Community Miner Design Discussion

[padrino]

I generally only support private initiatives however I have access and agreements in place for various chips... If you and sidehack already have something in play is there anything in particular you are looking for, or just a general conversation?

[1713569039]

1713569039

[1713569039]

1713569039

[sobe-it]

Very exciting project...thanks

Just a side note: Personally, i don't have S1, S3 or S5-had all three types, but sold them off as they became obsolete.
Buying used S3 or S5 again just to fit the new board?
Miner might have higher demand than boards, but I might be mistaken.

An s1 can be had on ebay for $25 possibly, I'm sure around here easily at that price. Buying an older miner to retrofit new boards will still be more cost effective than new gear. Also look at how much gear failed last few rounds from all the asic companies, this will be overbuilt with quality parts with the end user in mind and will be very flexable with underclock/undervolt and possibly overclock/overvolt. Look at how much an s7 was brand new, now imagine if you could have gotten it 2/3 of the price at the same time. Everything is calculated into a new miners price including having the asic built, sure this project will have some markup (why work for nothing) but nothing like batch 1 of any miner.

Watch the computer hardware thread, I guarantee when these boards hit that everyone who has extras will be selling their frames they have laying around.

[CanaryInTheMine]

I don't think the finances will make this feasible...  It's not possible to "compete" with a mass produced product once they hit the market.  so a small batch produced by a community isn't viable financially, plus someone still has to be the face of the project, legally, to enter into agreements with manufacturers etc...  Unfortunately the days of making your own miner as a project are over.  It's now more mainstream with several manufacturers.

[valkir]

I don't think the finances will make this feasible...  It's not possible to "compete" with a mass produced product once they hit the market.  so a small batch produced by a community isn't viable financially, plus someone still has to be the face of the project, legally, to enter into agreements with manufacturers etc...  Unfortunately the days of making your own miner as a project are over.  It's now more mainstream with several manufacturers.

Actually Sidehack already did some usb stick and everything was perfect. He now have a pick and place machine to be able to best faster.

Yeah its more hard to do it when you are not a big corporate but you can do a home miner 

I have faith in kilo17 and Sidehack

[kilo17]

I don't think the finances will make this feasible...  It's not possible to "compete" with a mass produced product once they hit the market.  so a small batch produced by a community isn't viable financially, plus someone still has to be the face of the project, legally, to enter into agreements with manufacturers etc...  Unfortunately the days of making your own miner as a project are over.  It's now more mainstream with several manufacturers.

Well the good news is the only person that would lose money is me.  I am not familiar with the WASP project but I can assume it didn't go well.
I think I need to define the term "Community Project" in the way I am using it.  The objective is to produce a small efficient miner that is affordable.  We aren't saying it will have a nice pretty shipping box and/or a nice pretty enclosure.  There are a lot of ways to keep cost low and that's what we are aiming to do.

Other important design issues I will leave to SideHack et al.  Imagine if you could underclock and undervolt S7's,  Avalon6 etc.  By adding things that the big wigs leave off increases the life span and value to the miner.  

I/We are NOT asking for any money or Bitcoin from anyone.  I think this thread can allow people to have a voice in the miner and make suggestions.

[dogie]

I am not familiar with the WASP project but I can assume it didn't go well.

Everyone lost out of WASP, everyone.

Other important design issues I will leave to SideHack et al.  Imagine if you could underclock and undervolt S7's,  Avalon6 etc.  By adding things that the big wigs leave off increases the life span and value to the miner. 

You can underlock S7s and Avalon6s, but why exactly would you want to undervolt them? Increasing life of hardware which is quickly becoming obsolete doesn't make much sense, especially when they may last 3-5 years anyway.

[padrino]

I am not familiar with the WASP project but I can assume it didn't go well.

Everyone lost out of WASP, everyone.

Other important design issues I will leave to SideHack et al.  Imagine if you could underclock and undervolt S7's,  Avalon6 etc.  By adding things that the big wigs leave off increases the life span and value to the miner. 

You can underlock S7s and Avalon6s, but why exactly would you want to undervolt them? Increasing life of hardware which is quickly becoming obsolete doesn't make much sense, especially when they may last 3-5 years anyway.

In the spirit of efficiency is how I read that.. The value of undervolting depends on the design of the chip and its performance charactistics..

Speaking of which, Bitmain for example made some noise about 9V for a while but it seems to have faded..

[Entropize]

I'll start by saying that this is really exciting stuff.

