if there are a limited amount of these selling 20 to one person chances are is a lot of them.    And you really do not want to mine with 20 devices when you have an option for 1 or 2.

If your just trying to fill up 3k make your life easier and get a S5 or S7.   But these pods will be a ton of fun to play with I already want one.  I think I have an addiction.

The builder himself doesn't know my situation and what is my best target for ROI. Seeing how so many more people voted 50$, he probably did not imagine all the situations that people maybe be in. Or why people people are so interested.
Because at the price he suggest, This would be a much better option for me than getting a s7.(Which i just can't, the price is ridiculous) And is also better than buying more S5's which are insanely hard to find in Canada, at least at a reasonable price/proper seller.

For example the S7 is about the worse ROI of all my active options. Because the $/GH is just horrible. And more S5 is already what i'm doing but there are a few reasons why the pod is better;
-I could get all that i need from one deal and pay shipping once.
-Possibly better overclocking than S5
-Volt control (Awesome sauce)
-Lower heat density = less noise
-And i suspect easier management as i believe Sidehack said they would daisy chain connect to make it easy to run all the miners on one controller.

While the deciding factor is i can't run a S5 next to my head to me when i sleep,... while maybe 12 pods running at moderate freq might be. I need miners that will run in my room quiet.

If the amount is limited, then of course. I said "I'd fill 3kW with them" before knowing there could be so few, after all, he's going to have made over 1000 usb sticks.

But who knows maybe some people are sitting on dozens of dead board that would let sidehack make a few hundreds pods.

If that is not the case a single lot of 6 would be just as nice since i got quite a bit of S1 and old Scrypt miner to replace.

It's very likely that I'll start this project. Even if I don't sell them, it's worth it to me to mine on these pods myself because, since most of the parts are scrounge, the $/GH is better than anything else I could buy and especially better when you consider the combination of efficiency and adjustability.

The real concern is how many I build, which is going to depend heavily on how many other people want to get it on it.

Yes, the deal would be you send S5 boards and I send you working pods. You'd have to pay shipping for the boards and the pods, but as mentioned before, if all shipping is within the US that'd be in the $15 neighborhood. I really wish I was able to ship internationally for cheaper. Even shipping to Canada costs $45 for something I can ship for $15 within the US, and it'd be $65 to anywhere overseas.

A small pod should be 30$ with USPS. A lot of 6 should be 45-50$. From experience with all the stuff i bought from sellers in the US so far. Its bad for one pod, but okay for 5-6.
If you decide to keep them, then i could see value in you doing that until they become unprofitable and you need to sell them to others. At either point i'd be interested in buying them.

[chiguireitor]

Welp, the moment you say "go", i'll start hunting borked s5 boards 

[sidehack]

Obsolete -> one generation behind; yes I get it now. By pure efficiency it's pretty good compared to current stuff only by merit of its adjustability; if someone else makes an S7-caliber miner that's adjustable this guy starts looking worse.

I have talked to Spondoolies about their new chip. Spondoolies likes NDAs so I can't talk about anything that I know because NDA. I don't know anything about what Avalon or InnoSilicon are doing right now. But I'm not too worried about "2 step behind before 2016". That said, I'm very much looking forward to working with current-gen chips. This project is another step toward the end goal of making not-crappy miners for not-crappy people.

The pods won't be daisychainable. Each pod will have its own USB connection so you'd have to hub multiples together, but they could still all be run on a single controller - say, a Pi with cgminer. That controller could have a wifi adapter. I don't know.

1000 sticks requires 1000 chips. Chips are the limiting factor. I have 1000 chips budgeted for Compacs, and another 200 for whatever else, and then a bunch of pulls. For 1000 chips I could make 125 pods. The chips I have available after Compacs are over can make about 50 pods. If someone wants to scrounge up a berjillion S5 boards that'd be pretty smexy. With the six boards I should expect from survey responders that's about 16 more pods (not counting the 6 that would be given in trade for said boards) so it'd take a lot of scrounging to get a big batch.

[chiguireitor]

...

With the six boards I should expect from survey responders that's about 16 more pods (not counting the 6 that would be given in trade for said boards) so it'd take a lot of scrounging to get a big batch.

So, say... 1 board + $100 would net 3 pods?

