GekkoScience BM1384 Project Development Discussion

[philipma1957]

Right now they're drawing less than 3W (probably less than 2.7, really) at 8.25GH per stick. The heatsink gets toasty in still air (especially the already-warm air in the shop) but it's not danger-zone hot for sure. If you want to push them up past probably 10GH (estimated 3.1W or so) I'd recommend air. I've tested the sticks for starting up to 250MHz (~14GH, estimated 5.5W) but I didn't run them for endurance so I don't know how well the stock heatsink with airflow works up that high. Yet.

Proto PCBs have been ordered for a chips-only 18-board (basically, a passive TypeZero Spec1, does not have power or control on board) and for the revised (version 0.4) Compac PCB which implements the changes to the previous PCB that got these sticks working. Technically only the silver-heatsink one has all the changes, as it's the one with the deadbugged modified level shifter circuit. The rest are still using the old circuit, which has both more complexity and worse performance.

Also, to note, three of the five sticks there are using chips pulled from one of those donated dead S5 miners from a while back. The sixth stick (what with the hardware errors) also has a pulled chip, which might explain a bit.

I'll probably be sending some of these guys out to my first-wave testers, maybe tomorrow if they run overnight without issue. In a week or two when I have the sample heatsinks and fresh PCBs in hand I'll build what I hope is the final version and get them out. If we have good word from Bitmain by then about fetching chips, and the boards test out well (I'll be looking forward to a review thread, whether it's good news or bad news I like having honest feedback), maybe we can start taking orders. We'll have a pick-and-place and a nice new SMD oven up and going here soon so it shouldn't take forever+day to assemble these things.

If can image the sdcard and send a copy to me so

I can run it on a rasp pi.  I would copy sdcards for rasp pi's. offering a sdcard /hub/stick package for a new buyer may be a good selling idea.

A real plug and play for the masses.

[1714097726]

1714097726

[1714097726]

1714097726

[ManeBjorn]

I agree this would be a great idea.
+1

Right now they're drawing less than 3W (probably less than 2.7, really) at 8.25GH per stick. The heatsink gets toasty in still air (especially the already-warm air in the shop) but it's not danger-zone hot for sure. If you want to push them up past probably 10GH (estimated 3.1W or so) I'd recommend air. I've tested the sticks for starting up to 250MHz (~14GH, estimated 5.5W) but I didn't run them for endurance so I don't know how well the stock heatsink with airflow works up that high. Yet.

Proto PCBs have been ordered for a chips-only 18-board (basically, a passive TypeZero Spec1, does not have power or control on board) and for the revised (version 0.4) Compac PCB which implements the changes to the previous PCB that got these sticks working. Technically only the silver-heatsink one has all the changes, as it's the one with the deadbugged modified level shifter circuit. The rest are still using the old circuit, which has both more complexity and worse performance.

Also, to note, three of the five sticks there are using chips pulled from one of those donated dead S5 miners from a while back. The sixth stick (what with the hardware errors) also has a pulled chip, which might explain a bit.

I'll probably be sending some of these guys out to my first-wave testers, maybe tomorrow if they run overnight without issue. In a week or two when I have the sample heatsinks and fresh PCBs in hand I'll build what I hope is the final version and get them out. If we have good word from Bitmain by then about fetching chips, and the boards test out well (I'll be looking forward to a review thread, whether it's good news or bad news I like having honest feedback), maybe we can start taking orders. We'll have a pick-and-place and a nice new SMD oven up and going here soon so it shouldn't take forever+day to assemble these things.

If can image the sdcard and send a copy to me so

I can run it on a rasp pi.  I would copy sdcards for rasp pi's. offering a sdcard /hub/stick package for a new buyer may be a good selling idea.

A real plug and play for the masses.

[richardamullens]

Maybe if you push the raspberry pi aspect of the project it could gain traction among users of the Pi and Pi 2.

Certainly if you are able to start making the stick miners I would use a Pi.  the B+ and the Pi 2 use less power than the original machine.

I guess that is a stripped down USB hub you have there on the wooden block.

I hardly ever bother looking at any other threads on bitcointalk now

[philipma1957]

Maybe if you push the raspberry pi aspect of the project it could gain traction among users of the Pi and Pi 2.

