GekkoScience BM1384 Project Development Discussion

[VirosaGITS]

Oh hey look, one hundred pages. How 'bout that. Maybe I should actually get around to demonstrating progress beyond a stickminer, eh?

I'll take a couple of Zeroboard, or whatever you called them, please. Some cost effective pods would be interesting too. Have you decided if you're going to go with the simpler 4-5 chips pods or the S AntMiner replacement/upgrade boards first?

[1714204954]

1714204954

[1714204954]

1714204954

[1714204954]

1714204954

[1714204954]

1714204954

[sidehack]

My next step is prototyping an 8-chip small board, as a first test of power systems and integrated controls. All I have to work with is BM1384, unfortunately, which are noncompetitive and can't really be had in quantity anyway. If it's good enough I might try to work out an exchange thing where folks with dead S5 boards can trade them for working pods made with pull chips. Sort of a version of Technobit's Minion exchange for HEX4M boards, except of course for the part where everyone gets ripped off.

In case anyone's wondering, I just now thought up that exchange idea. It's entirely possible that I'll make exactly one board as an in-house test and then do nothing at all with the design, proceeding straight to a 30-chip board sized for an S1 instead. I like the idea of a nice adjustable CPU-cooler'd pod but it has to be economically feasible.

[VirosaGITS]

My next step is prototyping an 8-chip small board, as a first test of power systems and integrated controls. All I have to work with is BM1384, unfortunately, which are noncompetitive and can't really be had in quantity anyway. If it's good enough I might try to work out an exchange thing where folks with dead S5 boards can trade them for working pods made with pull chips. Sort of a version of Technobit's Minion exchange for HEX4M boards, except of course for the part where everyone gets ripped off.

In case anyone's wondering, I just now thought up that exchange idea. It's entirely possible that I'll make exactly one board as an in-house test and then do nothing at all with the design, proceeding straight to a 30-chip board sized for an S1 instead. I like the idea of a nice adjustable CPU-cooler'd pod but it has to be economically feasible.

I throw my money at the screen but nothing happen! Jk.

Well if i could easily start mining with those boards without having to buy stuff to go along with it, i could be interested in those boards. Basically efficiency is a secondary concern to me. GH/$ is more important, as long as its acceptably efficient.

I may be interested in putting money into this and maybe that dead PCB i mentioned before.

[sidehack]

The idea I've had for a pod was USB-tethered (like the U3 and basically every other pod) and could run off either a brick or 6-pin. 8 chips would mean it'd run under 100W at the top end (for around 180GH), and under 20W at the dead bottom. You'd have to scrounge up a CPU cooler, and either a PSU or a good brick. Assume efficiency is about 10% worse than Bitmain's BM1384 chart because of VRM losses and peripheral power use, but you also get full control which means the bottom-end efficiency is close to S7 stock (say 44GH at about 15W)

I can't guarantee price. Chip cost would be about half the electronics cost, so trading a dead board would knock off at least that much as a source of chips. The problem with a pod is you still have all the same controls requirements of a larger mining PCB, which means you still have a decently high BOM. Of course one S5 board would have chips for three pods, so a one-to-one trade of board for pod might be possible.

I'll worry about a functional prototype first and see what happens from there. Might put up a poll and see if enough people would get behind the idea.

[philipma1957]

My next step is prototyping an 8-chip small board, as a first test of power systems and integrated controls. All I have to work with is BM1384, unfortunately, which are noncompetitive and can't really be had in quantity anyway. If it's good enough I might try to work out an exchange thing where folks with dead S5 boards can trade them for working pods made with pull chips. Sort of a version of Technobit's Minion exchange for HEX4M boards, except of course for the part where everyone gets ripped off.

In case anyone's wondering, I just now thought up that exchange idea. It's entirely possible that I'll make exactly one board as an in-house test and then do nothing at all with the design, proceeding straight to a 30-chip board sized for an S1 instead. I like the idea of a nice adjustable CPU-cooler'd pod but it has to be economically feasible.

 Okay I have 1 s-7 in house and 1 s-7 on order  that makes for 6 boards.
I would send a board to you when you need it.
I would love to see you get some better chips.

[Meech]

Sigh......  I only have S2  boards to send if wanted.

