Why is it not possible to home build an ASIC?

[jamiemw]

Hi Guys

Has anyone tried or can anyone tell me why we cannot build an ASIC 'DIY' before BFL or anyone else does in a corporate capacity?

Jamie

[1714128252]

1714128252

[1714128252]

1714128252

[greyhawk]

There's nothing stopping you.

All you need is:
 - a facility to grow monocrystalline silicon into boules
 - a facility to cut boules into wafers
 - buy or reinvent the IP cores to etch onto the wafers
 - design the pathing and interface to the cores to etch onto the wafers
 - a facility to etch the wafers
 - a facility to bump the wafers
 - a facility to dice the wafers
 - a facility to package the dies
 - design a PCB
 - a facility to thermo flow solder the packaged die and electrical components onto the PCB

Let me know how it goes.

[fluffypony]

How are you planning on making your own integrated circuit? That's where the IC in ASIC comes from. Your question is akin to someone asking if it's not possible to home build a GPU or home build an Intel-compatible processor:)

[fluffypony]

There's nothing stopping you.

All you need is:
 - a facility to grow monocrystalline silicon into boules
 - a facility to cut boules into wafers
 - buy or reinvent the IP cores to etch onto the wafers
 - design the pathing and interface to the cores to etch onto the wafers
 - a facility to etch the wafers
 - a facility to bump the wafers
 - a facility to dice the wafers
 - a facility to package the dies
 - design a PCB
 - a facility to thermo flow solder the packaged die and electrical components onto the PCB

Let me know how it goes.

Don't forget the specialised knowledge required to figure out what circuitry would SHA256 more efficiently:-P

[greyhawk]

Don't forget the specialised knowledge required to figure out what circuitry would SHA256 more efficiently:-P

Yeah, well, BFL seems to cope without that, so....

[fluffypony]

Don't forget the specialised knowledge required to figure out what circuitry would SHA256 more efficiently:-P

Yeah, well, BFL seems to cope without that, so....

Dude come on, they're shipping "in about a week", give them some credit - I'm sure we can reasonably expect everyone's pre-orders to ship by the end of April...2019 :-P

[midnightcoder]

Anyone speculate on the amount of miniaturization of the ASIC miners? Like is there any relation to the <30nm processes that the big chip makers are cranking out.
 

[greyhawk]

Don't forget the specialised knowledge required to figure out what circuitry would SHA256 more efficiently:-P

Yeah, well, BFL seems to cope without that, so....

Dude come on, they're shipping "in about a week", give them some credit - I'm sure we can reasonably expect everyone's pre-orders to ship by the end of April...2019 :-P

Funny enough my prediction I made in early last December is still largely on track:

 - Early December: Finally understanding how Asian fabs work, BFL raises their order from 20000 chips to 100000 to bump up their place in the production queue.

 - Fab decides they finally have time for that weirdo BFL company to do their miniscule run on the 11th. This is under provision no big orders from LG come in in the mean time because boy that Nexus 4 is selling like hotcakes. If that happens BFLs run is gonna be pushed back to "whenever".

 - Assuming the 11th happens and the run actually starts, you can add about 4-5 weeks for production of these things.

 - On 21st of December the world ends

 - In the middle of January the chips are finished. Inaba decides to hop in a plane and fly over to the fab to hand receive the chips. Turns out, all planes crashed during the end of the world. Instead the chips are shipped over to 'merica by boat.

 - Middle of March the chips finally arrive in what is now called the Independent State of California. Customs officials hold what they assume to be weapons grade military devices until the war with Texas finally ends in a stalemate.

- 2015: ASICs arrive at BFL studios (a subsidiary of bAvalon International). No one knows what to do with them because all bitcoins were wiped out by the EMP pulse from Planet Rth'Nag'Ar.

- 2016: bAvalon strings the chips up on metal bars to produce a first-of-its-kind mathematics device they call The Abacus

[Larynth]

Actually, you could there are companies who sell FPGA and ASIC chips, they just won't be optimized for bitcoin mining. All you would need to do is find someone willing to design a board if there were enough interest, and someone else willing to write the firmware. I suspect there are people already doing so who were legit companies and didn't need to raise funding to get started and we just won't hear anything about it till they have a working mass-producible product ready to ship in 60 days.

[fluffypony]

Anyone speculate on the amount of miniaturization of the ASIC miners? Like is there any relation to the <30nm processes that the big chip makers are cranking out.
 

Yes - if you look at what little technical info Avalon has put out there (and BFL, although they haven't shipped anything) it would seem that there is plenty of room for more efficient chips. But I think we're at an apex where there is very little room for "new" miners, I think the mining space will be dominated by a handful ( < 40 000) of people around the world with what will be current gen ASICs who will go on to generate enough money to get the next generation and dump their old hardware on unsuspecting folk who will still imagine there's money to be made in mining.

[fluffypony]

Actually, you could there are companies who sell FPGA and ASIC chips, they just won't be optimized for bitcoin mining. All you would need to do is find someone willing to design a board if there were enough interest, and someone else willing to write the firmware. I suspect there are people already doing so who were legit companies and didn't need to raise funding to get started and we just won't hear anything about it till they have a working mass-producible product ready to ship in 60 days.