I get the direction of making boards that you can swap into existing hardware, it will work the way people are used to. It raises some questions though, the higher yield hardware all has built in management boards, why is that the norm? Raspberry Pi and cubie board and probably countless other development boards have i2c interface to manage hardware. Would it cut cost to make unmanaged boards that we could control with a pi and cgminer? I'm picturing a pi hooked up to 5 Frankenstein s7s running at double speed and efficiency and maybe slightly less heat.

[kilo17]

I am not familiar with the WASP project but I can assume it didn't go well.

Everyone lost out of WASP, everyone.

Other important design issues I will leave to SideHack et al.  Imagine if you could underclock and undervolt S7's,  Avalon6 etc.  By adding things that the big wigs leave off increases the life span and value to the miner.  

You can underlock S7s and Avalon6s, but why exactly would you want to undervolt them? Increasing life of hardware which is quickly becoming obsolete doesn't make much sense, especially when they may last 3-5 years anyway.

Underclock/Undervolt - I understand you can underclock both the Antminers and Avalon but you cannot do both.  As stated by Padrino, it is for efficiency.  The compac is way more efficient than any S5 as far as I know although they use the same ASIC.

In regards to the life span, I am not referring to how long the miner will actually work but the Lifespan as a Miner - if the efficiency can be increased by making adjustments then the lifespan is increased.

Other uses would be if Bitcoin shoots up to $900 then you can dial it up to a greater Hash and accept the decreased efficiency.

[philipma1957]

I am not familiar with the WASP project but I can assume it didn't go well.

Everyone lost out of WASP, everyone.

Other important design issues I will leave to SideHack et al.  Imagine if you could underclock and undervolt S7's,  Avalon6 etc.  By adding things that the big wigs leave off increases the life span and value to the miner.  

You can underlock S7s and Avalon6s, but why exactly would you want to undervolt them? Increasing life of hardware which is quickly becoming obsolete doesn't make much sense, especially when they may last 3-5 years anyway.

Underclock/Undervolt - I understand you can underclock both the Antminers and Avalon but you cannot do both.  As stated by Padrino, it is for efficiency.  The compac is way more efficient than any S5 as far as I know although they use the same ASIC.

In regards to the life span, I am not referring to how long the miner will actually work but the Lifespan as a Miner - if the efficiency can be increased by making adjustments then the lifespan is increased.

Other uses would be if Bitcoin shoots up to $900 then you can dial it up to a greater Hash and accept the decreased efficiency.

If you have good power prices better efficiency can let a miner earn a long time.

[sidehack]

A miner which can be volt-adjusted has a much longer viable lifetime. The BM1384 chips on the S5 could be run down to 600mV and 0.25J/GH versus the 800mV/0.45J/GH setting of the S5. If the machine was undervoltable it could run as efficient as an S7. Sure you'd only see 500GH at that setpoint, but that's not bad for something you paid for a year ago and it's probably already paid off. Look how long the SP20 was around - some are still around, because you could run it hot or run it cold. When did that come out, October 2014?

I don't like the idea of designing a board for a specific controller, because then either I have to make that controller or design around one that already exists (in which case you have to have one or get one). And then you limit the software to using exactly that controller, especially if you're using low-level IO. I'd much rather put the extra $2 in parts directly onto the board to make it more generic, able to connect and run off anything with USB and cgminer. If the software and hardware package is going to be generalized or in any way reusable, that is a mandated requirement. Why build a reusable framework that requires you to hook up to a Pi, when some folks would rather have something that hooks up to a Cubie or an old XP laptop? I can find no convincing reasons to settle on that arbitrary limitation when it's no more difficult to make it a lot more flexible. And then if you wanted, you could hub a dozen boards together and run them off a single controller instead of requiring a Pi for every 2, or 4, or whatever.

Heck, with that flexibility, the same controls framework for the TypeZero boards could be used on a pod or a stickminer as well.

[Gyrsur]

I'm all in.

[kilo17]

That sounds much better than my explanation -  

[Entropize]

A miner which can be volt-adjusted has a much longer viable lifetime. The BM1384 chips on the S5 could be run down to 600mV and 0.25J/GH versus the 800mV/0.45J/GH setting of the S5. If the machine was undervoltable it could run as efficient as an S7. Sure you'd only see 500GH at that setpoint, but that's not bad for something you paid for a year ago and it's probably already paid off. Look how long the SP20 was around - some are still around, because you could run it hot or run it cold. When did that come out, October 2014?

I don't like the idea of designing a board for a specific controller, because then either I have to make that controller or design around one that already exists (in which case you have to have one or get one). And then you limit the software to using exactly that controller, especially if you're using low-level IO. I'd much rather put the extra $2 in parts directly onto the board to make it more generic, able to connect and run off anything with USB and cgminer. If the software and hardware package is going to be generalized or in any way reusable, that is a mandated requirement. Why build a reusable framework that requires you to hook up to a Pi, when some folks would rather have something that hooks up to a Cubie or an old XP laptop? I can find no convincing reasons to settle on that arbitrary limitation when it's no more difficult to make it a lot more flexible. And then if you wanted, you could hub a dozen boards together and run them off a single controller instead of requiring a Pi for every 2, or 4, or whatever.

