Topic: Community Miner Design Discussion - page 26. (Read 34221 times)

Most of what I learned from the failed WASP project backed up my opinions about the evils of beaurocracy, and also how important it is to design a prototype that you can actually build and test. There will be no design-by-committee racking up hundreds of hours of conversation while one guy blows through the entire budget on a fleet of impossible layouts. There's just a couple guys with particular skills and an overarching design mentality of simplicity and reliability.

I guess this is kilo's nod that we're okay to talk about what's going on. I'm talking with NotFuzzyWarm and jstefanop about helping so we can get things done quickly and show off a prototype. The first model will probably be an Avalon A3218-based board, if we can figure out how to make 'em work. The plan is to design a power and controls structure that gets open-sourced, so it can be used on other designs. What Novak and I had started with the pods and such, it'll continue from that. 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. Packets sent to the controller would have two segments - one for controls and one for chip signals. The controls are things that would get implemented at the micro - reporting temperatures, adjusting voltages, stuff like that. The chip signals would get unpacked and relayed directly on the chip bus (whichever it is for that particular design). A cgminer driver would handle running the controls and encapsulating the chip work.
Alongside that, I'd really like to design a dual-phase buck regulator that'll handle at least 25A per phase and adjust within the range of about 5V to 9V with a conversion efficiency (off 12Vin) of at least 95% across a typical load range. Adjustment would be done digitally, with measurement alteration being handled by the aforementioned microcontroller.

If this turns out how I think it can, we'd have a basic yet fairly powerful framework that can be adjusted to different numbers (and types) of chips, with only minimal changes required to firmware and cgminer driver. Of course a different chip is going to require a different work protocol and stuff, but you'd have to put that in anyway. A lot of the USB packeting and such would be already done.

If possible, we'd like to secure Bitfury ASICs for a "final product" that would be compatible with S1/3/5 chassis. Looking at something 4TH/480W wall (stock), undervoltable to about 2.6TH/200W wall. I really want to be able to sell a two-board set for $600 but I don't know if that'll be possible (which is to say, more than cost but less than expected mining revenue).

A second goal, since we're trying to figure out Avalon stuff, is use the framework to build a beefy Novec board with A3218 and see how it handles.

Never again

Nor I for a BTC Community miner. If the failed WASP project teaches anything it is to stay focused on 1 thing at a time whenever possible.

Myself and several others are looking into designing a board, SideHack has mentioned an S1 replacement board and he is working on a board with me.  The design will be open source and we are wanting to bring it to the Home Miners here on the forum.

BTW - I personally have No interest in Scrypt Miners or Dual Miners

others  say the new 14/16 nm  will with the right power price but the bottom line is no one really knows for sure till it happen, right now it's all guessing and hoping. imo it won't be any diff then it is now except the older miners like s5 and down might become useless and we will get 12.5 less coins per block etc  and in four years it happens again.That's how i see it if I'm wrong then I'm wrong and I'll adjust .

Hey, i'm not sure of it but none of those hardware will be profitable once the halving is reached no? At least not with the current price!

I'm enjoying a silent basement but I had to build a gpu rig just to scratch "the itch".  Hopefully some of the new gear will quiet enough for in home use.

Don't forget the new script miner coming there gonna be pretty bad ass .  

They may not be direct bitcoin miners but are bitcoin miners non the less they just do it a different way and should be included in with any new hard ware coming .


he left out

Innosilicon they make both miners direct bit coin miners and in direct ones aka script miners. sense bw aka lk makes there own chips now .

Hey, that sounds like fun.

Competition for us would be great but rather depends on the pricing model the mfgr's are wanting to follow. It's a slippery slope between what market will bear (good profits) and Consumer electronics pricing with next to nil profits.

Once the various new chips come out and get any needed batch-2 tweaks it will only take 1 vendor to decide to sell a near loss to kill the others.

Reserved for future hardware/chip updates

I wanted to bring up the discussion of hardware that is out and coming out into one place.  In trying to follow some of the discussion about hardware it has become a little difficult.  I am hoping to generate some of the great conversations that tend to overflow from one thread to another (often times on the wrong companies thread)

I will try to get the discussion going with a few key points.

AVALON
Avalon6 - 3.5 Th/s retail price $900 at 0.29 watts/GH.  Recently with Batch 2 it is up to 3.8 Th/s

BITMAIN
Antminer S7 - Batch 11 - $720 for 4.73 Th/s at around .28-.31 watts/GH

BITFURY-
Bitfury has stated on the Bitfury discussion thread that a miner with Bitfury chips will be available to the public BEFORE April 1.  They are claiming 100GH/s at 0.1 J/GH and the chip goes up to 180GH/s with immersion cooling.  They have not made it apparent if those are different chip batches or PCB's.  

They only reference to price I have seen is from asiabtc for 10 Th at $2000 - old thread

BW - LK Group
Obviously mining with their new miners. Posting and updating the news frequently.  I have found 2 pieces of info intriguing:
1 - They have 2 chips:
          BW-LK1401 - 160 W/Th
          BW-LK1402 - 95 W/Th
          Link to info - https://www.bw.com/news/show-65
2 - They have also announced that hardware will be available to the public in April.

Spondoolies - SP50
KNC- Not Available to the Public
          

Hopefully it will also be a bit easier to get raw chips from one of them.

Edited to clarify - Trying to facilitate a discussion around building a community miner and chips available with design implications.  Things of importance are the design, the chips to use and the availability of the chips etc.

SideHack has mentioned in another thread (aka SideHack Thread ) about the design going on for a miner- I am the party he mentioned and currently I am in talks with a couple of different companies about chips.