Are the pods going to get external 12v? with standard PCI-e connectors? (that would be rad)

[VirosaGITS]

Obsolete -> one generation behind; yes I get it now. By pure efficiency it's pretty good compared to current stuff only by merit of its adjustability; if someone else makes an S7-caliber miner that's adjustable this guy starts looking worse.

I have talked to Spondoolies about their new chip. Spondoolies likes NDAs so I can't talk about anything that I know because NDA. I don't know anything about what Avalon or InnoSilicon are doing right now. But I'm not too worried about "2 step behind before 2016". That said, I'm very much looking forward to working with current-gen chips. This project is another step toward the end goal of making not-crappy miners for not-crappy people.

The pods won't be daisychainable. Each pod will have its own USB connection so you'd have to hub multiples together, but they could still all be run on a single controller - say, a Pi with cgminer. That controller could have a wifi adapter. I don't know.

1000 sticks requires 1000 chips. Chips are the limiting factor. I have 1000 chips budgeted for Compacs, and another 200 for whatever else, and then a bunch of pulls. For 1000 chips I could make 125 pods. The chips I have available after Compacs are over can make about 50 pods. If someone wants to scrounge up a berjillion S5 boards that'd be pretty smexy. With the six boards I should expect from survey responders that's about 16 more pods (not counting the 6 that would be given in trade for said boards) so it'd take a lot of scrounging to get a big batch.

Now i know the limit amount of BM1384 chips you have/had.
chiguireitor seem up to the challenge, maybe he can scrounge up that berjillion S5 boards ^_^"

For the non daisy, usb hub still work, its a small extra expense, but very manageable, especially since i won't be getting 100 of them apparently. Good idea for slapping a Wifi adapter on the pi if ever.

Cool for SP NDA stuff. I guess this mean at least there is a channel of communication open. Hopefully at one point you get to tell us you'll have a batch of their chips.

Maybe we can setup something for the dead board. For instance i wouldn't mind buying dead S5 board and have them shipped directly to you.No point in having them shipped to me, then i reship them to you, especially since there's probably a bunch of them, dead already in the US.

[philipma1957]

Most of the network runs at .51 to .65 watts per gh. Yeah the s-7 is the best at .24 to .26 but it will take a long time for most of the network to be at .25 or less.

The future for home miners is effiecnt smaller gear pointed at a solo pool.  But I will never hit a block. So one hundred watts doing four hundred gh makes sense with twenty five  partners doing it . You have ten th mining in a team.


If your gear is really efficient and the club has good size you actually make money.

The stick club I am running is just a bit too small we do two to nine th each day.

But the math is there to make the idea work

[sidehack]

Chig - "Power comes in through either a 12V brick or a 6-pin PCIe (both jacks would be standard)." First post, first paragraph. I figure why limit it to one or the other and potentially force the customer to have extra expense? At the same time, someone with only a brick is limited to the brick's output power so full hashrate wouldn't be available.

VirosaGITS - I don't really care where the dead boards come from. You provide a dead board, you get a pod. I mean there's practical limits - if the board had Prisma'd and all the chips were charred or something I wouldn't take that in trade.

I probably shouldn't make the "you get a pod" guarantee since this whole project is still somewhere between "is it worth doing this" and "how do we make this happen" on the stages of project completion. So nobody move just yet.

I've been fairly annoyed by Spondoolies' engineering pretty much since I first learned about them. They really like to make things really complex, which ends up making things expensive and failure-prone (well, more potential failure points anyways, not necessarily actual failures). From a business standpoint, regarding customer service and general communication and not-being-a-dickery they've always done well. Better than any other mining manufacturer I've talked to. I don't know if we'll be able to make good use of their new chip, as there's more to consider in a design than just efficiency, but I'm not ruling it out yet.

[VirosaGITS]

VirosaGITS - I don't really care where the dead boards come from. You provide a dead board, you get a pod. I mean there's practical limits - if the board had Prisma'd and all the chips were charred or something I wouldn't take that in trade.

I probably shouldn't make the "you get a pod" guarantee since this whole project is still somewhere between "is it worth doing this" and "how do we make this happen" on the stages of project completion. So nobody move just yet.