Certainly if you are able to start making the stick miners I would use a Pi.  the B+ and the Pi 2 use less power than the original machine.

I guess that is a stripped down USB hub you have there on the wooden block.

I hardly ever bother looking at any other threads on bitcointalk now

I have a rasp pi b  older but not oldest model
I have a rasp pi 2 newest model.

The idea of 3-5 sticks on a hub to a rasp pi. power cost lets say 5 x 3.1 = 15.5 watts  the rasp pi + hub total watt amount = 25 watts
 5 sticks = 40gh  if software allows 5 pools  32gh can mine on slush or btcguld or f2pool  the other 8gh can solo mine.

gear is so  efficient it comes close to breaking even while giving you a shot at every block via the 1 stick on solo pool.

I love the idea of 100 usd to 125 usd out of pocket that pays itself off while allowing me a shot at each and every block.

or mine all your sticks at a 'real' and actually get to break even in 150 days to 180 days.

[sidehack]

Right now it's a stock Minera 0.5.0 image (or whatever's the version number for the current build) running stock cgminer 4.9.1 and enumerating the sticks as U3. That's suboptimal, of course, but it works well enough for a single chip. The only real annoyance is the stock U3 driver has a limited selection of frequencies to choose from, and not much granularity in the low end where the stick is likely to operate. Oh yeah and the device reset timeout is too low because it's expecting five times the hashrate, but that doesn't have much effect on performance.

For now I'm just gonna hand out Compacs and let people test with stock cgminer as U3. By the time we get to Amita work, hopefully Novak has some sort of driver figured out to get two chips working, but since I won't have an Amita PCB to test until after the final Compac PCB is tested we do have several weeks to twiddle the bits, and the two-chip board on the USB-UART adapter we've had running for a while now is a pretty accurate testbench to code with.

Yeah, that old hub was stripped down and beefed up for overclocked Block Erupters back in the day. I've got extra capacitors on it and starred power wires dedicated per socket. I trust it up to at least 5A, possibly more if I use the Molex instead of the pin header power.

Also, regarding price, I need to go back over my BOM and make some changes to reflect the new design. Odds are it won't change much, since some of the parts I'm knocking off should pretty well offset the parts I have to add.

Thing about doing that stuff you just said, Phil, is that for the price of five stick miners and a hub you could probably fetch a Spec1 board, basic heatsink and a fan and have something drawing four times the hashrate at a comparable efficiency. But it would require more power and probably be more loud.

[philipma1957]

Right now it's a stock Minera 0.5.0 image (or whatever's the version number for the current build) running stock cgminer 4.9.1 and enumerating the sticks as U3. That's suboptimal, of course, but it works well enough for a single chip. The only real annoyance is the stock U3 driver has a limited selection of frequencies to choose from, and not much granularity in the low end where the stick is likely to operate. Oh yeah and the device reset timeout is too low because it's expecting five times the hashrate, but that doesn't have much effect on performance.

For now I'm just gonna hand out Compacs and let people test with stock cgminer as U3. By the time we get to Amita work, hopefully Novak has some sort of driver figured out to get two chips working, but since I won't have an Amita PCB to test until after the final Compac PCB is tested we do have several weeks to twiddle the bits, and the two-chip board on the USB-UART adapter we've had running for a while now is a pretty accurate testbench to code with.

Yeah, that old hub was stripped down and beefed up for overclocked Block Erupters back in the day. I've got extra capacitors on it and starred power wires dedicated per socket. I trust it up to at least 5A, possibly more if I use the Molex instead of the pin header power.

Also, regarding price, I need to go back over my BOM and make some changes to reflect the new design. Odds are it won't change much, since some of the parts I'm knocking off should pretty well offset the parts I have to add.

Thing about doing that stuff you just said, Phil, is that for the price of five stick miners and a hub you could probably fetch a Spec1 board, basic heatsink and a fan and have something drawing four times the hashrate at a comparable efficiency. But it would require more power and probably be more loud.

The board is a nice idea.  but I mention the 4 stick real pool 5th stick solo as a way for small mining to stay around.