[VirosaGITS]

My next step is prototyping an 8-chip small board, as a first test of power systems and integrated controls. All I have to work with is BM1384, unfortunately, which are noncompetitive and can't really be had in quantity anyway. If it's good enough I might try to work out an exchange thing where folks with dead S5 boards can trade them for working pods made with pull chips. Sort of a version of Technobit's Minion exchange for HEX4M boards, except of course for the part where everyone gets ripped off.

In case anyone's wondering, I just now thought up that exchange idea. It's entirely possible that I'll make exactly one board as an in-house test and then do nothing at all with the design, proceeding straight to a 30-chip board sized for an S1 instead. I like the idea of a nice adjustable CPU-cooler'd pod but it has to be economically feasible.

 Okay I have 1 s-7 in house and 1 s-7 on order  that makes for 6 boards.
I would send a board to you when you need it.
I would love to see you get some better chips.

I think he was planning to do this on S5 ships. If he could get his hand on S7 chips, that would be awesome, but i suspect they might be very hard for him to get ahold of in sufficient amounts to start producing batches of those hardware.

He'd probably be happy being able to mess around with them however (?)

[sidehack]

I'm still working on an official source for BM1385 and other current-gen chips. I'd rather not tear apart a $600 board that still works.

I also talked to Novak about the miner-for-trade thing. We looked over some basic design ideas, approximate pricing and I put up a poll (over herefor opinions and such. I took the afternoon to hack up a bit of a proof-of-concept using mostly S5 and BTCGarden parts. Tomorrow I'll clean it up a bit, try to get some efficiency estimates and start working on a proper PCB layout.

I'm also potentially changing my mind (again) about what I want to build TypeZero power around, so I need to take a day and look at some fresh parts. While waiting for prototype parts for the pod, I'll probably rip into prototyping TypeZero power. The pod will integrate all the controls and a lot of the software so we'll be a few steps ahead of the game already.

We're gathering funds to start the second round of Compac manufacture, which should go more smoothly than the first half. All the long-lead parts have already been acquired; we just need to get some of the various components in quantity and reload the robot. I'll probably reopen the sales queue (with an estimated 14-day lead time for the first orders) Friday or so.

[TheRealSteve]

Now that I've got my stock of Compacs (6 via-via + the 2 direct that I'll keep, bitshopper hasn't opened sales yet unfortunately), I finally updated the packaging stuff.  The booklet remains much the same, other than the change in pot direction and a few wording and URL changes.

The back now has a bit of text and puts the GekkoScience logo over the heatsink:


I'll send GekkoScience a box along with some other goodies (@sidehack: is the send-from address you've been using also the mailing address to use?).

If anybody wanted to use it as a basis for their own, the InkScape design files are up here (for as long as that URL is valid): http://www.filedropper.com/gekkosciencecompacpackaging-quartetsize
The box is a "quartet" size PP box, which is larger than a poker card size box.  The Compac won't fit (well) in most poker size boxes length-wise, so if you were to get boxes, just keep that in mind.

Included in the design files:

	front	getting started	clocking	LEDs, FAQ, specs	back opaque	back transparent	foam template

[sidehack]

That is darn sexy, sir.

Yes the return address is correct.

Also, with the V0.5 there is no longer a "danger zone" on the pot. If you circle past the top-end voltage it automatically drops back down to 550mV. All 360° of rotation are within the safe operating voltage of the ASIC.

[Mikestang]

The back now has a bit of text and puts the GekkoScience logo over the heatsink:

Well done, that looks really good.  Did you put it together for fun, or is it for resale?  Seems like it would add a bit to shipping costs, especially on multiple stick orders.

Does the case provide static discharge protection?  I guess that it would since plastic doesn't conduct, but I'm no EE.

[spazzdla]

just bought the GekkoScience USB Compac BTC Miner (8-16 GH/s @ 0.3w/GH). 

[TheRealSteve]

Also, with the V0.5 there is no longer a "danger zone" on the pot. If you circle past the top-end voltage it automatically drops back down to 550mV. All 360° of rotation are within the safe operating voltage of the ASIC.

Yeah, it's no longer 'dangerous', but it'll at least provide too low a voltage for higher hash rates.. might as well designate a "just don't turn it here" area

Well done, that looks really good.  Did you put it together for fun, or is it for resale?  Seems like it would add a bit to shipping costs, especially on multiple stick orders.

Thanks!  Bit of both, 4 were gifted away, 2 went to people who bought from me.  Haven't got any others lined up and I've got a bunch of travels ahead of me or I'd see about making more to have some in stock 'just in case' I'd order some more out of USA direct.  Most of the cost is probably in 'labor', but FUN > ROI and all that.