No offense, but I don't think you understand what an ASIC is. There is NO REASON for any company to invest a fortune in developing a SHA256 ASIC except for Bitcoin mining. ASIC boards used to speed up SSL on servers that handle thousands of concurrent HTTPS handshakes will typically focus on speeding up relevant ciphers like AES256 and ECDHE. There is no market for SHA256 ASICs except here.

Also - "find someone willing to design a board" - there are VERY few people around the world with the level of knowledge required to design an efficient IC for this sort of relatively complex operation.

[greyhawk]

Actually, you could there are companies who sell FPGA and ASIC chips, they just won't be optimized for bitcoin mining.

Dude, that's an oxymoron. An ASIC is always optimized towards its task. That's what the AS part means: application specific. In the case of bitcoin mining that would be double-hashing SHA256. The only ASICs  in existence dedicated towards that task are friedcats ASICMiners (and he doesn't sell) and the Avalons. And maybe that single one that's running (poorly) in a test board at BFLs.

[davidshrugged]

Doing a home made ASIC is not a trivial. ASIC chips for Bitcoin mining require to be designed for that specific purpose: mining bitcoins. Any particular ASIC like that one can only be manufactured and mass-produced by hundreds of thousands in order for it to make sense financially.

[teodor87]

There's nothing stopping you.

All you need is:
 - a facility to grow monocrystalline silicon into boules
 - a facility to cut boules into wafers
 - buy or reinvent the IP cores to etch onto the wafers
 - design the pathing and interface to the cores to etch onto the wafers
 - a facility to etch the wafers
 - a facility to bump the wafers
 - a facility to dice the wafers
 - a facility to package the dies
 - design a PCB
 - a facility to thermo flow solder the packaged die and electrical components onto the PCB

Let me know how it goes.

Actually you could buy wafers etched and doped.

What you need is design for the IC and layout for the PCB as well as very expensive equipment to manufacture, assemble and test those.

BTW: The whole manufacture is a "clean" process which means that you have to buy a clean room, which might cost up from 1000$ upto 10000$ per square meter. You need air purifiers, protective suits and adequate cooling and ventilation because one grain of dust could ruin a 1000$ wafer.

That's why.

[Ru7ru5]

Anyone is encouraged to design a free hardware ASIC?


Let's do it!!

[ReCat]

http://www.amazon.com/General-Purpose-Transistor-2N3904-NPN/dp/B004J5ZWEQ/ref=sr_1_1?ie=UTF8&qid=1365805741&sr=8-1&keywords=NPN+transistors

Buy about 50,000 of those, then get started by turning them into standard logic gates, and then design a hardware description, and then build it with your transistors. Simple and easy. Except it'll take you a few thousand years to construct it by hand.



On a serious note. Your best bet for making your own hardware miner is forgetting about ASIC and going with FPGA.

http://www.altera.com/education/univ/materials/boards/de2/unv-de2-board.html

It'll cost you 200-ish if you can convince them you're buying for educational purposes. Just get some existing FPGA miner and plonk it down.

[mmeijeri]

Doing a home made ASIC is not a trivial.

Not just not trivial, totally impossible. You can design an ASIC at home, but to produce it you need a multi billion dollar factory. If you bring enough dollars, and buy enough units, the companies owning those factories will happily produce them for you. That's what's happening with Avalon and BFL.

[Ru7ru5]

http://www.amazon.com/General-Purpose-Transistor-2N3904-NPN/dp/B004J5ZWEQ/ref=sr_1_1?ie=UTF8&qid=1365805741&sr=8-1&keywords=NPN+transistors

Buy about 50,000 of those, then get started by turning them into standard logic gates, and then design a hardware description, and then build it with your transistors. Simple and easy. Except it'll take you a few thousand years to construct it by hand.



On a serious note. Your best bet for making your own hardware miner is forgetting about ASIC and going with FPGA.

http://www.altera.com/education/univ/materials/boards/de2/unv-de2-board.html

It'll cost you 200-ish if you can convince them you're buying for educational purposes. Just get some existing FPGA miner and plonk it down.

Doing a home made ASIC is not a trivial.

Not just not trivial, totally impossible. You can design an ASIC at home, but to produce it you need a multi billion dollar factory. If you bring enough dollars, and buy enough units, the companies owning those factories will happily produce them for you. That's what's happening with Avalon and BFL.

Then I can't make a ASIC only buying some FPGAs and some wires... 

Well, let's buy some ASICs as a co-op, when we have enough money I'll call Foxconn to assembly in line production (and sell it only a bit cheaper than AVALON)

[ripple]

By the time most private people get their ASIC rigs the increase in 'difficulty level' from mining will be so high as to make most close to break-even.




 

[Hobsie]

The thing I find quite interesting about ASIC rigs on sale (or anything they'll give better Hash/s than anything currently out there) is that economically you'd want to mine (test) things yourself until difficulty is sufficiently high that selling it off becomes more economical.

It seems to me that you'd only sell it off when you're done with it O_o