[sidehack]

We are intentionally designing the TypeZero board as capable of being a moderate size and power standalone miner, similar to the New R-Box. If all we cared about was the S1 Upgrade part of it we'd be making it a two-board direct replacement instead, but a ~100-150W efficient miner should be good for small. Sticks are probably easier to work with though, especially for beginners.

[philipma1957]

below the screen shots show how I am running s-3's in a 3 cent per kwatt power spot.    an s-3 in a 3 cent per kwatt is the same as your sticks in a 7.9 cent a kwatt spot.

I run 1 s-3 in balance and 2 s-3 in   failover.     so 8/9 of the hash is at btcguild and 1/9 of the hash is at solo.ckpool.org

your .31 watt gear can be set to run this way.  giving a small profit from the stock generic pool and a free shot at a full block.  this will sell to newcomers   since 3 sticks -5 sticks are low cost low power.

if I go a 1 board  like an r-box method  I need a gui with 5 pools and balanced load.  lets me run 80gh on 4 stock pools and 20 gh on the solo pool.  You need to be able to run most of the hash on the stocks pools so that the small solo part gives you the shot at a block but does not bust you out.  All this is more of promo and making widespread low cost mining a fun thing that does not break the bank.

[sidehack]

You should be able to get close to 300GH out of the Spec1 at 0.4W/GH, which isn't much worse than the sticks for efficiency. That's the goal anyways.

Can cgminer split up pools proportionally? The TypeZero boards are also still USB-connected so they'd presumably work with the same controller setup as sticks.

[quakefiend420]

Right now they're drawing less than 3W (probably less than 2.7, really) at 8.25GH per stick. The heatsink gets toasty in still air (especially the already-warm air in the shop) but it's not danger-zone hot for sure. If you want to push them up past probably 10GH (estimated 3.1W or so) I'd recommend air. I've tested the sticks for starting up to 250MHz (~14GH, estimated 5.5W) but I didn't run them for endurance so I don't know how well the stock heatsink with airflow works up that high. Yet.

Proto PCBs have been ordered for a chips-only 18-board (basically, a passive TypeZero Spec1, does not have power or control on board) and for the revised (version 0.4) Compac PCB which implements the changes to the previous PCB that got these sticks working. Technically only the silver-heatsink one has all the changes, as it's the one with the deadbugged modified level shifter circuit. The rest are still using the old circuit, which has both more complexity and worse performance.

Also, to note, three of the five sticks there are using chips pulled from one of those donated dead S5 miners from a while back. The sixth stick (what with the hardware errors) also has a pulled chip, which might explain a bit.

I'll probably be sending some of these guys out to my first-wave testers, maybe tomorrow if they run overnight without issue. In a week or two when I have the sample heatsinks and fresh PCBs in hand I'll build what I hope is the final version and get them out. If we have good word from Bitmain by then about fetching chips, and the boards test out well (I'll be looking forward to a review thread, whether it's good news or bad news I like having honest feedback), maybe we can start taking orders. We'll have a pick-and-place and a nice new SMD oven up and going here soon so it shouldn't take forever+day to assemble these things.

Nice man, glad to see this coming to fruition.  I'd be happy to help test if you need another person, I just shipped off the last of my SP20s today and need something fun to tinker with  The 4 S3s just sit there and work...and work....and work...and work...and work........

[aarons6]

below the screen shots show how I am running s-3's in a 3 cent per kwatt power spot.    an s-3 in a 3 cent per kwatt is the same as your sticks in a 7.9 cent a kwatt spot.

I run 1 s-3 in balance and 2 s-3 in   failover.     so 8/9 of the hash is at btcguild and 1/9 of the hash is at solo.ckpool.org

your .31 watt gear can be set to run this way.  giving a small profit from the stock generic pool and a free shot at a full block.  this will sell to newcomers   since 3 sticks -5 sticks are low cost low power.

if I go a 1 board  like an r-box method  I need a gui with 5 pools and balanced load.  lets me run 80gh on 4 stock pools and 20 gh on the solo pool.  You need to be able to run most of the hash on the stocks pools so that the small solo part gives you the shot at a block but does not bust you out.  All this is more of promo and making widespread low cost mining a fun thing that does not break the bank.

i have almost the same setup as you do but one difference..
i have 1 s5 on kanos pool and 3 oc'd s3s on antpool, one on balance to ck.solo.

this gives me roughly the same 1.15T on both kano and antpool with about 225ghs on ck.solo
they are also set to mine on westhash if the pool pays more then .0125, i cant deny the extra btc.
 