I've been fairly annoyed by Spondoolies' engineering pretty much since I first learned about them. They really like to make things really complex, which ends up making things expensive and failure-prone (well, more potential failure points anyways, not necessarily actual failures). From a business standpoint, regarding customer service and general communication and not-being-a-dickery they've always done well. Better than any other mining manufacturer I've talked to. I don't know if we'll be able to make good use of their new chip, as there's more to consider in a design than just efficiency, but I'm not ruling it out yet.

Roger that. I will wait and when we get the go ahead, i'll see if getting some boards to you would be doable at that point.

Thats rather vague comments regarding the SP chips and i'm not really "in" that circle so, i glean that there would be engineering challenge with their chip, but nothing certain yet.

Meanwhile I haven't heard much about Avalon's or InnoSilicon's chips. And who know for how long Bitmain will keep their BM1385 to themselves. They aren't ultra efficient but then maybe it would be a more familiar chip to work with.

It would be a shame if SP does not work out since they seemed to be the only one announcing they would be inclined to sell batches of chips SoonTM afaik.

[sidehack]

Yeah, the Spondoolies chip has potential. You can tell that from the posted specs, it's promising. Without even considering priviliged information, look at everything they've built so far and note that it's always super complex and hard to work with. I didn't want to work with Rockerbox, just like I don't want to work with SFARDS, because BGA sucks. If I can help it I'd like to stay with smaller chips that don't make routing evil. SP's new chip package appears to be smaller and less evil than the Rockerbox at least. The BM1385 looks like it'd be super easy to work with. I really like the pinout and all the integration saves a lot of hassle and additional parts. It's quite a bit closer to a chip I'd have designed myself, as far as that goes, and I'd really enjoy working with it. InnoSilicon made a darn good chip in the A1 and I want to see what they can do with new, but when PlanetCrypto talked to them about samples all they were interested in was someone to lay down a million dollars on a full batch preorder, apparently sight-unseen. Nope. Avalon hasn't said a word to me since about June and I honestly have no idea what they're doing these days.

But anyways. Whoops I seem to have dropped this stats page screenshot of my 8-chip proof-of-concept hashing away at 200MHz on 650mV core.

[VirosaGITS]

Yeah, the Spondoolies chip has potential. You can tell that from the posted specs, it's promising. Without even considering priviliged information, look at everything they've built so far and note that it's always super complex and hard to work with. I didn't want to work with Rockerbox, just like I don't want to work with SFARDS, because BGA sucks. If I can help it I'd like to stay with smaller chips that don't make routing evil. SP's new chip package appears to be smaller and less evil than the Rockerbox at least. The BM1385 looks like it'd be super easy to work with. I really like the pinout and all the integration saves a lot of hassle and additional parts. It's quite a bit closer to a chip I'd have designed myself, as far as that goes, and I'd really enjoy working with it. InnoSilicon made a darn good chip in the A1 and I want to see what they can do with new, but when PlanetCrypto talked to them about samples all they were interested in was someone to lay down a million dollars on a full batch preorder, apparently sight-unseen. Nope. Avalon hasn't said a word to me since about June and I honestly have no idea what they're doing these days.

But anyways. Whoops I seem to have dropped this stats page screenshot of my 8-chip proof-of-concept hashing away at 200MHz on 650mV core.

Hopefully BM stop dicking around and sell their chips but, their communication skills in English seem very poor.

Is that a prototype board with BM1384 that you already have working? Sound to be too soon to have PCB of these set up already, are you running this without a PCB? I wonder how that would look like o.o;

[sidehack]

It's chunks of several different things all jerry-rigged together, and falls back on some of the work I did this spring and summer which led to Compac design. I don't have a PCB layout yet (I'll start on that tomorrow) and since we don't have multichip code yet it's running off an S5 controller. Mostly I wanted to prove the electronics - the basic topology and power systems - but there's definitely some improvements to be made yet. I don't expect to have a single-board prototype for at least two weeks. I need to do more testing before finishing the electronics enough to send off for a PCB, and there's still Novak's input on controls interfacing which needs to be done. I don't know how long firmware will take for full integration.

Sometime in the next few days I'll probably hook this up on a meter and get some efficiency curve estimates, but I have some stability improvements to make first.