Does the case provide static discharge protection?  I guess that it would since plastic doesn't conduct, but I'm no EE.

Technically not - it's PP, but I'm reasonable certain it doesn't have any of the dissipative agents mixed in, so if you're very unlucky you might be able to build up a charge on it or yourself and poke at just the right place to fry something.  Testing that requires equipment we certainly don't have laying around   You'd have to try pretty hard, though - the way it's in the box you're most likely to grab it either by the USB plug, heat sink or any of the sides, which is all ground plane connected to All The Things™.  It's also the same materials combination used in the packaging for the Block Erupter USB, the Antminer U1 and U2 and the iMiner (though that one doesn't count, being enclosed itself).  I'm pretty confident none of those 6 are going to come back with tales of rubbing cats on their heads and zapping a something on there

[sidehack]

So today I spent a bit more time with a multi-chip board, a PCB I had worked up a while back but never actually got around to testing. Only problem with it is the topology isn't quite right so I'm having to jumper a few things, but I want it working to test some node-level stuff I'll need to know before I can work up a pod miner. Right now I have two chips on there, but the second chip doesn't initialize reliably. I wasn't sure if that was hardware or software so I grabbed an old two-chip test board and fired it up (UART adapter to stock cgminer) and it lit up both chips from the get-go so it must be a hardware issue. I hope to work on it more tomorrow, get at least four chips to light up which should tell me pretty much everything I need to know for the 8-chip layout I have planned. Maybe when it's working I'll put up some pictures.

The fun thing about using cgminer is apparently U3 code shares out work differently than these chips require because, while I can get it to initialize two chips off the U3 driver, the second chip returns the same shares as the first so while I see two chips' hashrate I only see one chip's WU and multiple errors per second because cgminer labeles duplicates as errors.

The pod survey is still open over here, where I'm looking at community/customer interest in a BM1384-based pod miner. Like the Compac, this will basically be considered a dev step that becomes a full product; it'll include all of the controls we'd put on a larger board and give a bit more data on topology and efficiency without having to build a whole big board.

[notlist3d]

So today I spent a bit more time with a multi-chip board, a PCB I had worked up a while back but never actually got around to testing. Only problem with it is the topology isn't quite right so I'm having to jumper a few things, but I want it working to test some node-level stuff I'll need to know before I can work up a pod miner. Right now I have two chips on there, but the second chip doesn't initialize reliably. I wasn't sure if that was hardware or software so I grabbed an old two-chip test board and fired it up (UART adapter to stock cgminer) and it lit up both chips from the get-go so it must be a hardware issue. I hope to work on it more tomorrow, get at least four chips to light up which should tell me pretty much everything I need to know for the 8-chip layout I have planned. Maybe when it's working I'll put up some pictures.

The fun thing about using cgminer is apparently U3 code shares out work differently than these chips require because, while I can get it to initialize two chips off the U3 driver, the second chip returns the same shares as the first so while I see two chips' hashrate I only see one chip's WU and multiple errors per second because cgminer labeles duplicates as errors.

The pod survey is still open over here, where I'm looking at community/customer interest in a BM1384-based pod miner. Like the Compac, this will basically be considered a dev step that becomes a full product; it'll include all of the controls we'd put on a larger board and give a bit more data on topology and efficiency without having to build a whole big board.

I think the survey results no matter what is take our money .  Only thing I hope with pods is possibly a limit of 1 per person at start.  I really want a chance of getting one and know it will be a limited amount.

But I am still willing to pre-order once you do decide to sale.  And I think a lot of others will to once you decide to start selling.  But I respect your decision on whatever you decide.

[sobe-it]

sidehack whats the heat dissipation on the chips? I have read somewhere a top to bottom percentage but most older s5s have them only on the top while others newer ones have little heat sinks on the back of the board too. Last night I was thinking about the s3 dual heat sink setup.

[VirosaGITS]

So today I spent a bit more time with a multi-chip board, a PCB I had worked up a while back but never actually got around to testing. Only problem with it is the topology isn't quite right so I'm having to jumper a few things, but I want it working to test some node-level stuff I'll need to know before I can work up a pod miner. Right now I have two chips on there, but the second chip doesn't initialize reliably. I wasn't sure if that was hardware or software so I grabbed an old two-chip test board and fired it up (UART adapter to stock cgminer) and it lit up both chips from the get-go so it must be a hardware issue. I hope to work on it more tomorrow, get at least four chips to light up which should tell me pretty much everything I need to know for the 8-chip layout I have planned. Maybe when it's working I'll put up some pictures.