[philipma1957]

You should be able to get close to 300GH out of the Spec1 at 0.4W/GH, which isn't much worse than the sticks for efficiency. That's the goal anyways.

Can cgminer split up pools proportionally? The TypeZero boards are also still USB-connected so they'd presumably work with the same controller setup as sticks.

well I got balanced to work on s-3's  gui but that is a tweaked version cgminer.

I got balanced to work on bfgminer I listed 6 pools and split up the bitfury usb sticks pretty easy.

I am sure cgminer can do multiple pools and run balanced.

It also did it on the sp20's gui but you had to ssh the command.

I know my goal will be to get the test stick to run balanced and spilt the hash onto:

 btcguild worker 1
 btcguild worker 2
 btcguild worker 3
 solo.ckpool.org

if that runs balanced    75% of the hash will go to real pools and 25 % to solo mining.

this is a decent mix brings in some sure coin and give me the chance at a block.

[ZenFr]

A 5 sticks kit seems to be a good idea : 4 for pool mining and 1 for solo :-).
A full kit with the rasp pi + USB  hub + system to download, why not ?
And what about the sells out of USA ?

[elrippo]

A 5 sticks kit seems to be a good idea : 4 for pool mining and 1 for solo :-).
A full kit with the rasp pi + USB  hub + system to download, why not ?
And what about the sells out of USA ?

This is going to get hunky. You need a decent service with decent pricing and international tracking, despite the taxes 

[ZenFr]

A 5 sticks kit seems to be a good idea : 4 for pool mining and 1 for solo :-).
A full kit with the rasp pi + USB  hub + system to download, why not ?
And what about the sells out of USA ?

This is going to get hunky. You need a decent service with decent pricing and international tracking, despite the taxes 

Yes : i'm affraid for non american :-).
We will see...

[cavaliersrus]

pretty cool seeing this working with minera as thats what i use for my gridseed's and with my pi and 3 gridseeds running scrypt only they run about 25-28 watts for the 3 5 chip miners on scrypt but sha256 takes there power way up that power is also with a tiny 15mm fan running on my pi.. if i put my 2 80mm fans on my gridseeds my power spikes up to 35-38 watts so seeing this kinda power from this project is very nice but then again im still a newbie at bitcoin / alt coin mining

[sidehack]

I'm dropping off some packages at the post office today, which are three of the stickminer prototypes going to my designated first-wave testers: Philipma1957, vs3 and CrazyGuy. I need to get an address from MrTeal and he'll be receiving the fourth (if he's still interested in helping out).

These guys are my first picks for testing, because between them they have a pretty good understanding of electronics, durable engineering, consumer use-case testing and marketplace conditions. I figure I can draw advice from their various knowledges to move things along where a second perspective and/or expert opinion is desired. What I'm thinking is if one of those guys (whoever gets there first, I reckon) can start a separate review/criticism thread to put up pictures, opinions, complaints and such.

I'll be getting ahold of my "top picks" for the second half of the testing bunch probably sometime today, and when I have another four guys lined up and then when I get the next revision PCBs assembled and tested, I'll be sending out a hopefully-final-version stick to all 8 for official review.

The reasons for this are manifold. In the first, as mentioned above I would like to have opinions and criticisms from people I trust to know what they are looking at regarding design and sellability. In the second, I want to make sure that there's enough solid information coming from trusted members that folks don't have to worry I'm gonna rip them off with vaporware. Third, it'd be nice to have a unified and fairly visible discussion with actual product photos and specs right at the start instead of tucked somewhere around page 50 of the thread.

[HerbPean]

Can't wait for Philipma1957 review ^_^

[valkir]

Great! Good pick for the first tester!

Hope to be on second round!  

[alh]

I'll throw my hat in the ring as a 2nd round tester.

Does it help that I am a Rolla graduate from last century? 

Congratulations as well. I hope that Bitmain is willing/able to deliver chips at a reasonable price and volume.