[VirosaGITS]

It's chunks of several different things all jerry-rigged together, and falls back on some of the work I did this spring and summer which led to Compac design. I don't have a PCB layout yet (I'll start on that tomorrow) and since we don't have multichip code yet it's running off an S5 controller. Mostly I wanted to prove the electronics - the basic topology and power systems - but there's definitely some improvements to be made yet. I don't expect to have a single-board prototype for at least two weeks. I need to do more testing before finishing the electronics enough to send off for a PCB, and there's still Novak's input on controls interfacing which needs to be done. I don't know how long firmware will take for full integration.

Sometime in the next few days I'll probably hook this up on a meter and get some efficiency curve estimates, but I have some stability improvements to make first.

I'm impressed, this is pretty awesome. If you're up for it, i'd invite you to take some pictures sometimes from development stages. Seeing a jury rig of parts would be pretty cool to me. But if you want to keep your trade secrets, i'm sure people will understand

[sidehack]

If I weren't worried about trade secrets I'd have already posted pictures. Once I have a single-board prototype you better believe it'll be going up here though.

[AJRGale]

If he really shipped all of the boards he ends up pulling to you for scrapping the gold-silver then double-bag them in wall mart sacks and put them in duffel bags.
By the time they get to you, they will be already ground in powder from being kicked, thrown, and beaten dust by the shipping company.

you know, that'll make my life easer, I'd ball-mill them then wash the glass/epoxy away, leaving the metals behind..

Welp, the moment you say "go", i'll start hunting borked s5 boards  

... why is ebay aus so barren with any miner (anyone for a U3 for $115 excluding post?)? you'd half expect there would be some on there now.. so it'll be worth me getting some dead S5's and shipping it to side..

2xS5's +$25 nice deal make that 1 S5 for 1 pod.. but I has 4x non-communicating NRB, can i haz 1?
I'll endeavor to find some BM1384's or S5 boards to trade, that should help the project more

[sidehack]

To be honest, right now in trade negotiations I'd prioritize dead S5 boards much more highly than I would boards for power parts (NRB, S1 etc) because I currently have power parts for four times more pods than I have ASICs for, and that's a conservative estimate. Unless a lot of ASICs start arriving, it's entirely possible I could build every pod there's demand for and still have power parts left over from my own stash.

[quakefiend420]

Any use for an S4 board?

[QuintLeo]

Please do keep in mind that there are many "common" CPU cooler families around with different mounting hole requirements - Intel and AMD have not uses the same size mounting holes since the days of clip-on coolers (and the only heatsinks that used "holes" in those days were the Swiftech 370 and Alpha 8045 types), and even within Intel and AMD there have been a few different "families" of mounting holes. It's possible to design a PCB to support more than one set of mounting holes.

 I'd actually vote for the "clip-on" type cooler mounts, though you'd probably need one cooler for 1-2 chips if I remember the max power draw of the BM1384 correctly.

[Rabinovitch]

Please do keep in mind that there are many "common" CPU cooler families around with different mounting hole requirements - Intel and AMD have not uses the same size mounting holes

Please keep in mind also that almost every CPU cooler contains now a mounting accesoires for all existing sockets, so the different configuration of mounting holes should not be a problem (sure in case with "standart" configuration of those holes, i.e. 'as for LGA-1155' or as for any other (more convenient for developer in certain circumstances) socket).

[Searing]

It is "unlikely" to say the least you would want to 'modify' a 2013 KNC 550gh Jupiter (although you guys are certifiable as being crazy on your projects imho)

So when this 'madness' of a project takes off (hey looney bin...lots of folks lots of stories ..you get to 'act out" I'm in)

Would someone such as me ...be better off getting a S5 or whatever off Ebay for your modifications...or is there enough old stuff gonna flow into the
project ..someone like me ..who went the KNC route would not be left out? (unsure if i need to get my own 'junker' to get in this club/shop or not?)

but yeah ..if it is really about 50 bucks or so (hell could be more) I'm in....we could all maybe jump in and  pump another sha-256 coin huh? (lol)

(man I really really am gonna miss home mining.....when home mining dies ....I mean really we are at the point in time where it is out of desperation

"PIMP YOUr ASIC" (TM) (TV show rights can be negotiated)

(so low I've fallen so low...I thought I'd at least get a home miner that kinda/sorta worked up till 2017...alas I was a misled newbie back in the day)

Anyway I'm in