The fun thing about using cgminer is apparently U3 code shares out work differently than these chips require because, while I can get it to initialize two chips off the U3 driver, the second chip returns the same shares as the first so while I see two chips' hashrate I only see one chip's WU and multiple errors per second because cgminer labeles duplicates as errors.

The pod survey is still open over here, where I'm looking at community/customer interest in a BM1384-based pod miner. Like the Compac, this will basically be considered a dev step that becomes a full product; it'll include all of the controls we'd put on a larger board and give a bit more data on topology and efficiency without having to build a whole big board.

I think the survey results no matter what is take our money .  Only thing I hope with pods is possibly a limit of 1 per person at start.  I really want a chance of getting one and know it will be a limited amount.

But I am still willing to pre-order once you do decide to sale.  And I think a lot of others will to once you decide to start selling.  But I respect your decision on whatever you decide.

I'd be interested but i can't do it with the expensive shipping if its only one unit. I could give a dead S5 board + money and that would be possibly enough for nearly 4 pods, which would then be doable.

[notlist3d]

So today I spent a bit more time with a multi-chip board, a PCB I had worked up a while back but never actually got around to testing. Only problem with it is the topology isn't quite right so I'm having to jumper a few things, but I want it working to test some node-level stuff I'll need to know before I can work up a pod miner. Right now I have two chips on there, but the second chip doesn't initialize reliably. I wasn't sure if that was hardware or software so I grabbed an old two-chip test board and fired it up (UART adapter to stock cgminer) and it lit up both chips from the get-go so it must be a hardware issue. I hope to work on it more tomorrow, get at least four chips to light up which should tell me pretty much everything I need to know for the 8-chip layout I have planned. Maybe when it's working I'll put up some pictures.

The fun thing about using cgminer is apparently U3 code shares out work differently than these chips require because, while I can get it to initialize two chips off the U3 driver, the second chip returns the same shares as the first so while I see two chips' hashrate I only see one chip's WU and multiple errors per second because cgminer labeles duplicates as errors.

The pod survey is still open over here, where I'm looking at community/customer interest in a BM1384-based pod miner. Like the Compac, this will basically be considered a dev step that becomes a full product; it'll include all of the controls we'd put on a larger board and give a bit more data on topology and efficiency without having to build a whole big board.

I think the survey results no matter what is take our money .  Only thing I hope with pods is possibly a limit of 1 per person at start.  I really want a chance of getting one and know it will be a limited amount.

But I am still willing to pre-order once you do decide to sale.  And I think a lot of others will to once you decide to start selling.  But I respect your decision on whatever you decide.

I'd be interested but i can't do it with the expensive shipping if its only one unit. I could give a dead S5 board + money and that would be possibly enough for nearly 4 pods, which would then be doable.

I think you missed the idea behind the dead board.  Re-read the topic here: https://bitcointalk.org/index.php?topic=1203190.0

Shipping 1x dead S5 board would be payment for 1 pod (I believe).  Sidehack can confirm If I'm reading it right.   But I do not think shipping in 1 dead board get's you 4 pods.   The other chips would be "recycled" and sold in other pods.  And you get one for the dead board.

[VirosaGITS]

So today I spent a bit more time with a multi-chip board, a PCB I had worked up a while back but never actually got around to testing. Only problem with it is the topology isn't quite right so I'm having to jumper a few things, but I want it working to test some node-level stuff I'll need to know before I can work up a pod miner. Right now I have two chips on there, but the second chip doesn't initialize reliably. I wasn't sure if that was hardware or software so I grabbed an old two-chip test board and fired it up (UART adapter to stock cgminer) and it lit up both chips from the get-go so it must be a hardware issue. I hope to work on it more tomorrow, get at least four chips to light up which should tell me pretty much everything I need to know for the 8-chip layout I have planned. Maybe when it's working I'll put up some pictures.

The fun thing about using cgminer is apparently U3 code shares out work differently than these chips require because, while I can get it to initialize two chips off the U3 driver, the second chip returns the same shares as the first so while I see two chips' hashrate I only see one chip's WU and multiple errors per second because cgminer labeles duplicates as errors.