Heck, with that flexibility, the same controls framework for the TypeZero boards could be used on a pod or a stickminer as well.

That's exactly what I'm talking about though, sorry if I din't get the idea out right. I focused on the pi because I've got some laying around that i'd love to put to use. What I meant for the new hardware is that I like the idea of making high powered hardware usb manageable so you could just plug it into existing or cheap hardware and upgrade your farm. 

[johnny5]

A sidehack project?  Sweet!  /Auto-good
I'm definitely interested -  Best of luck with it.

[NotFuzzyWarm]

<snip> I focused on the pi because I've got some laying around that i'd love to put to use. What I meant for the new hardware is that I like the idea of making high powered hardware usb manageable so you could just plug it into existing or cheap hardware and upgrade your farm. 

You still get to use your PI's. The boards would plug into a USB hub or ports on a computer running CGminer, the computer connects to the 'net. Can be an old laptop, an el'cheapo fanless Atom(R), or a PI. Anything that will run CGminer and has USB and LAN ports.

[notlist3d]

Very exciting project...thanks

Just a side note: Personally, i don't have S1, S3 or S5-had all three types, but sold them off as they became obsolete.
Buying used S3 or S5 again just to fit the new board?
Miner might have higher demand than boards, but I might be mistaken.

I'm in the same boat for my electricity price I got rid of all older generation miners.  I felt it was best to sell and buy BTC vs mining when I did calculations I came out with more BTC this way.

So I would have to buy them for this.  And chances are as soon as you say X miner it fit's parts miners take a jump in price on that model.   It might not be possible but I would love to see a complete miner without using other miners gut's cases.  And that might be to hard of a feat when keeping prices low I do realize that.

But best of luck with project.  

[sidehack]

Well the good thing about using Something like S1 chassis is it'd fit S1, or S3, or S5. So prices might jump a bit but in a supply/demand market there'll be a heck of a lot of supply. And it also saves me having to contract out all the mechanical manufacture and, for a lot of people who still have those miners laying around, it saves a lot on shipping. I could probably send a dozen miners' worth of boards for the price of one complete miner. Maybe someone could arrange a big ol' bulk purchase of retired S1 from some farm's back room and do some complete unit resale.

It'd be nice to eventually build a complete unit, but for now I'll settle with tackling electronics and save the mechanicals for another time.

[Witrebel]

Very exciting project...thanks

Just a side note: Personally, i don't have S1, S3 or S5-had all three types, but sold them off as they became obsolete.
Buying used S3 or S5 again just to fit the new board?
Miner might have higher demand than boards, but I might be mistaken.

I'm in the same boat for my electricity price I got rid of all older generation miners.  I felt it was best to sell and buy BTC vs mining when I did calculations I came out with more BTC this way.

So I would have to buy them for this.  And chances are as soon as you say X miner it fit's parts miners take a jump in price on that model.   It might not be possible but I would love to see a complete miner without using other miners gut's cases.  And that might be to hard of a feat when keeping prices low I do realize that.

But best of luck with project.  

To your point, I wonder how feasible it would be to come up with a heatsink/chassis/fan solution that could handle 2 boards and come in around $20ish.  Certainly could be interesting.  I mean, we have a form factor already defined, so lead time might be a non issue.  I will investigate this on my own, but as this is (I hope) a community oriented project, perhaps someone else on the forum has better contacts than I. Anyone with an S1/S3/S5 wins because they can upgrade on the cheap.  Anyone without wins because they could buy this chassis kit on the cheap.  Sidehack/the dev group wins because the market for their product expands.  Home miners win because we have a cost effective alternative to the commercial products, and potentially those products have to lower their respective price points. Not to mention, whoever could facilitate a low cost chassis kit might stand to make a bit of coin for themselves in the process.

[2112]

What my basic plan is, is to use a basic USB-capable AVR microcontroller as a board-level brain. The chip Novak liked had a decent array of IO bus, including UART, SPI and I2C. We can interface to sensors and controls using I2C, use GPIO for fan speed, and interface to ASICs on UART or SPI.

I'm dying to know what part did you guys selected. The Atmel web site isn't of much help, it took me over an hour to find the difference between ATxmega256A3U and ATxmega256A3BU. Are you shooting for something like ATmega16U4 or ATmega8U2?