The pod survey is still open over here, where I'm looking at community/customer interest in a BM1384-based pod miner. Like the Compac, this will basically be considered a dev step that becomes a full product; it'll include all of the controls we'd put on a larger board and give a bit more data on topology and efficiency without having to build a whole big board.

I think the survey results no matter what is take our money .  Only thing I hope with pods is possibly a limit of 1 per person at start.  I really want a chance of getting one and know it will be a limited amount.

But I am still willing to pre-order once you do decide to sale.  And I think a lot of others will to once you decide to start selling.  But I respect your decision on whatever you decide.

I'd be interested but i can't do it with the expensive shipping if its only one unit. I could give a dead S5 board + money and that would be possibly enough for nearly 4 pods, which would then be doable.

I think you missed the idea behind the dead board.  Re-read the topic here: https://bitcointalk.org/index.php?topic=1203190.0

Shipping 1x dead S5 board would be payment for 1 pod (I believe).  Sidehack can confirm If I'm reading it right.   But I do not think shipping in 1 dead board get's you 4 pods.   The other chips would be "recycled" and sold in other pods.  And you get one for the dead board.

I said one dead board, enough chips for 3 pods and a bit more assuming they would all be in working condition and to that add money to cover the rest. Since i am supplying my own chips, i'd be taking idealistically under 1 pod worth of supplies from sidehack, hereby leaving plenty to others.

So i'm pretty sure i did not miss the idea behind the dead board. The idea is he only has chips for 50 pods. If i supply him the parts, i do not see a problem.

[notlist3d]

So today I spent a bit more time with a multi-chip board, a PCB I had worked up a while back but never actually got around to testing. Only problem with it is the topology isn't quite right so I'm having to jumper a few things, but I want it working to test some node-level stuff I'll need to know before I can work up a pod miner. Right now I have two chips on there, but the second chip doesn't initialize reliably. I wasn't sure if that was hardware or software so I grabbed an old two-chip test board and fired it up (UART adapter to stock cgminer) and it lit up both chips from the get-go so it must be a hardware issue. I hope to work on it more tomorrow, get at least four chips to light up which should tell me pretty much everything I need to know for the 8-chip layout I have planned. Maybe when it's working I'll put up some pictures.

The fun thing about using cgminer is apparently U3 code shares out work differently than these chips require because, while I can get it to initialize two chips off the U3 driver, the second chip returns the same shares as the first so while I see two chips' hashrate I only see one chip's WU and multiple errors per second because cgminer labeles duplicates as errors.

The pod survey is still open over here, where I'm looking at community/customer interest in a BM1384-based pod miner. Like the Compac, this will basically be considered a dev step that becomes a full product; it'll include all of the controls we'd put on a larger board and give a bit more data on topology and efficiency without having to build a whole big board.

I think the survey results no matter what is take our money .  Only thing I hope with pods is possibly a limit of 1 per person at start.  I really want a chance of getting one and know it will be a limited amount.

But I am still willing to pre-order once you do decide to sale.  And I think a lot of others will to once you decide to start selling.  But I respect your decision on whatever you decide.

I'd be interested but i can't do it with the expensive shipping if its only one unit. I could give a dead S5 board + money and that would be possibly enough for nearly 4 pods, which would then be doable.

I think you missed the idea behind the dead board.  Re-read the topic here: https://bitcointalk.org/index.php?topic=1203190.0

Shipping 1x dead S5 board would be payment for 1 pod (I believe).  Sidehack can confirm If I'm reading it right.   But I do not think shipping in 1 dead board get's you 4 pods.   The other chips would be "recycled" and sold in other pods.  And you get one for the dead board.

I said one dead board, enough chips for 3 pods and a bit more assuming they would all be in working condition and to that add money to cover the rest. Since i am supplying my own chips, i'd be taking idealistically under 1 pod worth of supplies from sidehack, hereby leaving plenty to others.

So i'm pretty sure i did not miss the idea behind the dead board. The idea is he only has chips for 50 pods. If i supply him the parts, i do not see a problem.

But you are saying:

I'd be interested but i can't do it with the expensive shipping if its only one unit. I could give a dead S5 board + money and that would be possibly enough for nearly 4 pods, which would then be doable.

You said you could not do it for 1 pod.  So it sounded like you want multiple pods so it really does not supply chips to other pods if you get 3-4 made from it.  It would still be pretty limited. 

But this is taking away from thread.  Feel free to PM me.  I don't want sidehacks thread to contain debates that are really not needed.